Re: [PATCH 0/6] locking/rtmutex: Avoid PI state recursion through sched_submit_work()

[Peter Zijlstra <peterz@infradead.org>]

On Wed, Aug 16, 2023 at 04:58:18PM +0200, Peter Zijlstra wrote:

> I've ended up with the below, but it is quite horrible.. but let me go
> stare at the futex wreckage before trying to clean things up.

OK, I think the below covers the simple case, now lets see if I can make
sense of futex_wait_requeue_pi()... :/

---
--- a/kernel/futex/pi.c
+++ b/kernel/futex/pi.c
@@ -1002,6 +1002,12 @@ int futex_lock_pi(u32 __user *uaddr, uns
 		goto no_block;
 	}
 
+	/*
+	 * Must be done before we enqueue the waiter, here is unfortunately
+	 * under the hb lock, but that *should* work.
+	 */
+	rt_mutex_pre_schedule();
+
 	rt_mutex_init_waiter(&rt_waiter);
 
 	/*
@@ -1052,6 +1058,10 @@ int futex_lock_pi(u32 __user *uaddr, uns
 	if (ret && !rt_mutex_cleanup_proxy_lock(&q.pi_state->pi_mutex, &rt_waiter))
 		ret = 0;
 
+	/*
+	 * Waiter is unqueued.
+	 */
+	rt_mutex_post_schedule();
 no_block:
 	/*
 	 * Fixup the pi_state owner and possibly acquire the lock if we