aboutsummaryrefslogtreecommitdiffstats
path: root/target/linux/generic/backport-5.4/080-wireguard-0080-wireguard-socket-mark-skbs-as-not-on-list-when-recei.patch
blob: 779491c8db0e1f535da6df8caab314bff5b2e352 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: "Jason A. Donenfeld" <Jason@zx2c4.com>
Date: Thu, 2 Jan 2020 17:47:51 +0100
Subject: [PATCH] wireguard: socket: mark skbs as not on list when receiving
 via gro

commit 736775d06bac60d7a353e405398b48b2bd8b1e54 upstream.

Certain drivers will pass gro skbs to udp, at which point the udp driver
simply iterates through them and passes them off to encap_rcv, which is
where we pick up. At the moment, we're not attempting to coalesce these
into bundles, but we also don't want to wind up having cascaded lists of
skbs treated separately. The right behavior here, then, is to just mark
each incoming one as not on a list. This can be seen in practice, for
example, with Qualcomm's rmnet_perf driver.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Tested-by: Yaroslav Furman <yaro330@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
---
 drivers/net/wireguard/socket.c | 1 +
 1 file changed, 1 insertion(+)

--- a/drivers/net/wireguard/socket.c
+++ b/drivers/net/wireguard/socket.c
@@ -333,6 +333,7 @@ static int wg_receive(struct sock *sk, s
 	wg = sk->sk_user_data;
 	if (unlikely(!wg))
 		goto err;
+	skb_mark_not_on_list(skb);
 	wg_packet_receive(wg, skb);
 	return 0;