diff --git a/include/qom/cpu.h b/include/qom/cpu.h
index 5f188d4c28c7aac16fe7599b16daeebf58b82ce1..25eefea7ab40d3ffd6864b9bd232398f2d7eec24 100644
--- a/include/qom/cpu.h
+++ b/include/qom/cpu.h
@@ -302,6 +302,8 @@ struct qemu_work_item;
  * @kvm_fd: vCPU file descriptor for KVM.
  * @work_mutex: Lock to prevent multiple access to queued_work_*.
  * @queued_work_first: First asynchronous work pending.
+ * @trace_dstate_delayed: Delayed changes to trace_dstate (includes all changes
+ *                        to @trace_dstate).
  * @trace_dstate: Dynamic tracing state of events for this vCPU (bitmask).
  *
  * State of one CPU core or thread.
@@ -372,6 +374,7 @@ struct CPUState {
     struct kvm_run *kvm_run;
 
     /* Used for events with 'vcpu' and *without* the 'disabled' properties */
+    DECLARE_BITMAP(trace_dstate_delayed, CPU_TRACE_DSTATE_MAX_EVENTS);
     DECLARE_BITMAP(trace_dstate, CPU_TRACE_DSTATE_MAX_EVENTS);
 
     /* TODO Move common fields from CPUArchState here. */
diff --git a/trace/control-target.c b/trace/control-target.c
index 99a8ed5157d51502519e4ffbb587eb534806510c..31c2ed11a09c6415fa7c4a04306877e573a835f6 100644
--- a/trace/control-target.c
+++ b/trace/control-target.c
@@ -61,6 +61,13 @@ void trace_event_set_state_dynamic(TraceEvent *ev, bool state)
     }
 }
 
+static void trace_event_synchronize_vcpu_state_dynamic(
+    CPUState *vcpu, run_on_cpu_data ignored)
+{
+    bitmap_copy(vcpu->trace_dstate, vcpu->trace_dstate_delayed,
+                CPU_TRACE_DSTATE_MAX_EVENTS);
+}
+
 void trace_event_set_vcpu_state_dynamic(CPUState *vcpu,
                                         TraceEvent *ev, bool state)
 {
@@ -73,13 +80,20 @@ void trace_event_set_vcpu_state_dynamic(CPUState *vcpu,
     if (state_pre != state) {
         if (state) {
             trace_events_enabled_count++;
-            set_bit(vcpu_id, vcpu->trace_dstate);
+            set_bit(vcpu_id, vcpu->trace_dstate_delayed);
             (*ev->dstate)++;
         } else {
             trace_events_enabled_count--;
-            clear_bit(vcpu_id, vcpu->trace_dstate);
+            clear_bit(vcpu_id, vcpu->trace_dstate_delayed);
             (*ev->dstate)--;
         }
+        /*
+         * Delay changes until next TB; we want all TBs to be built from a
+         * single set of dstate values to ensure consistency of generated
+         * tracing code.
+         */
+        async_run_on_cpu(vcpu, trace_event_synchronize_vcpu_state_dynamic,
+                         RUN_ON_CPU_NULL);
     }
 }