diff --git a/include/uc_priv.h b/include/uc_priv.h
index 4bd806cc..869cf685 100644
--- a/include/uc_priv.h
+++ b/include/uc_priv.h
@@ -357,6 +357,8 @@ struct uc_struct {
 
     sigjmp_buf jmp_bufs[UC_MAX_NESTED_LEVEL]; // To support nested uc_emu_start
     int nested_level;                         // Current nested_level
+
+    struct TranslationBlock* last_tb; // The real last tb we executed.
 };
 
 // Metadata stub for the variable-size cpu context used with uc_context_*()
diff --git a/qemu/accel/tcg/cpu-exec.c b/qemu/accel/tcg/cpu-exec.c
index 09cffc24..cb70582c 100644
--- a/qemu/accel/tcg/cpu-exec.c
+++ b/qemu/accel/tcg/cpu-exec.c
@@ -258,10 +258,9 @@ static inline TranslationBlock *tb_find(CPUState *cpu,
         /* We add the TB in the virtual pc hash table for the fast lookup */
         cpu->tb_jmp_cache[tb_jmp_cache_hash_func(cpu->uc, pc)] = tb;
 
-        UC_TB_COPY(&cur_tb, tb);
-
-        if (last_tb) {
-            UC_TB_COPY(&prev_tb, last_tb);
+        if (uc->last_tb) {
+            UC_TB_COPY(&cur_tb, tb);
+            UC_TB_COPY(&prev_tb, uc->last_tb);
             for (cur = uc->hook[UC_HOOK_EDGE_GENERATED_IDX].head;
                 cur != NULL && (hook = (struct hook *)cur->data); cur = cur->next) {
                 if (hook->to_delete) {
@@ -494,6 +493,7 @@ static inline void cpu_loop_exec_tb(CPUState *cpu, TranslationBlock *tb,
 
     // trace_exec_tb(tb, tb->pc);
     ret = cpu_tb_exec(cpu, tb);
+    cpu->uc->last_tb = tb; // Trace the last tb we executed.
     tb = (TranslationBlock *)(ret & ~TB_EXIT_MASK);
     *tb_exit = ret & TB_EXIT_MASK;
     if (*tb_exit != TB_EXIT_REQUESTED) {