Commit f00bf5dc authored by Jason Xing's avatar Jason Xing Committed by David S. Miller
Browse files

net: rps: protect filter locklessly 



As we can see, rflow->filter can be written/read concurrently, so
lockless access is needed.

Signed-off-by: default avatarJason Xing <kernelxing@tencent.com>
Reviewed-by: default avatarEric Dumazet <edumazet@google.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 84b6823c

net/core/dev.c

+4 −4
Original line number Diff line number Diff line
@@ -4530,9 +4530,9 @@ set_rps_cpu(struct net_device *dev, struct sk_buff *skb, 
			goto out;

		old_rflow = rflow;

		rflow = &flow_table->flows[flow_id];

		rflow->filter = rc;

		if (old_rflow->filter == rflow->filter)

			old_rflow->filter = RPS_NO_FILTER;

		WRITE_ONCE(rflow->filter, rc);

		if (old_rflow->filter == rc)

			WRITE_ONCE(old_rflow->filter, RPS_NO_FILTER);

	out:

#endif

		head = READ_ONCE(per_cpu(softnet_data, next_cpu).input_queue_head);
@@ -4672,7 +4672,7 @@ bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, 
	if (flow_table && flow_id <= flow_table->mask) {

		rflow = &flow_table->flows[flow_id];

		cpu = READ_ONCE(rflow->cpu);

		if (rflow->filter == filter_id && cpu < nr_cpu_ids &&

		if (READ_ONCE(rflow->filter) == filter_id && cpu < nr_cpu_ids &&

		    ((int)(READ_ONCE(per_cpu(softnet_data, cpu).input_queue_head) -

			   READ_ONCE(rflow->last_qtail)) <

		     (int)(10 * flow_table->mask)))