Re: [PATCH 6/9] lockdep: Introduce wait-type checks

[Peter Zijlstra]

On Tue, Mar 31, 2020 at 03:25:21PM +0200, Geert Uytterhoeven wrote:
> On arm64 (e.g. R-Car H3 ES2.0):
> 
> +=============================
> +[ BUG: Invalid wait context ]
> +5.6.0-salvator-x-09423-gb29514ba13a9c459-dirty #679 Not tainted
> +-----------------------------
> +swapper/5/0 is trying to lock:
> +ffffff86ff76f398 (&pool->lock){..-.}-{3:3}, at: __queue_work+0x134/0x430
> +other info that might help us debug this:
> +1 lock held by swapper/5/0:
> + #0: ffffffc01103a4a0 (rcu_read_lock){....}-{1:3}, at:
> rcu_lock_acquire.constprop.59+0x0/0x38
> +stack backtrace:
> +CPU: 5 PID: 0 Comm: swapper/5 Not tainted
> 5.6.0-salvator-x-09423-gb29514ba13a9c459-dirty #679
> +Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
> +Call trace:
> + dump_backtrace+0x0/0x180
> + show_stack+0x14/0x1c
> + dump_stack+0xdc/0x12c
> + __lock_acquire+0x37c/0xf9c
> + lock_acquire+0x258/0x288
> + _raw_spin_lock+0x34/0x48
> + __queue_work+0x134/0x430
> + queue_work_on+0x48/0x8c
> + timers_update_nohz+0x24/0x2c
> + tick_nohz_activate.isra.15.part.16+0x5c/0x80
> + tick_setup_sched_timer+0xe0/0xf0
> + hrtimer_run_queues+0x88/0xf8

So this is complaining that it cannot take pool->lock, which is
WAIT_CONFIG while holding RCU, which presents a WAIT_CONFIG context.

This seems to implicate something is amiss, because that should be
allowed. The thing it doesn't print is the context, which in the above
case is a (hrtimer) interrupt.

I suspect this really is a hardirq context and the next patch won't cure
things. It looks nohz (full?) related.

Frederic, can you untangle this?

---

diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 1511690e4de7..ac10db66cc63 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -3952,10 +3952,36 @@ static int mark_lock(struct task_struct *curr, struct held_lock *this,
 	return ret;
 }
 
+static inline short task_wait_context(struct task_struct *curr)
+{
+	/*
+	 * Set appropriate wait type for the context; for IRQs we have to take
+	 * into account force_irqthread as that is implied by PREEMPT_RT.
+	 */
+	if (curr->hardirq_context) {
+		/*
+		 * Check if force_irqthreads will run us threaded.
+		 */
+		if (curr->hardirq_threaded || curr->irq_config)
+			return LD_WAIT_CONFIG;
+
+		return LD_WAIT_SPIN;
+	} else if (curr->softirq_context) {
+		/*
+		 * Softirqs are always threaded.
+		 */
+		return LD_WAIT_CONFIG;
+	}
+
+	return LD_WAIT_MAX;
+}
+
 static int
 print_lock_invalid_wait_context(struct task_struct *curr,
 				struct held_lock *hlock)
 {
+	short curr_inner;
+
 	if (!debug_locks_off())
 		return 0;
 	if (debug_locks_silent)
@@ -3971,6 +3997,10 @@ print_lock_invalid_wait_context(struct task_struct *curr,
 	print_lock(hlock);
 
 	pr_warn("other info that might help us debug this:\n");
+
+	curr_inner = task_wait_context(curr);
+	pr_warn("context-{%d:%d}\n", curr_inner, curr_inner);
+
 	lockdep_print_held_locks(curr);
 
 	pr_warn("stack backtrace:\n");
@@ -4017,26 +4047,7 @@ static int check_wait_context(struct task_struct *curr, struct held_lock *next)
 	}
 	depth++;
 
-	/*
-	 * Set appropriate wait type for the context; for IRQs we have to take
-	 * into account force_irqthread as that is implied by PREEMPT_RT.
-	 */
-	if (curr->hardirq_context) {
-		/*
-		 * Check if force_irqthreads will run us threaded.
-		 */
-		if (curr->hardirq_threaded || curr->irq_config)
-			curr_inner = LD_WAIT_CONFIG;
-		else
-			curr_inner = LD_WAIT_SPIN;
-	} else if (curr->softirq_context) {
-		/*
-		 * Softirqs are always threaded.
-		 */
-		curr_inner = LD_WAIT_CONFIG;
-	} else {
-		curr_inner = LD_WAIT_MAX;
-	}
+	curr_inner = task_wait_context(curr);
 
 	for (; depth < curr->lockdep_depth; depth++) {
 		struct held_lock *prev = curr->held_locks + depth;