W dniu 9/16/25 o 12:49, Sabrina Dubroca pisze:

2025-09-12, 13:24:20 +0200, Marek Mietus wrote:

quoted

ovpn_udp4_output unnecessarily references the dst_entry from the
dst_cache.

This should probably include a description of why it's safe to skip
the reference in this context.

Noted, will do.

quoted

Reduce this overhead by using the newly implemented
udp_tunnel_xmit_skb_noref function and dst_cache helpers.

Signed-off-by: Marek Mietus <redacted>
---
 drivers/net/ovpn/udp.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ovpn/udp.c b/drivers/net/ovpn/udp.c
index d6a0f7a0b75d..c5d289c23d2b 100644
--- a/drivers/net/ovpn/udp.c
+++ b/drivers/net/ovpn/udp.c

@@ -158,7 +158,7 @@ static int ovpn_udp4_output(struct ovpn_peer *peer, struct ovpn_bind *bind,
 	int ret;
 
 	local_bh_disable();
-	rt = dst_cache_get_ip4(cache, &fl.saddr);
+	rt = dst_cache_get_ip4_rcu(cache, &fl.saddr);
 	if (rt)
 		goto transmit;
 

@@ -194,12 +194,12 @@ static int ovpn_udp4_output(struct ovpn_peer *peer, struct ovpn_bind *bind,
 				    ret);
 		goto err;
 	}
-	dst_cache_set_ip4(cache, &rt->dst, fl.saddr);
+	dst_cache_steal_ip4(cache, &rt->dst, fl.saddr);
 
 transmit:
-	udp_tunnel_xmit_skb(rt, sk, skb, fl.saddr, fl.daddr, 0,
-			    ip4_dst_hoplimit(&rt->dst), 0, fl.fl4_sport,
-			    fl.fl4_dport, false, sk->sk_no_check_tx, 0);
+	udp_tunnel_xmit_skb_noref(rt, sk, skb, fl.saddr, fl.daddr, 0,
+				  ip4_dst_hoplimit(&rt->dst), 0, fl.fl4_sport,
+				  fl.fl4_dport, false, sk->sk_no_check_tx, 0);
 	ret = 0;
 err:
 	local_bh_enable();

@@ -269,7 +269,7 @@ static int ovpn_udp6_output(struct ovpn_peer *peer, struct ovpn_bind *bind,
 	 * fragment packets if needed.
 	 *
 	 * NOTE: this is not needed for IPv4 because we pass df=0 to
-	 * udp_tunnel_xmit_skb()
+	 * udp_tunnel_xmit_skb_noref()
 	 */
 	skb->ignore_df = 1;
 	udp_tunnel6_xmit_skb(dst, sk, skb, skb->dev, &fl.saddr, &fl.daddr, 0,

Why are you changing only ipv4? Is there something in the ipv6 code
that prevents this?

I'm not sure. I'm not as acquainted with IPv6 as I am with IPv4. (and thought I'd hold off
until I got a positive response about the series)
IPv4 already has some noref xmit optimizations, so it just felt like the right place to start.