On Thu, Aug 17, 2017 at 09:30:28AM +0200, Sebastian Andrzej Siewior wrote:

On 2017-08-16 15:42:28 [-0500], Julia Cartwright wrote:

[..]

quoted

diff --git a/kernel/cpu_pm.c b/kernel/cpu_pm.c
index 9656a3c36503..9da42f83ee03 100644
--- a/kernel/cpu_pm.c
+++ b/kernel/cpu_pm.c

@@ -22,14 +22,13 @@
 #include <linux/spinlock.h>
 #include <linux/syscore_ops.h>
 
-static DEFINE_RWLOCK(cpu_pm_notifier_lock);
-static RAW_NOTIFIER_HEAD(cpu_pm_notifier_chain);
+static ATOMIC_NOTIFIER_HEAD(cpu_pm_notifier_chain);
 
 static int cpu_pm_notify(enum cpu_pm_event event, int nr_to_call, int *nr_calls)
 {
 	int ret;
 
-	ret = __raw_notifier_call_chain(&cpu_pm_notifier_chain, event, NULL,
+	ret = __atomic_notifier_call_chain(&cpu_pm_notifier_chain, event, NULL,
 		nr_to_call, nr_calls);

there is a piece missing, upstream has a different change queued. I know
that this is the same as in latest RT, this is just to let you know in
case someone complains about an RCU backtrace???

Ah!  Thanks. Indeed.  I did see the upstream-queued version get cc'd to
linux-rt-users, but didn't see the rcu_irq_{enter,exit}_irqson()
addition, assuming it to be what landed in rt-devel.

Will you be fixing this up in 4.11-rt?

   Julia