From dc01e7d3447793fd9e4090aa9d50c549848b5a18 Mon Sep 17 00:00:00 2001 From: Or Gerlitz Date: Mon, 20 Jan 2014 13:59:21 +0200 Subject: [PATCH] net: Add GRO support for vxlan traffic Add GRO handlers for vxlann, by using the UDP GRO infrastructure. For single TCP session that goes through vxlan tunneling I got nice improvement from 6.8Gbs to 11.5Gbs --> UDP/VXLAN GRO disabled $ netperf -H 192.168.52.147 -c -C $ netperf -t TCP_STREAM -H 192.168.52.147 -c -C MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.52.147 () port 0 AF_INET Recv Send Send Utilization Service Demand Socket Socket Message Elapsed Send Recv Send Recv Size Size Size Time Throughput local remote local remote bytes bytes bytes secs. 10^6bits/s % S % S us/KB us/KB 87380 65536 65536 10.00 6799.75 12.54 24.79 0.604 1.195 --> UDP/VXLAN GRO enabled $ netperf -t TCP_STREAM -H 192.168.52.147 -c -C MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.52.147 () port 0 AF_INET Recv Send Send Utilization Service Demand Socket Socket Message Elapsed Send Recv Send Recv Size Size Size Time Throughput local remote local remote bytes bytes bytes secs. 10^6bits/s % S % S us/KB us/KB 87380 65536 65536 10.00 11562.72 24.90 20.34 0.706 0.577 Signed-off-by: Shlomo Pongratz Signed-off-by: Or Gerlitz Signed-off-by: David S. Miller --- drivers/net/vxlan.c | 117 +++++++++++++++++++++++++++++++++++++++++--- include/net/vxlan.h | 1 + 2 files changed, 111 insertions(+), 7 deletions(-) diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c index 0fa4b9108e82..942acc20d394 100644 --- a/drivers/net/vxlan.c +++ b/drivers/net/vxlan.c @@ -40,6 +40,7 @@ #include #include #include +#include #if IS_ENABLED(CONFIG_IPV6) #include #include @@ -554,13 +555,106 @@ static int vxlan_fdb_append(struct vxlan_fdb *f, return 1; } +static struct sk_buff **vxlan_gro_receive(struct sk_buff **head, struct sk_buff *skb) +{ + struct sk_buff *p, **pp = NULL; + struct vxlanhdr *vh, *vh2; + struct ethhdr *eh, *eh2; + unsigned int hlen, off_vx, off_eth; + const struct packet_offload *ptype; + __be16 type; + int flush = 1; + + off_vx = skb_gro_offset(skb); + hlen = off_vx + sizeof(*vh); + vh = skb_gro_header_fast(skb, off_vx); + if (skb_gro_header_hard(skb, hlen)) { + vh = skb_gro_header_slow(skb, hlen, off_vx); + if (unlikely(!vh)) + goto out; + } + skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ + + off_eth = skb_gro_offset(skb); + hlen = off_eth + sizeof(*eh); + eh = skb_gro_header_fast(skb, off_eth); + if (skb_gro_header_hard(skb, hlen)) { + eh = skb_gro_header_slow(skb, hlen, off_eth); + if (unlikely(!eh)) + goto out; + } + + flush = 0; + + for (p = *head; p; p = p->next) { + if (!NAPI_GRO_CB(p)->same_flow) + continue; + + vh2 = (struct vxlanhdr *)(p->data + off_vx); + eh2 = (struct ethhdr *)(p->data + off_eth); + if (vh->vx_vni != vh2->vx_vni || compare_ether_header(eh, eh2)) { + NAPI_GRO_CB(p)->same_flow = 0; + continue; + } + goto found; + } + +found: + type = eh->h_proto; + + rcu_read_lock(); + ptype = gro_find_receive_by_type(type); + if (ptype == NULL) { + flush = 1; + goto out_unlock; + } + + skb_gro_pull(skb, sizeof(*eh)); /* pull inner eth header */ + pp = ptype->callbacks.gro_receive(head, skb); + +out_unlock: + rcu_read_unlock(); +out: + NAPI_GRO_CB(skb)->flush |= flush; + + return pp; +} + +static int vxlan_gro_complete(struct sk_buff *skb, int nhoff) +{ + struct ethhdr *eh; + struct packet_offload *ptype; + __be16 type; + int vxlan_len = sizeof(struct vxlanhdr) + sizeof(struct ethhdr); + int err = -ENOSYS; + + eh = (struct ethhdr *)(skb->data + nhoff + sizeof(struct vxlanhdr)); + type = eh->h_proto; + + rcu_read_lock(); + ptype = gro_find_complete_by_type(type); + if (ptype != NULL) + err = ptype->callbacks.gro_complete(skb, nhoff + vxlan_len); + + rcu_read_unlock(); + return err; +} + /* Notify netdevs that UDP port started listening */ -static void vxlan_notify_add_rx_port(struct sock *sk) +static void vxlan_notify_add_rx_port(struct vxlan_sock *vs) { struct net_device *dev; + struct sock *sk = vs->sock->sk; struct net *net = sock_net(sk); sa_family_t sa_family = sk->sk_family; __be16 port = inet_sk(sk)->inet_sport; + int err; + + if (sa_family == AF_INET) { + err = udp_add_offload(&vs->udp_offloads); + if (err) + pr_warn("vxlan: udp_add_offload failed with status %d\n", err); + } rcu_read_lock(); for_each_netdev_rcu(net, dev) { @@ -572,9 +666,10 @@ static void vxlan_notify_add_rx_port(struct sock *sk) } /* Notify netdevs that UDP port is no more listening */ -static void vxlan_notify_del_rx_port(struct sock *sk) +static void vxlan_notify_del_rx_port(struct vxlan_sock *vs) { struct net_device *dev; + struct sock *sk = vs->sock->sk; struct net *net = sock_net(sk); sa_family_t sa_family = sk->sk_family; __be16 port = inet_sk(sk)->inet_sport; @@ -586,6 +681,9 @@ static void vxlan_notify_del_rx_port(struct sock *sk) port); } rcu_read_unlock(); + + if (sa_family == AF_INET) + udp_del_offload(&vs->udp_offloads); } /* Add new entry to forwarding table -- assumes lock held */ @@ -963,7 +1061,7 @@ void vxlan_sock_release(struct vxlan_sock *vs) spin_lock(&vn->sock_lock); hlist_del_rcu(&vs->hlist); rcu_assign_sk_user_data(vs->sock->sk, NULL); - vxlan_notify_del_rx_port(sk); + vxlan_notify_del_rx_port(vs); spin_unlock(&vn->sock_lock); queue_work(vxlan_wq, &vs->del_work); @@ -1124,8 +1222,8 @@ static void vxlan_rcv(struct vxlan_sock *vs, * leave the CHECKSUM_UNNECESSARY, the device checksummed it * for us. Otherwise force the upper layers to verify it. */ - if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation || - !(vxlan->dev->features & NETIF_F_RXCSUM)) + if ((skb->ip_summed != CHECKSUM_UNNECESSARY && skb->ip_summed != CHECKSUM_PARTIAL) || + !skb->encapsulation || !(vxlan->dev->features & NETIF_F_RXCSUM)) skb->ip_summed = CHECKSUM_NONE; skb->encapsulation = 0; @@ -2303,7 +2401,7 @@ static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port, struct sock *sk; unsigned int h; - vs = kmalloc(sizeof(*vs), GFP_KERNEL); + vs = kzalloc(sizeof(*vs), GFP_KERNEL); if (!vs) return ERR_PTR(-ENOMEM); @@ -2328,9 +2426,14 @@ static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port, vs->data = data; rcu_assign_sk_user_data(vs->sock->sk, vs); + /* Initialize the vxlan udp offloads structure */ + vs->udp_offloads.port = port; + vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive; + vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete; + spin_lock(&vn->sock_lock); hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); - vxlan_notify_add_rx_port(sk); + vxlan_notify_add_rx_port(vs); spin_unlock(&vn->sock_lock); /* Mark socket as an encapsulation socket. */ diff --git a/include/net/vxlan.h b/include/net/vxlan.h index 6b6d180fb91a..5deef1ae78c9 100644 --- a/include/net/vxlan.h +++ b/include/net/vxlan.h @@ -21,6 +21,7 @@ struct vxlan_sock { struct rcu_head rcu; struct hlist_head vni_list[VNI_HASH_SIZE]; atomic_t refcnt; + struct udp_offload udp_offloads; }; struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port, -- 2.20.1