diff --git a/include/metalsvm/tasks_types.h b/include/metalsvm/tasks_types.h
index c6e835e5..9725454b 100644
--- a/include/metalsvm/tasks_types.h
+++ b/include/metalsvm/tasks_types.h
@@ -50,7 +50,6 @@ extern "C" {
 #define TASK_DEFAULT_FLAGS		0
 #define TASK_FPU_INIT			(1 << 0)
 #define TASK_FPU_USED			(1 << 1)
-#define TASK_SWITCH_IN_PROGESS	(1 << 2)
 
 typedef int (*entry_point_t)(void*);
 typedef int (STDCALL *internal_entry_point_t)(void*);
diff --git a/kernel/tasks.c b/kernel/tasks.c
index bf57dda0..6800e662 100644
--- a/kernel/tasks.c
+++ b/kernel/tasks.c
@@ -149,7 +149,7 @@ static void NORETURN do_exit(int arg) {
 	curr_task->status = TASK_FINISHED;
 	reschedule();
 	
-	kputs("Kernel panic: scheduler found no valid task\n");
+	kprintf("Kernel panic: scheduler on core %d found no valid task\n", CORE_ID);
 	while(1) {
 		HALT;
 	}
@@ -174,22 +174,13 @@ void NORETURN abort(void) {
 }
 
 /*
- * @brief: if the task switch is finished, we reset
- * the TASK_SWITCH_IN_PROGRESS bit
+ * @brief: if the task gets the first time slice,
+ * the table_lock is hold and have to be released.
  */
-inline static void task_switch_finished(void)
+inline static void start_first_time_slice(void)
 {
 #if MAX_CORES > 1
-	uint32_t flags = irq_nested_disable();
-
-	// do we already reset the TASK_SWITCH_IN_PROGRESS bit?
-	task_t* old = per_core(old_task);
-	if (old) {
-		old->flags &= ~TASK_SWITCH_IN_PROGESS;
-		per_core(old_task) = NULL;
-	}
-
-	irq_nested_enable(flags);
+	spinlock_irqsave_unlock(&table_lock);
 #endif
 }
 
@@ -314,10 +305,7 @@ int sys_fork(void)
 				// Leave the function without releasing the locks
 				// because the locks are already released 
 				// by the parent task!
-
-				// first switch to the new current task
-				// => signalizes a successful task switch
-				task_switch_finished();
+				start_first_time_slice();
 				return 0; 
 			}
 
@@ -352,9 +340,7 @@ static int STDCALL kernel_entry(void* args)
 	int ret;
 	kernel_args_t* kernel_args = (kernel_args_t*) args;
 
-	// first switch to the new current task
-	// => signalizes a successful task switch
-	task_switch_finished();
+	start_first_time_slice();
 
 	if (BUILTIN_EXPECT(!kernel_args, 0))
 		return -EINVAL;
@@ -595,9 +581,7 @@ static int STDCALL user_entry(void* arg)
 {
 	int ret;
 
-	// first switch to the new current task
-	// => signalizes a successful task switch
-	task_switch_finished();
+	start_first_time_slice();
 
 	if (BUILTIN_EXPECT(!arg, 0))
 		return -EINVAL;
@@ -816,11 +800,6 @@ void scheduler(void)
 	unsigned int i;
 	unsigned int new_id;
 
-	// let's play it save
-	// => check if we already signalizes that the previous switch
-	// is finished
-	task_switch_finished();
-
 #if MAX_CORES > 1
 	spinlock_irqsave_lock(&table_lock);
 #endif
@@ -840,32 +819,16 @@ void scheduler(void)
 	for(i=1, new_id=(curr_task->id + 1) % MAX_TASKS; 
 		i<MAX_TASKS; i++, new_id=(new_id+1) % MAX_TASKS) 
 	{
-		if ((task_table[new_id].status == TASK_READY) && !(task_table[new_id].flags & TASK_SWITCH_IN_PROGESS)) {
-			if (curr_task->status == TASK_RUNNING) {
+		if (task_table[new_id].status == TASK_READY) {
+			if (curr_task->status == TASK_RUNNING)
 				curr_task->status = TASK_READY;
-#if MAX_CORES > 1
-				curr_task->flags |= TASK_SWITCH_IN_PROGESS;
-				per_core(old_task) = curr_task;
-#endif
-			}
-#if MAX_CORES > 1
-				else per_core(old_task) = NULL;
-#endif
 			task_table[new_id].status = TASK_RUNNING;
 			curr_task = per_core(current_task) = task_table+new_id;
 
 			goto get_task_out;
 		}
-
-		//if ((task_table[new_id].status == TASK_READY) && (task_table[new_id].flags & TASK_SWITCH_IN_PROGESS))
-		//	kprintf("task switch %d is in progress\n", new_id);
 	}
 
-#if MAX_CORES > 1
-	// old task will never rescheduled
-	per_core(old_task) = NULL;
-#endif
-
 	if ((curr_task->status == TASK_RUNNING) || (curr_task->status == TASK_IDLE))
 		goto get_task_out;
 
@@ -878,14 +841,13 @@ void scheduler(void)
 
 get_task_out:
 	//kprintf("schedule %d on core %d\n", per_core(current_task)->id, smp_id());
+
+	if (curr_task != orig_task)
+		switch_task(new_id);
+
 #if MAX_CORES > 1
 	spinlock_irqsave_unlock(&table_lock);
 #endif
-
-	if (curr_task != orig_task) {
-		switch_task(new_id);
-		task_switch_finished();
-	}
 }
 
 void reschedule(void)