From: "Paul E. McKenney" <paulmck@kernel.org>
Date: Fri, 3 Jan 2020 15:44:23 -0800
Subject: rcu: Add WRITE_ONCE() to rcu_node ->qsmaskinitnext
Patch-mainline: v5.7-rc1
Git-commit: 105abf82b0a62664edabcd4c5b4d9414c74ef399
References: bsc#1171828

The rcu_state structure's ->qsmaskinitnext field is read locklessly,
so this commit adds the WRITE_ONCE() to an update in order to provide
proper documentation and READ_ONCE()/WRITE_ONCE() pairing.

This data race was reported by KCSAN.  Not appropriate for backporting
due to failure being unlikely for systems not doing incessant CPU-hotplug
operations.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Acked-by: Daniel Wagner <dwagner@suse.de>
---
 kernel/rcu/tree.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -3083,7 +3083,7 @@ void rcu_cpu_starting(unsigned int cpu)
 	rnp = rdp->mynode;
 	mask = rdp->grpmask;
 	raw_spin_lock_irqsave_rcu_node(rnp, flags);
-	rnp->qsmaskinitnext |= mask;
+	WRITE_ONCE(rnp->qsmaskinitnext, rnp->qsmaskinitnext | mask);
 	oldmask = rnp->expmaskinitnext;
 	rnp->expmaskinitnext |= mask;
 	oldmask ^= rnp->expmaskinitnext;
@@ -3135,7 +3135,7 @@ void rcu_report_dead(unsigned int cpu)
 		rcu_report_qs_rnp(mask, rnp, rnp->gp_seq, flags);
 		raw_spin_lock_irqsave_rcu_node(rnp, flags);
 	}
-	rnp->qsmaskinitnext &= ~mask;
+	WRITE_ONCE(rnp->qsmaskinitnext, rnp->qsmaskinitnext & ~mask);
 	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
 	raw_spin_unlock(&rcu_state.ofl_lock);