The socket timestamping option SOF_TIMESTAMPING_OPT_PKTINFO needs the skb napi_id in order to provide the index of the device that captured the hardware receive timestamp. However, wireguard resets most of the skb headers, including the napi_id, which prevents the timestamping option from working as expected. The missing index prevents applications that rely on it (e.g. chrony) from processing hardware receive timestamps (unlike with transmit timestamps looped back to the error queue, where the IP_PKTINFO index identifies the device that captured the timestamp). Preserve the napi_id in wg_reset_packet() of received packets in order to make the timestamping option useful with wireguard tunnels and enable highly accurate synchronization. Cc: Jason A. Donenfeld Cc: Willem de Bruijn Signed-off-by: Miroslav Lichvar --- Notes: v2: - don't copy napi_id if only CONFIG_XPS is defined - improve commit message drivers/net/wireguard/queueing.h | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/net/wireguard/queueing.h b/drivers/net/wireguard/queueing.h index 79b6d70de236..a5b76cecf429 100644 --- a/drivers/net/wireguard/queueing.h +++ b/drivers/net/wireguard/queueing.h @@ -75,6 +75,7 @@ static inline bool wg_check_packet_protocol(struct sk_buff *skb) static inline void wg_reset_packet(struct sk_buff *skb, bool encapsulating) { + unsigned int napi_id = skb_napi_id(skb); u8 l4_hash = skb->l4_hash; u8 sw_hash = skb->sw_hash; u32 hash = skb->hash; @@ -84,6 +85,10 @@ static inline void wg_reset_packet(struct sk_buff *skb, bool encapsulating) skb->l4_hash = l4_hash; skb->sw_hash = sw_hash; skb->hash = hash; + } else { +#ifdef CONFIG_NET_RX_BUSY_POLL + skb->napi_id = napi_id; +#endif } skb->queue_mapping = 0; skb->nohdr = 0; -- 2.51.0