Re: [PATCH] tracing: Add trace_trigger kernel command line option

[Paul E. McKenney]

On Thu, Oct 20, 2022 at 02:53:40PM -0400, Steven Rostedt wrote:
> On Thu, 20 Oct 2022 10:55:12 -0700
> "Paul E. McKenney" <paulmck@xxxxxxxxxx> wrote:
> 
> > > synchronize_rcu() {
> > >     rcu_poll_gp_seq_start_unlocked() {
> > > 	struct rcu_node *rnp = rcu_get_root();
> > > 
> > > 	if (rcu_init_invoked()) {
> > > 		lockdep_assert_irqs_enabled();
> > > 		raw_spin_lock_irq_rcu_node(rnp);
> > > 	}
> > > 	rcu_poll_gp_seq_start(snap);
> > > 	if (rcu_init_invoked())
> > > 		raw_spin_unlock_irq_rcu_node(rnp);  <<-- Interrupts enabled here
> > >    }  
> > 
> > Thank you for digging into this!
> > 
> > Does the following help?
> 
> It does indeed!
> 
> I can remove the early_boot_irqs_disabled checks if you can add this to
> this rc cycle and mark it stable.

My plan is to push it in after a bit of testing.  The Fixes tag should
make the right -stable stuff happen.

							Thanx, Paul

------------------------------------------------------------------------

commit 952643575669332565c77066fb42d67987680f12
Author: Paul E. McKenney <paulmck@xxxxxxxxxx>
Date:   Thu Oct 20 10:58:14 2022 -0700

    rcu: Keep synchronize_rcu() from enabling irqs in early boot
    
    Making polled RCU grace periods account for expedited grace periods
    required acquiring the leaf rcu_node structure's lock during early boot,
    but after rcu_init() was called.  This lock is irq-disabled, but the
    code incorrectly assumes that irqs are always disabled when invoking
    synchronize_rcu().  The exception is early boot before the scheduler has
    started, which means that upon return from synchronize_rcu(), irqs will
    be incorrectly enabled.
    
    This commit fixes this bug by using irqsave/irqrestore locking primitives.
    
    Fixes: bf95b2bc3e42 ("rcu: Switch polled grace-period APIs to ->gp_seq_polled")
    
    Reported-by: Steven Rostedt <rostedt@xxxxxxxxxxx>
    Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxx>

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index abc615808b6e8..45ad9ef00d647 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -1425,30 +1425,32 @@ static void rcu_poll_gp_seq_end(unsigned long *snap)
 // where caller does not hold the root rcu_node structure's lock.
 static void rcu_poll_gp_seq_start_unlocked(unsigned long *snap)
 {
+	unsigned long flags;
 	struct rcu_node *rnp = rcu_get_root();
 
 	if (rcu_init_invoked()) {
 		lockdep_assert_irqs_enabled();
-		raw_spin_lock_irq_rcu_node(rnp);
+		raw_spin_lock_irqsave_rcu_node(rnp, flags);
 	}
 	rcu_poll_gp_seq_start(snap);
 	if (rcu_init_invoked())
-		raw_spin_unlock_irq_rcu_node(rnp);
+		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
 }
 
 // Make the polled API aware of the end of a grace period, but where
 // caller does not hold the root rcu_node structure's lock.
 static void rcu_poll_gp_seq_end_unlocked(unsigned long *snap)
 {
+	unsigned long flags;
 	struct rcu_node *rnp = rcu_get_root();
 
 	if (rcu_init_invoked()) {
 		lockdep_assert_irqs_enabled();
-		raw_spin_lock_irq_rcu_node(rnp);
+		raw_spin_lock_irqsave_rcu_node(rnp, flags);
 	}
 	rcu_poll_gp_seq_end(snap);
 	if (rcu_init_invoked())
-		raw_spin_unlock_irq_rcu_node(rnp);
+		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
 }
 
 /*