1 From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
2 From: "Jason A. Donenfeld" <Jason@zx2c4.com>
3 Date: Wed, 6 May 2020 15:33:04 -0600
4 Subject: [PATCH] wireguard: send/receive: cond_resched() when processing
7 commit 4005f5c3c9d006157ba716594e0d70c88a235c5e upstream.
9 Users with pathological hardware reported CPU stalls on CONFIG_
10 PREEMPT_VOLUNTARY=y, because the ringbuffers would stay full, meaning
11 these workers would never terminate. That turned out not to be okay on
12 systems without forced preemption, which Sultan observed. This commit
13 adds a cond_resched() to the bottom of each loop iteration, so that
14 these workers don't hog the core. Note that we don't need this on the
15 napi poll worker, since that terminates after its budget is expended.
17 Suggested-by: Sultan Alsawaf <sultan@kerneltoast.com>
18 Reported-by: Wang Jian <larkwang@gmail.com>
19 Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
20 Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
21 Signed-off-by: David S. Miller <davem@davemloft.net>
22 Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
24 drivers/net/wireguard/receive.c | 2 ++
25 drivers/net/wireguard/send.c | 4 ++++
26 2 files changed, 6 insertions(+)
28 --- a/drivers/net/wireguard/receive.c
29 +++ b/drivers/net/wireguard/receive.c
30 @@ -516,6 +516,8 @@ void wg_packet_decrypt_worker(struct wor
31 &PACKET_CB(skb)->keypair->receiving)) ?
32 PACKET_STATE_CRYPTED : PACKET_STATE_DEAD;
33 wg_queue_enqueue_per_peer_napi(skb, state);
39 --- a/drivers/net/wireguard/send.c
40 +++ b/drivers/net/wireguard/send.c
41 @@ -281,6 +281,8 @@ void wg_packet_tx_worker(struct work_str
43 wg_noise_keypair_put(keypair, false);
50 @@ -304,6 +306,8 @@ void wg_packet_encrypt_worker(struct wor
52 wg_queue_enqueue_per_peer(&PACKET_PEER(first)->tx_queue, first,