Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20211013' into staging

Use MO_128 for 16-byte atomic memory operations.
Add cpu_ld/st_mmu memory primitives.
Move helper_ld/st memory helpers out of tcg.h.
Canonicalize alignment flags in MemOp.

# gpg: Signature made Wed 13 Oct 2021 10:48:45 AM PDT
# gpg:                using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg:                issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [ultimate]

* remotes/rth/tags/pull-tcg-20211013:
  tcg: Canonicalize alignment flags in MemOp
  tcg: Move helper_*_mmu decls to tcg/tcg-ldst.h
  target/arm: Use cpu_*_mmu instead of helper_*_mmu
  target/sparc: Use cpu_*_mmu instead of helper_*_mmu
  target/s390x: Use cpu_*_mmu instead of helper_*_mmu
  target/mips: Use 8-byte memory ops for msa load/store
  target/mips: Use cpu_*_data_ra for msa load/store
  accel/tcg: Move cpu_atomic decls to exec/cpu_ldst.h
  accel/tcg: Add cpu_{ld,st}*_mmu interfaces
  target/hexagon: Implement cpu_mmu_index
  target/s390x: Use MO_128 for 16 byte atomics
  target/ppc: Use MO_128 for 16 byte atomics
  target/i386: Use MO_128 for 16 byte atomics
  target/arm: Use MO_128 for 16 byte atomics
  memory: Log access direction for invalid accesses

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

3 files changed, 572 insertions, 475 deletions

diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c
index 46140ccff3..b69a953447 100644
--- a/accel/tcg/cputlb.c
+++ b/accel/tcg/cputlb.c
@@ -39,6 +39,7 @@
 #ifdef CONFIG_PLUGIN
 #include "qemu/plugin-memory.h"
 #endif
+#include "tcg/tcg-ldst.h"
 
 /* DEBUG defines, enable DEBUG_TLB_LOG to log to the CPU_LOG_MMU target */
 /* #define DEBUG_TLB */
@@ -1840,6 +1841,25 @@ static void *atomic_mmu_lookup(CPUArchState *env, target_ulong addr,
 }
 
 /*
+ * Verify that we have passed the correct MemOp to the correct function.
+ *
+ * In the case of the helper_*_mmu functions, we will have done this by
+ * using the MemOp to look up the helper during code generation.
+ *
+ * In the case of the cpu_*_mmu functions, this is up to the caller.
+ * We could present one function to target code, and dispatch based on
+ * the MemOp, but so far we have worked hard to avoid an indirect function
+ * call along the memory path.
+ */
+static void validate_memop(MemOpIdx oi, MemOp expected)
+{
+#ifdef CONFIG_DEBUG_TCG
+    MemOp have = get_memop(oi) & (MO_SIZE | MO_BSWAP);
+    assert(have == expected);
+#endif
+}
+
+/*
  * Load Helpers
  *
  * We support two different access types. SOFTMMU_CODE_ACCESS is
@@ -1992,6 +2012,7 @@ load_helper(CPUArchState *env, target_ulong addr, MemOpIdx oi,
 static uint64_t full_ldub_mmu(CPUArchState *env, target_ulong addr,
                               MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_UB);
     return load_helper(env, addr, oi, retaddr, MO_UB, false, full_ldub_mmu);
 }
 
@@ -2004,6 +2025,7 @@ tcg_target_ulong helper_ret_ldub_mmu(CPUArchState *env, target_ulong addr,
 static uint64_t full_le_lduw_mmu(CPUArchState *env, target_ulong addr,
                                  MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_LEUW);
     return load_helper(env, addr, oi, retaddr, MO_LEUW, false,
                        full_le_lduw_mmu);
 }
@@ -2017,6 +2039,7 @@ tcg_target_ulong helper_le_lduw_mmu(CPUArchState *env, target_ulong addr,
 static uint64_t full_be_lduw_mmu(CPUArchState *env, target_ulong addr,
                                  MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_BEUW);
     return load_helper(env, addr, oi, retaddr, MO_BEUW, false,
                        full_be_lduw_mmu);
 }
@@ -2030,6 +2053,7 @@ tcg_target_ulong helper_be_lduw_mmu(CPUArchState *env, target_ulong addr,
 static uint64_t full_le_ldul_mmu(CPUArchState *env, target_ulong addr,
                                  MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_LEUL);
     return load_helper(env, addr, oi, retaddr, MO_LEUL, false,
                        full_le_ldul_mmu);
 }
@@ -2043,6 +2067,7 @@ tcg_target_ulong helper_le_ldul_mmu(CPUArchState *env, target_ulong addr,
 static uint64_t full_be_ldul_mmu(CPUArchState *env, target_ulong addr,
                                  MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_BEUL);
     return load_helper(env, addr, oi, retaddr, MO_BEUL, false,
                        full_be_ldul_mmu);
 }
@@ -2056,6 +2081,7 @@ tcg_target_ulong helper_be_ldul_mmu(CPUArchState *env, target_ulong addr,
 uint64_t helper_le_ldq_mmu(CPUArchState *env, target_ulong addr,
                            MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_LEQ);
     return load_helper(env, addr, oi, retaddr, MO_LEQ, false,
                        helper_le_ldq_mmu);
 }
@@ -2063,6 +2089,7 @@ uint64_t helper_le_ldq_mmu(CPUArchState *env, target_ulong addr,
 uint64_t helper_be_ldq_mmu(CPUArchState *env, target_ulong addr,
                            MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_BEQ);
     return load_helper(env, addr, oi, retaddr, MO_BEQ, false,
                        helper_be_ldq_mmu);
 }
@@ -2108,186 +2135,56 @@ tcg_target_ulong helper_be_ldsl_mmu(CPUArchState *env, target_ulong addr,
  */
 
 static inline uint64_t cpu_load_helper(CPUArchState *env, abi_ptr addr,
-                                       int mmu_idx, uintptr_t retaddr,
-                                       MemOp op, FullLoadHelper *full_load)
+                                       MemOpIdx oi, uintptr_t retaddr,
+                                       FullLoadHelper *full_load)
 {
-    MemOpIdx oi = make_memop_idx(op, mmu_idx);
     uint64_t ret;
 
     trace_guest_ld_before_exec(env_cpu(env), addr, oi);
-
     ret = full_load(env, addr, oi, retaddr);
-
     qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_R);
-
     return ret;
 }
 
-uint32_t cpu_ldub_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                            int mmu_idx, uintptr_t ra)
-{
-    return cpu_load_helper(env, addr, mmu_idx, ra, MO_UB, full_ldub_mmu);
-}
-
-int cpu_ldsb_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                       int mmu_idx, uintptr_t ra)
-{
-    return (int8_t)cpu_ldub_mmuidx_ra(env, addr, mmu_idx, ra);
-}
-
-uint32_t cpu_lduw_be_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                               int mmu_idx, uintptr_t ra)
-{
-    return cpu_load_helper(env, addr, mmu_idx, ra, MO_BEUW, full_be_lduw_mmu);
-}
-
-int cpu_ldsw_be_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                          int mmu_idx, uintptr_t ra)
-{
-    return (int16_t)cpu_lduw_be_mmuidx_ra(env, addr, mmu_idx, ra);
-}
-
-uint32_t cpu_ldl_be_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                              int mmu_idx, uintptr_t ra)
-{
-    return cpu_load_helper(env, addr, mmu_idx, ra, MO_BEUL, full_be_ldul_mmu);
-}
-
-uint64_t cpu_ldq_be_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                              int mmu_idx, uintptr_t ra)
-{
-    return cpu_load_helper(env, addr, mmu_idx, ra, MO_BEQ, helper_be_ldq_mmu);
-}
-
-uint32_t cpu_lduw_le_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                               int mmu_idx, uintptr_t ra)
-{
-    return cpu_load_helper(env, addr, mmu_idx, ra, MO_LEUW, full_le_lduw_mmu);
-}
-
-int cpu_ldsw_le_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                          int mmu_idx, uintptr_t ra)
-{
-    return (int16_t)cpu_lduw_le_mmuidx_ra(env, addr, mmu_idx, ra);
-}
-
-uint32_t cpu_ldl_le_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                              int mmu_idx, uintptr_t ra)
-{
-    return cpu_load_helper(env, addr, mmu_idx, ra, MO_LEUL, full_le_ldul_mmu);
-}
-
-uint64_t cpu_ldq_le_mmuidx_ra(CPUArchState *env, abi_ptr addr,
-                              int mmu_idx, uintptr_t ra)
+uint8_t cpu_ldb_mmu(CPUArchState *env, abi_ptr addr, MemOpIdx oi, uintptr_t ra)
 {
-    return cpu_load_helper(env, addr, mmu_idx, ra, MO_LEQ, helper_le_ldq_mmu);
+    return cpu_load_helper(env, addr, oi, ra, full_ldub_mmu);
 }
 
-uint32_t cpu_ldub_data_ra(CPUArchState *env, target_ulong ptr,
-                          uintptr_t retaddr)
+uint16_t cpu_ldw_be_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
-    return cpu_ldub_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
+    return cpu_load_helper(env, addr, oi, ra, full_be_lduw_mmu);
 }
 
-int cpu_ldsb_data_ra(CPUArchState *env, target_ulong ptr, uintptr_t retaddr)
+uint32_t cpu_ldl_be_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
-    return cpu_ldsb_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
+    return cpu_load_helper(env, addr, oi, ra, full_be_ldul_mmu);
 }
 
-uint32_t cpu_lduw_be_data_ra(CPUArchState *env, target_ulong ptr,
-                             uintptr_t retaddr)
+uint64_t cpu_ldq_be_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
-    return cpu_lduw_be_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
+    return cpu_load_helper(env, addr, oi, MO_BEQ, helper_be_ldq_mmu);
 }
 
-int cpu_ldsw_be_data_ra(CPUArchState *env, target_ulong ptr, uintptr_t retaddr)
+uint16_t cpu_ldw_le_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
-    return cpu_ldsw_be_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
+    return cpu_load_helper(env, addr, oi, ra, full_le_lduw_mmu);
 }
 
-uint32_t cpu_ldl_be_data_ra(CPUArchState *env, target_ulong ptr,
-                            uintptr_t retaddr)
+uint32_t cpu_ldl_le_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
-    return cpu_ldl_be_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
+    return cpu_load_helper(env, addr, oi, ra, full_le_ldul_mmu);
 }
 
-uint64_t cpu_ldq_be_data_ra(CPUArchState *env, target_ulong ptr,
-                            uintptr_t retaddr)
+uint64_t cpu_ldq_le_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
-    return cpu_ldq_be_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
-}
-
-uint32_t cpu_lduw_le_data_ra(CPUArchState *env, target_ulong ptr,
-                             uintptr_t retaddr)
-{
-    return cpu_lduw_le_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
-}
-
-int cpu_ldsw_le_data_ra(CPUArchState *env, target_ulong ptr, uintptr_t retaddr)
-{
-    return cpu_ldsw_le_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
-}
-
-uint32_t cpu_ldl_le_data_ra(CPUArchState *env, target_ulong ptr,
-                            uintptr_t retaddr)
-{
-    return cpu_ldl_le_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
-}
-
-uint64_t cpu_ldq_le_data_ra(CPUArchState *env, target_ulong ptr,
-                            uintptr_t retaddr)
-{
-    return cpu_ldq_le_mmuidx_ra(env, ptr, cpu_mmu_index(env, false), retaddr);
-}
-
-uint32_t cpu_ldub_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_ldub_data_ra(env, ptr, 0);
-}
-
-int cpu_ldsb_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_ldsb_data_ra(env, ptr, 0);
-}
-
-uint32_t cpu_lduw_be_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_lduw_be_data_ra(env, ptr, 0);
-}
-
-int cpu_ldsw_be_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_ldsw_be_data_ra(env, ptr, 0);
-}
-
-uint32_t cpu_ldl_be_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_ldl_be_data_ra(env, ptr, 0);
-}
-
-uint64_t cpu_ldq_be_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_ldq_be_data_ra(env, ptr, 0);
-}
-
-uint32_t cpu_lduw_le_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_lduw_le_data_ra(env, ptr, 0);
-}
-
-int cpu_ldsw_le_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_ldsw_le_data_ra(env, ptr, 0);
-}
-
-uint32_t cpu_ldl_le_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_ldl_le_data_ra(env, ptr, 0);
-}
-
-uint64_t cpu_ldq_le_data(CPUArchState *env, target_ulong ptr)
-{
-    return cpu_ldq_le_data_ra(env, ptr, 0);
+    return cpu_load_helper(env, addr, oi, ra, helper_le_ldq_mmu);
 }
 
 /*
@@ -2324,6 +2221,9 @@ store_memop(void *haddr, uint64_t val, MemOp op)
     }
 }
 
+static void full_stb_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
+                         MemOpIdx oi, uintptr_t retaddr);
+
 static void __attribute__((noinline))
 store_helper_unaligned(CPUArchState *env, target_ulong addr, uint64_t val,
                        uintptr_t retaddr, size_t size, uintptr_t mmu_idx,
@@ -2387,13 +2287,13 @@ store_helper_unaligned(CPUArchState *env, target_ulong addr, uint64_t val,
         for (i = 0; i < size; ++i) {
             /* Big-endian extract.  */
             uint8_t val8 = val >> (((size - 1) * 8) - (i * 8));
-            helper_ret_stb_mmu(env, addr + i, val8, oi, retaddr);
+            full_stb_mmu(env, addr + i, val8, oi, retaddr);
         }
     } else {
         for (i = 0; i < size; ++i) {
             /* Little-endian extract.  */
             uint8_t val8 = val >> (i * 8);
-            helper_ret_stb_mmu(env, addr + i, val8, oi, retaddr);
+            full_stb_mmu(env, addr + i, val8, oi, retaddr);
         }
     }
 }
@@ -2496,46 +2396,83 @@ store_helper(CPUArchState *env, target_ulong addr, uint64_t val,
     store_memop(haddr, val, op);
 }
 
-void __attribute__((noinline))
-helper_ret_stb_mmu(CPUArchState *env, target_ulong addr, uint8_t val,
-                   MemOpIdx oi, uintptr_t retaddr)
+static void __attribute__((noinline))
+full_stb_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
+             MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_UB);
     store_helper(env, addr, val, oi, retaddr, MO_UB);
 }
 
+void helper_ret_stb_mmu(CPUArchState *env, target_ulong addr, uint8_t val,
+                        MemOpIdx oi, uintptr_t retaddr)
+{
+    full_stb_mmu(env, addr, val, oi, retaddr);
+}
+
+static void full_le_stw_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
+                            MemOpIdx oi, uintptr_t retaddr)
+{
+    validate_memop(oi, MO_LEUW);
+    store_helper(env, addr, val, oi, retaddr, MO_LEUW);
+}
+
 void helper_le_stw_mmu(CPUArchState *env, target_ulong addr, uint16_t val,
                        MemOpIdx oi, uintptr_t retaddr)
 {
-    store_helper(env, addr, val, oi, retaddr, MO_LEUW);
+    full_le_stw_mmu(env, addr, val, oi, retaddr);
+}
+
+static void full_be_stw_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
+                            MemOpIdx oi, uintptr_t retaddr)
+{
+    validate_memop(oi, MO_BEUW);
+    store_helper(env, addr, val, oi, retaddr, MO_BEUW);
 }
 
 void helper_be_stw_mmu(CPUArchState *env, target_ulong addr, uint16_t val,
                        MemOpIdx oi, uintptr_t retaddr)
 {
-    store_helper(env, addr, val, oi, retaddr, MO_BEUW);
+    full_be_stw_mmu(env, addr, val, oi, retaddr);
+}
+
+static void full_le_stl_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
+                            MemOpIdx oi, uintptr_t retaddr)
+{
+    validate_memop(oi, MO_LEUL);
+    store_helper(env, addr, val, oi, retaddr, MO_LEUL);
 }
 
 void helper_le_stl_mmu(CPUArchState *env, target_ulong addr, uint32_t val,
                        MemOpIdx oi, uintptr_t retaddr)
 {
-    store_helper(env, addr, val, oi, retaddr, MO_LEUL);
+    full_le_stl_mmu(env, addr, val, oi, retaddr);
+}
+
+static void full_be_stl_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
+                            MemOpIdx oi, uintptr_t retaddr)
+{
+    validate_memop(oi, MO_BEUL);
+    store_helper(env, addr, val, oi, retaddr, MO_BEUL);
 }
 
 void helper_be_stl_mmu(CPUArchState *env, target_ulong addr, uint32_t val,
                        MemOpIdx oi, uintptr_t retaddr)
 {
-    store_helper(env, addr, val, oi, retaddr, MO_BEUL);
+    full_be_stl_mmu(env, addr, val, oi, retaddr);
 }
 
 void helper_le_stq_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
                        MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_LEQ);
     store_helper(env, addr, val, oi, retaddr, MO_LEQ);
 }
 
 void helper_be_stq_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
                        MemOpIdx oi, uintptr_t retaddr)
 {
+    validate_memop(oi, MO_BEQ);
     store_helper(env, addr, val, oi, retaddr, MO_BEQ);
 }
 
@@ -2543,137 +2480,61 @@ void helper_be_stq_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
  * Store Helpers for cpu_ldst.h
  */
 
-static inline void QEMU_ALWAYS_INLINE
-cpu_store_helper(CPUArchState *env, target_ulong addr, uint64_t val,
-                 int mmu_idx, uintptr_t retaddr, MemOp op)
-{
-    MemOpIdx oi = make_memop_idx(op, mmu_idx);
+typedef void FullStoreHelper(CPUArchState *env, target_ulong addr,
+                             uint64_t val, MemOpIdx oi, uintptr_t retaddr);
 
+static inline void cpu_store_helper(CPUArchState *env, target_ulong addr,
+                                    uint64_t val, MemOpIdx oi, uintptr_t ra,
+                                    FullStoreHelper *full_store)
+{
     trace_guest_st_before_exec(env_cpu(env), addr, oi);
-
-    store_helper(env, addr, val, oi, retaddr, op);
-
+    full_store(env, addr, val, oi, ra);
     qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_W);
 }
 
-void cpu_stb_mmuidx_ra(CPUArchState *env, target_ulong addr, uint32_t val,
-                       int mmu_idx, uintptr_t retaddr)
+void cpu_stb_mmu(CPUArchState *env, target_ulong addr, uint8_t val,
+                 MemOpIdx oi, uintptr_t retaddr)
 {
-    cpu_store_helper(env, addr, val, mmu_idx, retaddr, MO_UB);
+    cpu_store_helper(env, addr, val, oi, retaddr, full_stb_mmu);
 }
 
-void cpu_stw_be_mmuidx_ra(CPUArchState *env, target_ulong addr, uint32_t val,
-                          int mmu_idx, uintptr_t retaddr)
+void cpu_stw_be_mmu(CPUArchState *env, target_ulong addr, uint16_t val,
+                    MemOpIdx oi, uintptr_t retaddr)
 {
-    cpu_store_helper(env, addr, val, mmu_idx, retaddr, MO_BEUW);
+    cpu_store_helper(env, addr, val, oi, retaddr, full_be_stw_mmu);
 }
 
-void cpu_stl_be_mmuidx_ra(CPUArchState *env, target_ulong addr, uint32_t val,
-                          int mmu_idx, uintptr_t retaddr)
+void cpu_stl_be_mmu(CPUArchState *env, target_ulong addr, uint32_t val,
+                    MemOpIdx oi, uintptr_t retaddr)
 {
-    cpu_store_helper(env, addr, val, mmu_idx, retaddr, MO_BEUL);
+    cpu_store_helper(env, addr, val, oi, retaddr, full_be_stl_mmu);
 }
 
-void cpu_stq_be_mmuidx_ra(CPUArchState *env, target_ulong addr, uint64_t val,
-                          int mmu_idx, uintptr_t retaddr)
+void cpu_stq_be_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
+                    MemOpIdx oi, uintptr_t retaddr)
 {
-    cpu_store_helper(env, addr, val, mmu_idx, retaddr, MO_BEQ);
+    cpu_store_helper(env, addr, val, oi, retaddr, helper_be_stq_mmu);
 }
 
-void cpu_stw_le_mmuidx_ra(CPUArchState *env, target_ulong addr, uint32_t val,
-                          int mmu_idx, uintptr_t retaddr)
+void cpu_stw_le_mmu(CPUArchState *env, target_ulong addr, uint16_t val,
+                    MemOpIdx oi, uintptr_t retaddr)
 {
-    cpu_store_helper(env, addr, val, mmu_idx, retaddr, MO_LEUW);
+    cpu_store_helper(env, addr, val, oi, retaddr, full_le_stw_mmu);
 }
 
-void cpu_stl_le_mmuidx_ra(CPUArchState *env, target_ulong addr, uint32_t val,
-                          int mmu_idx, uintptr_t retaddr)
+void cpu_stl_le_mmu(CPUArchState *env, target_ulong addr, uint32_t val,
+                    MemOpIdx oi, uintptr_t retaddr)
 {
-    cpu_store_helper(env, addr, val, mmu_idx, retaddr, MO_LEUL);
+    cpu_store_helper(env, addr, val, oi, retaddr, full_le_stl_mmu);
 }
 
-void cpu_stq_le_mmuidx_ra(CPUArchState *env, target_ulong addr, uint64_t val,
-                          int mmu_idx, uintptr_t retaddr)
+void cpu_stq_le_mmu(CPUArchState *env, target_ulong addr, uint64_t val,
+                    MemOpIdx oi, uintptr_t retaddr)
 {
-    cpu_store_helper(env, addr, val, mmu_idx, retaddr, MO_LEQ);
+    cpu_store_helper(env, addr, val, oi, retaddr, helper_le_stq_mmu);
 }
 
-void cpu_stb_data_ra(CPUArchState *env, target_ulong ptr,
-                     uint32_t val, uintptr_t retaddr)
-{
-    cpu_stb_mmuidx_ra(env, ptr, val, cpu_mmu_index(env, false), retaddr);
-}
-
-void cpu_stw_be_data_ra(CPUArchState *env, target_ulong ptr,
-                        uint32_t val, uintptr_t retaddr)
-{
-    cpu_stw_be_mmuidx_ra(env, ptr, val, cpu_mmu_index(env, false), retaddr);
-}
-
-void cpu_stl_be_data_ra(CPUArchState *env, target_ulong ptr,
-                        uint32_t val, uintptr_t retaddr)
-{
-    cpu_stl_be_mmuidx_ra(env, ptr, val, cpu_mmu_index(env, false), retaddr);
-}
-
-void cpu_stq_be_data_ra(CPUArchState *env, target_ulong ptr,
-                        uint64_t val, uintptr_t retaddr)
-{
-    cpu_stq_be_mmuidx_ra(env, ptr, val, cpu_mmu_index(env, false), retaddr);
-}
-
-void cpu_stw_le_data_ra(CPUArchState *env, target_ulong ptr,
-                        uint32_t val, uintptr_t retaddr)
-{
-    cpu_stw_le_mmuidx_ra(env, ptr, val, cpu_mmu_index(env, false), retaddr);
-}
-
-void cpu_stl_le_data_ra(CPUArchState *env, target_ulong ptr,
-                        uint32_t val, uintptr_t retaddr)
-{
-    cpu_stl_le_mmuidx_ra(env, ptr, val, cpu_mmu_index(env, false), retaddr);
-}
-
-void cpu_stq_le_data_ra(CPUArchState *env, target_ulong ptr,
-                        uint64_t val, uintptr_t retaddr)
-{
-    cpu_stq_le_mmuidx_ra(env, ptr, val, cpu_mmu_index(env, false), retaddr);
-}
-
-void cpu_stb_data(CPUArchState *env, target_ulong ptr, uint32_t val)
-{
-    cpu_stb_data_ra(env, ptr, val, 0);
-}
-
-void cpu_stw_be_data(CPUArchState *env, target_ulong ptr, uint32_t val)
-{
-    cpu_stw_be_data_ra(env, ptr, val, 0);
-}
-
-void cpu_stl_be_data(CPUArchState *env, target_ulong ptr, uint32_t val)
-{
-    cpu_stl_be_data_ra(env, ptr, val, 0);
-}
-
-void cpu_stq_be_data(CPUArchState *env, target_ulong ptr, uint64_t val)
-{
-    cpu_stq_be_data_ra(env, ptr, val, 0);
-}
-
-void cpu_stw_le_data(CPUArchState *env, target_ulong ptr, uint32_t val)
-{
-    cpu_stw_le_data_ra(env, ptr, val, 0);
-}
-
-void cpu_stl_le_data(CPUArchState *env, target_ulong ptr, uint32_t val)
-{
-    cpu_stl_le_data_ra(env, ptr, val, 0);
-}
-
-void cpu_stq_le_data(CPUArchState *env, target_ulong ptr, uint64_t val)
-{
-    cpu_stq_le_data_ra(env, ptr, val, 0);
-}
+#include "ldst_common.c.inc"
 
 /*
  * First set of functions passes in OI and RETADDR.
diff --git a/accel/tcg/ldst_common.c.inc b/accel/tcg/ldst_common.c.inc
new file mode 100644
index 0000000000..bfefb275e7
--- /dev/null
+++ b/accel/tcg/ldst_common.c.inc
@@ -0,0 +1,307 @@
+/*
+ * Routines common to user and system emulation of load/store.
+ *
+ *  Copyright (c) 2003 Fabrice Bellard
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+uint32_t cpu_ldub_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                            int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_UB, mmu_idx);
+    return cpu_ldb_mmu(env, addr, oi, ra);
+}
+
+int cpu_ldsb_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                       int mmu_idx, uintptr_t ra)
+{
+    return (int8_t)cpu_ldub_mmuidx_ra(env, addr, mmu_idx, ra);
+}
+
+uint32_t cpu_lduw_be_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                               int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_BEUW | MO_UNALN, mmu_idx);
+    return cpu_ldw_be_mmu(env, addr, oi, ra);
+}
+
+int cpu_ldsw_be_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                          int mmu_idx, uintptr_t ra)
+{
+    return (int16_t)cpu_lduw_be_mmuidx_ra(env, addr, mmu_idx, ra);
+}
+
+uint32_t cpu_ldl_be_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                              int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_BEUL | MO_UNALN, mmu_idx);
+    return cpu_ldl_be_mmu(env, addr, oi, ra);
+}
+
+uint64_t cpu_ldq_be_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                              int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_BEQ | MO_UNALN, mmu_idx);
+    return cpu_ldq_be_mmu(env, addr, oi, ra);
+}
+
+uint32_t cpu_lduw_le_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                               int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_LEUW | MO_UNALN, mmu_idx);
+    return cpu_ldw_le_mmu(env, addr, oi, ra);
+}
+
+int cpu_ldsw_le_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                          int mmu_idx, uintptr_t ra)
+{
+    return (int16_t)cpu_lduw_le_mmuidx_ra(env, addr, mmu_idx, ra);
+}
+
+uint32_t cpu_ldl_le_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                              int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_LEUL | MO_UNALN, mmu_idx);
+    return cpu_ldl_le_mmu(env, addr, oi, ra);
+}
+
+uint64_t cpu_ldq_le_mmuidx_ra(CPUArchState *env, abi_ptr addr,
+                              int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_LEQ | MO_UNALN, mmu_idx);
+    return cpu_ldq_le_mmu(env, addr, oi, ra);
+}
+
+void cpu_stb_mmuidx_ra(CPUArchState *env, abi_ptr addr, uint32_t val,
+                       int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_UB, mmu_idx);
+    cpu_stb_mmu(env, addr, val, oi, ra);
+}
+
+void cpu_stw_be_mmuidx_ra(CPUArchState *env, abi_ptr addr, uint32_t val,
+                          int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_BEUW | MO_UNALN, mmu_idx);
+    cpu_stw_be_mmu(env, addr, val, oi, ra);
+}
+
+void cpu_stl_be_mmuidx_ra(CPUArchState *env, abi_ptr addr, uint32_t val,
+                          int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_BEUL | MO_UNALN, mmu_idx);
+    cpu_stl_be_mmu(env, addr, val, oi, ra);
+}
+
+void cpu_stq_be_mmuidx_ra(CPUArchState *env, abi_ptr addr, uint64_t val,
+                          int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_BEQ | MO_UNALN, mmu_idx);
+    cpu_stq_be_mmu(env, addr, val, oi, ra);
+}
+
+void cpu_stw_le_mmuidx_ra(CPUArchState *env, abi_ptr addr, uint32_t val,
+                          int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_LEUW | MO_UNALN, mmu_idx);
+    cpu_stw_le_mmu(env, addr, val, oi, ra);
+}
+
+void cpu_stl_le_mmuidx_ra(CPUArchState *env, abi_ptr addr, uint32_t val,
+                          int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_LEUL | MO_UNALN, mmu_idx);
+    cpu_stl_le_mmu(env, addr, val, oi, ra);
+}
+
+void cpu_stq_le_mmuidx_ra(CPUArchState *env, abi_ptr addr, uint64_t val,
+                          int mmu_idx, uintptr_t ra)
+{
+    MemOpIdx oi = make_memop_idx(MO_LEQ | MO_UNALN, mmu_idx);
+    cpu_stq_le_mmu(env, addr, val, oi, ra);
+}
+
+/*--------------------------*/
+
+uint32_t cpu_ldub_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return cpu_ldub_mmuidx_ra(env, addr, cpu_mmu_index(env, false), ra);
+}
+
+int cpu_ldsb_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return (int8_t)cpu_ldub_data_ra(env, addr, ra);
+}
+
+uint32_t cpu_lduw_be_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return cpu_lduw_be_mmuidx_ra(env, addr, cpu_mmu_index(env, false), ra);
+}
+
+int cpu_ldsw_be_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return (int16_t)cpu_lduw_be_data_ra(env, addr, ra);
+}
+
+uint32_t cpu_ldl_be_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return cpu_ldl_be_mmuidx_ra(env, addr, cpu_mmu_index(env, false), ra);
+}
+
+uint64_t cpu_ldq_be_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return cpu_ldq_be_mmuidx_ra(env, addr, cpu_mmu_index(env, false), ra);
+}
+
+uint32_t cpu_lduw_le_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return cpu_lduw_le_mmuidx_ra(env, addr, cpu_mmu_index(env, false), ra);
+}
+
+int cpu_ldsw_le_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return (int16_t)cpu_lduw_le_data_ra(env, addr, ra);
+}
+
+uint32_t cpu_ldl_le_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return cpu_ldl_le_mmuidx_ra(env, addr, cpu_mmu_index(env, false), ra);
+}
+
+uint64_t cpu_ldq_le_data_ra(CPUArchState *env, abi_ptr addr, uintptr_t ra)
+{
+    return cpu_ldq_le_mmuidx_ra(env, addr, cpu_mmu_index(env, false), ra);
+}
+
+void cpu_stb_data_ra(CPUArchState *env, abi_ptr addr,
+                     uint32_t val, uintptr_t ra)
+{
+    cpu_stb_mmuidx_ra(env, addr, val, cpu_mmu_index(env, false), ra);
+}
+
+void cpu_stw_be_data_ra(CPUArchState *env, abi_ptr addr,
+                        uint32_t val, uintptr_t ra)
+{
+    cpu_stw_be_mmuidx_ra(env, addr, val, cpu_mmu_index(env, false), ra);
+}
+
+void cpu_stl_be_data_ra(CPUArchState *env, abi_ptr addr,
+                        uint32_t val, uintptr_t ra)
+{
+    cpu_stl_be_mmuidx_ra(env, addr, val, cpu_mmu_index(env, false), ra);
+}
+
+void cpu_stq_be_data_ra(CPUArchState *env, abi_ptr addr,
+                        uint64_t val, uintptr_t ra)
+{
+    cpu_stq_be_mmuidx_ra(env, addr, val, cpu_mmu_index(env, false), ra);
+}
+
+void cpu_stw_le_data_ra(CPUArchState *env, abi_ptr addr,
+                        uint32_t val, uintptr_t ra)
+{
+    cpu_stw_le_mmuidx_ra(env, addr, val, cpu_mmu_index(env, false), ra);
+}
+
+void cpu_stl_le_data_ra(CPUArchState *env, abi_ptr addr,
+                        uint32_t val, uintptr_t ra)
+{
+    cpu_stl_le_mmuidx_ra(env, addr, val, cpu_mmu_index(env, false), ra);
+}
+
+void cpu_stq_le_data_ra(CPUArchState *env, abi_ptr addr,
+                        uint64_t val, uintptr_t ra)
+{
+    cpu_stq_le_mmuidx_ra(env, addr, val, cpu_mmu_index(env, false), ra);
+}
+
+/*--------------------------*/
+
+uint32_t cpu_ldub_data(CPUArchState *env, abi_ptr addr)
+{
+    return cpu_ldub_data_ra(env, addr, 0);
+}
+
+int cpu_ldsb_data(CPUArchState *env, abi_ptr addr)
+{
+    return (int8_t)cpu_ldub_data(env, addr);
+}
+
+uint32_t cpu_lduw_be_data(CPUArchState *env, abi_ptr addr)
+{
+    return cpu_lduw_be_data_ra(env, addr, 0);
+}
+
+int cpu_ldsw_be_data(CPUArchState *env, abi_ptr addr)
+{
+    return (int16_t)cpu_lduw_be_data(env, addr);
+}
+
+uint32_t cpu_ldl_be_data(CPUArchState *env, abi_ptr addr)
+{
+    return cpu_ldl_be_data_ra(env, addr, 0);
+}
+
+uint64_t cpu_ldq_be_data(CPUArchState *env, abi_ptr addr)
+{
+    return cpu_ldq_be_data_ra(env, addr, 0);
+}
+
+uint32_t cpu_lduw_le_data(CPUArchState *env, abi_ptr addr)
+{
+    return cpu_lduw_le_data_ra(env, addr, 0);
+}
+
+int cpu_ldsw_le_data(CPUArchState *env, abi_ptr addr)
+{
+    return (int16_t)cpu_lduw_le_data(env, addr);
+}
+
+uint32_t cpu_ldl_le_data(CPUArchState *env, abi_ptr addr)
+{
+    return cpu_ldl_le_data_ra(env, addr, 0);
+}
+
+uint64_t cpu_ldq_le_data(CPUArchState *env, abi_ptr addr)
+{
+    return cpu_ldq_le_data_ra(env, addr, 0);
+}
+
+void cpu_stb_data(CPUArchState *env, abi_ptr addr, uint32_t val)
+{
+    cpu_stb_data_ra(env, addr, val, 0);
+}
+
+void cpu_stw_be_data(CPUArchState *env, abi_ptr addr, uint32_t val)
+{
+    cpu_stw_be_data_ra(env, addr, val, 0);
+}
+
+void cpu_stl_be_data(CPUArchState *env, abi_ptr addr, uint32_t val)
+{
+    cpu_stl_be_data_ra(env, addr, val, 0);
+}
+
+void cpu_stq_be_data(CPUArchState *env, abi_ptr addr, uint64_t val)
+{
+    cpu_stq_be_data_ra(env, addr, val, 0);
+}
+
+void cpu_stw_le_data(CPUArchState *env, abi_ptr addr, uint32_t val)
+{
+    cpu_stw_le_data_ra(env, addr, val, 0);
+}
+
+void cpu_stl_le_data(CPUArchState *env, abi_ptr addr, uint32_t val)
+{
+    cpu_stl_le_data_ra(env, addr, val, 0);
+}
+
+void cpu_stq_le_data(CPUArchState *env, abi_ptr addr, uint64_t val)
+{
+    cpu_stq_le_data_ra(env, addr, val, 0);
+}
diff --git a/accel/tcg/user-exec.c b/accel/tcg/user-exec.c
index 65d3c9b286..e6bb29b42d 100644
--- a/accel/tcg/user-exec.c
+++ b/accel/tcg/user-exec.c
@@ -886,300 +886,227 @@ int cpu_signal_handler(int host_signum, void *pinfo,
 
 /* The softmmu versions of these helpers are in cputlb.c.  */
 
-uint32_t cpu_ldub_data(CPUArchState *env, abi_ptr ptr)
-{
-    MemOpIdx oi = make_memop_idx(MO_UB, MMU_USER_IDX);
-    uint32_t ret;
-
-    trace_guest_ld_before_exec(env_cpu(env), ptr, oi);
-    ret = ldub_p(g2h(env_cpu(env), ptr));
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_R);
-    return ret;
-}
-
-int cpu_ldsb_data(CPUArchState *env, abi_ptr ptr)
-{
-    return (int8_t)cpu_ldub_data(env, ptr);
-}
-
-uint32_t cpu_lduw_be_data(CPUArchState *env, abi_ptr ptr)
-{
-    MemOpIdx oi = make_memop_idx(MO_BEUW, MMU_USER_IDX);
-    uint32_t ret;
-
-    trace_guest_ld_before_exec(env_cpu(env), ptr, oi);
-    ret = lduw_be_p(g2h(env_cpu(env), ptr));
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_R);
-    return ret;
-}
-
-int cpu_ldsw_be_data(CPUArchState *env, abi_ptr ptr)
-{
-    return (int16_t)cpu_lduw_be_data(env, ptr);
-}
-
-uint32_t cpu_ldl_be_data(CPUArchState *env, abi_ptr ptr)
-{
-    MemOpIdx oi = make_memop_idx(MO_BEUL, MMU_USER_IDX);
-    uint32_t ret;
-
-    trace_guest_ld_before_exec(env_cpu(env), ptr, oi);
-    ret = ldl_be_p(g2h(env_cpu(env), ptr));
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_R);
-    return ret;
-}
-
-uint64_t cpu_ldq_be_data(CPUArchState *env, abi_ptr ptr)
-{
-    MemOpIdx oi = make_memop_idx(MO_BEQ, MMU_USER_IDX);
-    uint64_t ret;
-
-    trace_guest_ld_before_exec(env_cpu(env), ptr, oi);
-    ret = ldq_be_p(g2h(env_cpu(env), ptr));
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_R);
-    return ret;
-}
-
-uint32_t cpu_lduw_le_data(CPUArchState *env, abi_ptr ptr)
-{
-    MemOpIdx oi = make_memop_idx(MO_LEUW, MMU_USER_IDX);
-    uint32_t ret;
-
-    trace_guest_ld_before_exec(env_cpu(env), ptr, oi);
-    ret = lduw_le_p(g2h(env_cpu(env), ptr));
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_R);
-    return ret;
-}
-
-int cpu_ldsw_le_data(CPUArchState *env, abi_ptr ptr)
+/*
+ * Verify that we have passed the correct MemOp to the correct function.
+ *
+ * We could present one function to target code, and dispatch based on
+ * the MemOp, but so far we have worked hard to avoid an indirect function
+ * call along the memory path.
+ */
+static void validate_memop(MemOpIdx oi, MemOp expected)
 {
-    return (int16_t)cpu_lduw_le_data(env, ptr);
+#ifdef CONFIG_DEBUG_TCG
+    MemOp have = get_memop(oi) & (MO_SIZE | MO_BSWAP);
+    assert(have == expected);
+#endif
 }
 
-uint32_t cpu_ldl_le_data(CPUArchState *env, abi_ptr ptr)
+static void *cpu_mmu_lookup(CPUArchState *env, target_ulong addr,
+                            MemOpIdx oi, uintptr_t ra, MMUAccessType type)
 {
-    MemOpIdx oi = make_memop_idx(MO_LEUL, MMU_USER_IDX);
-    uint32_t ret;
-
-    trace_guest_ld_before_exec(env_cpu(env), ptr, oi);
-    ret = ldl_le_p(g2h(env_cpu(env), ptr));
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_R);
-    return ret;
-}
+    void *ret;
 
-uint64_t cpu_ldq_le_data(CPUArchState *env, abi_ptr ptr)
-{
-    MemOpIdx oi = make_memop_idx(MO_LEQ, MMU_USER_IDX);
-    uint64_t ret;
+    /* TODO: Enforce guest required alignment.  */
 
-    trace_guest_ld_before_exec(env_cpu(env), ptr, oi);
-    ret = ldq_le_p(g2h(env_cpu(env), ptr));
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_R);
+    ret = g2h(env_cpu(env), addr);
+    set_helper_retaddr(ra);
     return ret;
 }
 
-uint32_t cpu_ldub_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
+uint8_t cpu_ldb_mmu(CPUArchState *env, abi_ptr addr,
+                    MemOpIdx oi, uintptr_t ra)
 {
-    uint32_t ret;
+    void *haddr;
+    uint8_t ret;
 
-    set_helper_retaddr(retaddr);
-    ret = cpu_ldub_data(env, ptr);
+    validate_memop(oi, MO_UB);
+    trace_guest_ld_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_LOAD);
+    ret = ldub_p(haddr);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_R);
     return ret;
 }
 
-int cpu_ldsb_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
-{
-    return (int8_t)cpu_ldub_data_ra(env, ptr, retaddr);
-}
-
-uint32_t cpu_lduw_be_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
+uint16_t cpu_ldw_be_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
-    uint32_t ret;
+    void *haddr;
+    uint16_t ret;
 
-    set_helper_retaddr(retaddr);
-    ret = cpu_lduw_be_data(env, ptr);
+    validate_memop(oi, MO_BEUW);
+    trace_guest_ld_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_LOAD);
+    ret = lduw_be_p(haddr);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_R);
     return ret;
 }
 
-int cpu_ldsw_be_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
-{
-    return (int16_t)cpu_lduw_be_data_ra(env, ptr, retaddr);
-}
-
-uint32_t cpu_ldl_be_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
+uint32_t cpu_ldl_be_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
+    void *haddr;
     uint32_t ret;
 
-    set_helper_retaddr(retaddr);
-    ret = cpu_ldl_be_data(env, ptr);
+    validate_memop(oi, MO_BEUL);
+    trace_guest_ld_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_LOAD);
+    ret = ldl_be_p(haddr);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_R);
     return ret;
 }
 
-uint64_t cpu_ldq_be_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
+uint64_t cpu_ldq_be_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
+    void *haddr;
     uint64_t ret;
 
-    set_helper_retaddr(retaddr);
-    ret = cpu_ldq_be_data(env, ptr);
+    validate_memop(oi, MO_BEQ);
+    trace_guest_ld_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_LOAD);
+    ret = ldq_be_p(haddr);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_R);
     return ret;
 }
 
-uint32_t cpu_lduw_le_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
+uint16_t cpu_ldw_le_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
-    uint32_t ret;
+    void *haddr;
+    uint16_t ret;
 
-    set_helper_retaddr(retaddr);
-    ret = cpu_lduw_le_data(env, ptr);
+    validate_memop(oi, MO_LEUW);
+    trace_guest_ld_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_LOAD);
+    ret = lduw_le_p(haddr);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_R);
     return ret;
 }
 
-int cpu_ldsw_le_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
-{
-    return (int16_t)cpu_lduw_le_data_ra(env, ptr, retaddr);
-}
-
-uint32_t cpu_ldl_le_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
+uint32_t cpu_ldl_le_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
+    void *haddr;
     uint32_t ret;
 
-    set_helper_retaddr(retaddr);
-    ret = cpu_ldl_le_data(env, ptr);
+    validate_memop(oi, MO_LEUL);
+    trace_guest_ld_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_LOAD);
+    ret = ldl_le_p(haddr);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_R);
     return ret;
 }
 
-uint64_t cpu_ldq_le_data_ra(CPUArchState *env, abi_ptr ptr, uintptr_t retaddr)
+uint64_t cpu_ldq_le_mmu(CPUArchState *env, abi_ptr addr,
+                        MemOpIdx oi, uintptr_t ra)
 {
+    void *haddr;
     uint64_t ret;
 
-    set_helper_retaddr(retaddr);
-    ret = cpu_ldq_le_data(env, ptr);
+    validate_memop(oi, MO_LEQ);
+    trace_guest_ld_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_LOAD);
+    ret = ldq_le_p(haddr);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_R);
     return ret;
 }
 
-void cpu_stb_data(CPUArchState *env, abi_ptr ptr, uint32_t val)
-{
-    MemOpIdx oi = make_memop_idx(MO_UB, MMU_USER_IDX);
-
-    trace_guest_st_before_exec(env_cpu(env), ptr, oi);
-    stb_p(g2h(env_cpu(env), ptr), val);
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_W);
-}
-
-void cpu_stw_be_data(CPUArchState *env, abi_ptr ptr, uint32_t val)
-{
-    MemOpIdx oi = make_memop_idx(MO_BEUW, MMU_USER_IDX);
-
-    trace_guest_st_before_exec(env_cpu(env), ptr, oi);
-    stw_be_p(g2h(env_cpu(env), ptr), val);
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_W);
-}
-
-void cpu_stl_be_data(CPUArchState *env, abi_ptr ptr, uint32_t val)
-{
-    MemOpIdx oi = make_memop_idx(MO_BEUL, MMU_USER_IDX);
-
-    trace_guest_st_before_exec(env_cpu(env), ptr, oi);
-    stl_be_p(g2h(env_cpu(env), ptr), val);
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_W);
-}
-
-void cpu_stq_be_data(CPUArchState *env, abi_ptr ptr, uint64_t val)
+void cpu_stb_mmu(CPUArchState *env, abi_ptr addr, uint8_t val,
+                 MemOpIdx oi, uintptr_t ra)
 {
-    MemOpIdx oi = make_memop_idx(MO_BEQ, MMU_USER_IDX);
+    void *haddr;
 
-    trace_guest_st_before_exec(env_cpu(env), ptr, oi);
-    stq_be_p(g2h(env_cpu(env), ptr), val);
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_W);
+    validate_memop(oi, MO_UB);
+    trace_guest_st_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE);
+    stb_p(haddr, val);
+    clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_W);
 }
 
-void cpu_stw_le_data(CPUArchState *env, abi_ptr ptr, uint32_t val)
+void cpu_stw_be_mmu(CPUArchState *env, abi_ptr addr, uint16_t val,
+                    MemOpIdx oi, uintptr_t ra)
 {
-    MemOpIdx oi = make_memop_idx(MO_LEUW, MMU_USER_IDX);
+    void *haddr;
 
-    trace_guest_st_before_exec(env_cpu(env), ptr, oi);
-    stw_le_p(g2h(env_cpu(env), ptr), val);
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_W);
+    validate_memop(oi, MO_BEUW);
+    trace_guest_st_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE);
+    stw_be_p(haddr, val);
+    clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_W);
 }
 
-void cpu_stl_le_data(CPUArchState *env, abi_ptr ptr, uint32_t val)
+void cpu_stl_be_mmu(CPUArchState *env, abi_ptr addr, uint32_t val,
+                    MemOpIdx oi, uintptr_t ra)
 {
-    MemOpIdx oi = make_memop_idx(MO_LEUL, MMU_USER_IDX);
+    void *haddr;
 
-    trace_guest_st_before_exec(env_cpu(env), ptr, oi);
-    stl_le_p(g2h(env_cpu(env), ptr), val);
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_W);
+    validate_memop(oi, MO_BEUL);
+    trace_guest_st_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE);
+    stl_be_p(haddr, val);
+    clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_W);
 }
 
-void cpu_stq_le_data(CPUArchState *env, abi_ptr ptr, uint64_t val)
+void cpu_stq_be_mmu(CPUArchState *env, abi_ptr addr, uint64_t val,
+                    MemOpIdx oi, uintptr_t ra)
 {
-    MemOpIdx oi = make_memop_idx(MO_LEQ, MMU_USER_IDX);
-
-    trace_guest_st_before_exec(env_cpu(env), ptr, oi);
-    stq_le_p(g2h(env_cpu(env), ptr), val);
-    qemu_plugin_vcpu_mem_cb(env_cpu(env), ptr, oi, QEMU_PLUGIN_MEM_W);
-}
+    void *haddr;
 
-void cpu_stb_data_ra(CPUArchState *env, abi_ptr ptr,
-                     uint32_t val, uintptr_t retaddr)
-{
-    set_helper_retaddr(retaddr);
-    cpu_stb_data(env, ptr, val);
+    validate_memop(oi, MO_BEQ);
+    trace_guest_st_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE);
+    stq_be_p(haddr, val);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_W);
 }
 
-void cpu_stw_be_data_ra(CPUArchState *env, abi_ptr ptr,
-                        uint32_t val, uintptr_t retaddr)
+void cpu_stw_le_mmu(CPUArchState *env, abi_ptr addr, uint16_t val,
+                    MemOpIdx oi, uintptr_t ra)
 {
-    set_helper_retaddr(retaddr);
-    cpu_stw_be_data(env, ptr, val);
-    clear_helper_retaddr();
-}
+    void *haddr;
 
-void cpu_stl_be_data_ra(CPUArchState *env, abi_ptr ptr,
-                        uint32_t val, uintptr_t retaddr)
-{
-    set_helper_retaddr(retaddr);
-    cpu_stl_be_data(env, ptr, val);
+    validate_memop(oi, MO_LEUW);
+    trace_guest_st_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE);
+    stw_le_p(haddr, val);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_W);
 }
 
-void cpu_stq_be_data_ra(CPUArchState *env, abi_ptr ptr,
-                        uint64_t val, uintptr_t retaddr)
+void cpu_stl_le_mmu(CPUArchState *env, abi_ptr addr, uint32_t val,
+                    MemOpIdx oi, uintptr_t ra)
 {
-    set_helper_retaddr(retaddr);
-    cpu_stq_be_data(env, ptr, val);
-    clear_helper_retaddr();
-}
+    void *haddr;
 
-void cpu_stw_le_data_ra(CPUArchState *env, abi_ptr ptr,
-                        uint32_t val, uintptr_t retaddr)
-{
-    set_helper_retaddr(retaddr);
-    cpu_stw_le_data(env, ptr, val);
+    validate_memop(oi, MO_LEUL);
+    trace_guest_st_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE);
+    stl_le_p(haddr, val);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_W);
 }
 
-void cpu_stl_le_data_ra(CPUArchState *env, abi_ptr ptr,
-                        uint32_t val, uintptr_t retaddr)
+void cpu_stq_le_mmu(CPUArchState *env, abi_ptr addr, uint64_t val,
+                    MemOpIdx oi, uintptr_t ra)
 {
-    set_helper_retaddr(retaddr);
-    cpu_stl_le_data(env, ptr, val);
-    clear_helper_retaddr();
-}
+    void *haddr;
 
-void cpu_stq_le_data_ra(CPUArchState *env, abi_ptr ptr,
-                        uint64_t val, uintptr_t retaddr)
-{
-    set_helper_retaddr(retaddr);
-    cpu_stq_le_data(env, ptr, val);
+    validate_memop(oi, MO_LEQ);
+    trace_guest_st_before_exec(env_cpu(env), addr, oi);
+    haddr = cpu_mmu_lookup(env, addr, oi, ra, MMU_DATA_STORE);
+    stq_le_p(haddr, val);
     clear_helper_retaddr();
+    qemu_plugin_vcpu_mem_cb(env_cpu(env), addr, oi, QEMU_PLUGIN_MEM_W);
 }
 
 uint32_t cpu_ldub_code(CPUArchState *env, abi_ptr ptr)
@@ -1222,6 +1149,8 @@ uint64_t cpu_ldq_code(CPUArchState *env, abi_ptr ptr)
     return ret;
 }
 
+#include "ldst_common.c.inc"
+
 /*
  * Do not allow unaligned operations to proceed.  Return the host address.
  *