On Fri, 20 Jun 2025 10:15:42 +0200
Peter Zijlstra [off-list ref] wrote:

On Tue, Jun 10, 2025 at 08:54:29PM -0400, Steven Rostedt wrote:

quoted

 void unwind_deferred_cancel(struct unwind_work *work)
 {
+	struct task_struct *g, *t;
+
 	if (!work)
 		return;
 
 	guard(mutex)(&callback_mutex);
 	list_del(&work->list);
+
+	clear_bit(work->bit, &unwind_mask);  

atomic bitop

Yeah, it just seemed cleaner than: unwind_mask &= ~(work->bit);

It's not needed as the update of unwind_mask is done within the
callback_mutex.

quoted

+
+	guard(rcu)();
+	/* Clear this bit from all threads */
+	for_each_process_thread(g, t) {
+		clear_bit(work->bit, &t->unwind_info.unwind_mask);
+	}
 }
 
 int unwind_deferred_init(struct unwind_work *work, unwind_callback_t func)

@@ -256,6 +278,14 @@ int unwind_deferred_init(struct unwind_work *work, unwind_callback_t func)
 	memset(work, 0, sizeof(*work));
 
 	guard(mutex)(&callback_mutex);
+
+	/* See if there's a bit in the mask available */
+	if (unwind_mask == ~0UL)
+		return -EBUSY;
+
+	work->bit = ffz(unwind_mask);
+	unwind_mask |= BIT(work->bit);  

regular or

quoted

+
 	list_add(&work->list, &callbacks);
 	work->func = func;
 	return 0;

@@ -267,6 +297,7 @@ void unwind_task_init(struct task_struct *task)
 
 	memset(info, 0, sizeof(*info));
 	init_task_work(&info->work, unwind_deferred_task_work);
+	info->unwind_mask = 0;
 }  

Which is somewhat inconsistent;

  __clear_bit()/__set_bit()

Hmm, are the above non-atomic?

or:

  unwind_mask &= ~BIT() / unwind_mask |= BIT()

although, because the update is always guarded, this may be the better
approach, as it shows there's no atomic needed.

-- Steve