| b.liu | e958203 | 2025-04-17 19:18:16 +0800 | [diff] [blame] | 1 | From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001 |
| 2 | From: "Jason A. Donenfeld" <Jason@zx2c4.com> |
| 3 | Date: Thu, 2 Jan 2020 17:47:51 +0100 |
| 4 | Subject: [PATCH] wireguard: socket: mark skbs as not on list when receiving |
| 5 | via gro |
| 6 | |
| 7 | commit 736775d06bac60d7a353e405398b48b2bd8b1e54 upstream. |
| 8 | |
| 9 | Certain drivers will pass gro skbs to udp, at which point the udp driver |
| 10 | simply iterates through them and passes them off to encap_rcv, which is |
| 11 | where we pick up. At the moment, we're not attempting to coalesce these |
| 12 | into bundles, but we also don't want to wind up having cascaded lists of |
| 13 | skbs treated separately. The right behavior here, then, is to just mark |
| 14 | each incoming one as not on a list. This can be seen in practice, for |
| 15 | example, with Qualcomm's rmnet_perf driver. |
| 16 | |
| 17 | Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> |
| 18 | Tested-by: Yaroslav Furman <yaro330@gmail.com> |
| 19 | Signed-off-by: David S. Miller <davem@davemloft.net> |
| 20 | Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> |
| 21 | --- |
| 22 | drivers/net/wireguard/socket.c | 1 + |
| 23 | 1 file changed, 1 insertion(+) |
| 24 | |
| 25 | --- a/drivers/net/wireguard/socket.c |
| 26 | +++ b/drivers/net/wireguard/socket.c |
| 27 | @@ -333,6 +333,7 @@ static int wg_receive(struct sock *sk, s |
| 28 | wg = sk->sk_user_data; |
| 29 | if (unlikely(!wg)) |
| 30 | goto err; |
| 31 | + skb_mark_not_on_list(skb); |
| 32 | wg_packet_receive(wg, skb); |
| 33 | return 0; |
| 34 | |