summary refs log tree commit diff stats
diff options
context:
space:
mode:
-rw-r--r--target/riscv/translate.c46
1 files changed, 34 insertions, 12 deletions
diff --git a/target/riscv/translate.c b/target/riscv/translate.c
index d7a6de02df..9ddef2d6e2 100644
--- a/target/riscv/translate.c
+++ b/target/riscv/translate.c
@@ -1217,13 +1217,35 @@ const RISCVDecoder decoder_table[] = {
 
 const size_t decoder_table_size = ARRAY_SIZE(decoder_table);
 
-static void decode_opc(CPURISCVState *env, DisasContext *ctx, uint16_t opcode)
+static void decode_opc(CPURISCVState *env, DisasContext *ctx)
 {
+    uint32_t opcode;
+    bool pc_is_4byte_align = ((ctx->base.pc_next % 4) == 0);
+
     ctx->virt_inst_excp = false;
-    ctx->cur_insn_len = insn_len(opcode);
+    if (pc_is_4byte_align) {
+        /*
+         * Load 4 bytes at once to make instruction fetch atomically.
+         *
+         * Note: When pc is 4-byte aligned, 4-byte instruction wouldn't be
+         * across pages. We could preload 4 bytes instruction no matter
+         * real one is 2 or 4 bytes. Instruction preload wouldn't trigger
+         * additional page fault.
+         */
+        opcode = translator_ldl(env, &ctx->base, ctx->base.pc_next);
+    } else {
+        /*
+         * For unaligned pc, instruction preload may trigger additional
+         * page fault so we only load 2 bytes here.
+         */
+        opcode = (uint32_t) translator_lduw(env, &ctx->base, ctx->base.pc_next);
+    }
+    ctx->ol = ctx->xl;
+
+    ctx->cur_insn_len = insn_len((uint16_t)opcode);
     /* Check for compressed insn */
     if (ctx->cur_insn_len == 2) {
-        ctx->opcode = opcode;
+        ctx->opcode = (uint16_t)opcode;
         /*
          * The Zca extension is added as way to refer to instructions in the C
          * extension that do not include the floating-point loads and stores
@@ -1233,15 +1255,17 @@ static void decode_opc(CPURISCVState *env, DisasContext *ctx, uint16_t opcode)
             return;
         }
     } else {
-        uint32_t opcode32 = opcode;
-        opcode32 = deposit32(opcode32, 16, 16,
-                             translator_lduw(env, &ctx->base,
-                                             ctx->base.pc_next + 2));
-        ctx->opcode = opcode32;
+        if (!pc_is_4byte_align) {
+            /* Load last 2 bytes of instruction here */
+            opcode = deposit32(opcode, 16, 16,
+                               translator_lduw(env, &ctx->base,
+                                               ctx->base.pc_next + 2));
+        }
+        ctx->opcode = opcode;
 
         for (guint i = 0; i < ctx->decoders->len; ++i) {
             riscv_cpu_decode_fn func = g_ptr_array_index(ctx->decoders, i);
-            if (func(ctx, opcode32)) {
+            if (func(ctx, opcode)) {
                 return;
             }
         }
@@ -1319,10 +1343,8 @@ static void riscv_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
 {
     DisasContext *ctx = container_of(dcbase, DisasContext, base);
     CPURISCVState *env = cpu_env(cpu);
-    uint16_t opcode16 = translator_lduw(env, &ctx->base, ctx->base.pc_next);
 
-    ctx->ol = ctx->xl;
-    decode_opc(env, ctx, opcode16);
+    decode_opc(env, ctx);
     ctx->base.pc_next += ctx->cur_insn_len;
 
     /*