[PATCH 11/18] rcu/tree: Introduce expedited_drain flag

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
To: linux-kernel@vger.kernel.org
Cc: "Uladzislau Rezki (Sony)" <urezki@gmail.com>,
	Joel Fernandes <joel@joelfernandes.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	Ingo Molnar <mingo@redhat.com>,
	Josh Triplett <josh@joshtriplett.org>,
	Lai Jiangshan <jiangshanlai@gmail.com>,
	linux-mm@kvack.org,
	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
	"Paul E. McKenney" <paulmck@kernel.org>,
	"Rafael J. Wysocki" <rafael.j.wysocki@intel.com>,
	rcu@vger.kernel.org, Steven Rostedt <rostedt@goodmis.org>
Subject: [PATCH 11/18] rcu/tree: Introduce expedited_drain flag
Date: Sun, 29 Mar 2020 22:32:41 -0400	[thread overview]
Message-ID: <20200330023248.164994-12-joel@joelfernandes.org> (raw)
In-Reply-To: <20200330023248.164994-1-joel@joelfernandes.org>

From: "Uladzislau Rezki (Sony)" <urezki@gmail.com>

It is used and set to true when the bulk array can not
be maintained, it happens under low memory condition
and memory pressure.

In that case the drain work is scheduled right away and
not after KFREE_DRAIN_JIFFIES. It tends to speed up the
reclaim path. On the other hand, there is no data showing
the difference yet.

Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 kernel/rcu/tree.c | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 8fbc8450284db..3b94526f490cb 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -3128,14 +3128,16 @@ kvfree_call_rcu_add_ptr_to_bulk(struct kfree_rcu_cpu *krcp, void *ptr)
  * due to memory pressure.
  *
  * Each kvfree_call_rcu() request is added to a batch. The batch will be drained
- * every KFREE_DRAIN_JIFFIES number of jiffies. All the objects in the batch will
- * be free'd in workqueue context. This allows us to: batch requests together to
- * reduce the number of grace periods during heavy kfree_rcu()/kvfree_rcu() load.
+ * every KFREE_DRAIN_JIFFIES number of jiffies or can be scheduled right away if
+ * a low memory is detected. All the objects in the batch will be free'd in
+ * workqueue context. This allows us to: batch requests together to reduce the
+ * number of grace periods during heavy kfree_rcu()/kvfree_rcu() load.
  */
 void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
 {
 	unsigned long flags;
 	struct kfree_rcu_cpu *krcp;
+	bool expedited_drain = false;
 	void *ptr;
 
 	local_irq_save(flags);	// For safely calling this_cpu_ptr().
@@ -3161,6 +3163,14 @@ void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
 		head->func = func;
 		head->next = krcp->head;
 		krcp->head = head;
+
+		/*
+		 * There was an issue to place the pointer directly
+		 * into array, due to memory pressure. Initiate an
+		 * expedited drain to accelerate lazy invocation of
+		 * appropriate free calls.
+		 */
+		expedited_drain = true;
 	}
 
 	WRITE_ONCE(krcp->count, krcp->count + 1);
@@ -3169,7 +3179,9 @@ void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
 	if (rcu_scheduler_active == RCU_SCHEDULER_RUNNING &&
 	    !krcp->monitor_todo) {
 		krcp->monitor_todo = true;
-		schedule_delayed_work(&krcp->monitor_work, KFREE_DRAIN_JIFFIES);
+
+		schedule_delayed_work(&krcp->monitor_work,
+			expedited_drain ? 0 : KFREE_DRAIN_JIFFIES);
 	}
 
 unlock_return:
-- 
2.26.0.rc2.310.g2932bb562d-goog

