2 * Copyright (c) 2007-2013 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
18 * This code is derived from kernel vxlan module.
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 #include <linux/kernel.h>
24 #include <linux/types.h>
25 #include <linux/module.h>
26 #include <linux/errno.h>
27 #include <linux/slab.h>
28 #include <linux/skbuff.h>
29 #include <linux/rculist.h>
30 #include <linux/netdevice.h>
33 #include <linux/udp.h>
34 #include <linux/igmp.h>
35 #include <linux/etherdevice.h>
36 #include <linux/if_ether.h>
37 #include <linux/if_vlan.h>
38 #include <linux/hash.h>
39 #include <linux/ethtool.h>
41 #include <net/ndisc.h>
43 #include <net/ip_tunnels.h>
46 #include <net/rtnetlink.h>
47 #include <net/route.h>
48 #include <net/dsfield.h>
49 #include <net/inet_ecn.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 #include <net/vxlan.h>
59 #define PORT_HASH_BITS 8
60 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
62 #define VXLAN_N_VID (1u << 24)
63 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
64 /* IP header + UDP + VXLAN + Ethernet header */
65 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
66 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
68 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
70 /* VXLAN protocol header */
76 static int vxlan_net_id;
78 /* per-network namespace private data for this module */
80 struct hlist_head sock_list[PORT_HASH_SIZE];
81 struct mutex sock_lock; /* RTNL lock nests inside this lock. */
84 /* Socket hash table head */
85 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
87 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
89 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
92 /* Find VXLAN socket based on network namespace and UDP port */
93 static struct vxlan_sock *vxlan_find_port(struct net *net, __be16 port)
95 struct vxlan_sock *vs;
97 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
98 if (inet_sport(vs->sock->sk) == port)
104 /* Callback from net/ipv4/udp.c to receive packets */
105 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
107 struct vxlan_handler *vh;
108 struct vxlan_sock *vs;
109 struct vxlanhdr *vxh;
111 /* Need Vxlan and inner Ethernet header to be present */
112 if (!pskb_may_pull(skb, VXLAN_HLEN))
115 /* Return packets with reserved bits set */
116 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
117 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
118 (vxh->vx_vni & htonl(0xff))) {
119 pr_warn("invalid vxlan flags=%#x vni=%#x\n",
120 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
124 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
127 vs = vxlan_find_port(sock_net(sk), inet_sport(sk));
131 list_for_each_entry_rcu(vh, &vs->handler_list, node) {
132 if (vh->rcv(vh, skb, vxh->vx_vni) == PACKET_RCVD)
137 /* Consume bad packet */
142 /* Return non vxlan pkt */
146 static void vxlan_sock_put(struct sk_buff *skb)
151 /* On transmit, associate with the tunnel socket */
152 static void vxlan_set_owner(struct sock *sk, struct sk_buff *skb)
157 skb->destructor = vxlan_sock_put;
160 /* Compute source port for outgoing packet
161 * first choice to use L4 flow hash since it will spread
162 * better and maybe available from hardware
163 * secondary choice is to use jhash on the Ethernet header
165 __be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
167 unsigned int range = (port_max - port_min) + 1;
170 hash = skb_get_rxhash(skb);
172 hash = jhash(skb->data, 2 * ETH_ALEN,
173 (__force u32) skb->protocol);
175 return htons((((u64) hash * range) >> 32) + port_min);
178 static void vxlan_gso(struct sk_buff *skb)
180 int udp_offset = skb_transport_offset(skb);
184 uh->len = htons(skb->len - udp_offset);
186 /* csum segment if tunnel sets skb with csum. */
187 if (unlikely(uh->check)) {
188 struct iphdr *iph = ip_hdr(skb);
190 uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
191 skb->len - udp_offset,
193 uh->check = csum_fold(skb_checksum(skb, udp_offset,
194 skb->len - udp_offset, 0));
197 uh->check = CSUM_MANGLED_0;
200 skb->ip_summed = CHECKSUM_NONE;
203 static int handle_offloads(struct sk_buff *skb)
205 if (skb_is_gso(skb)) {
206 OVS_GSO_CB(skb)->fix_segment = vxlan_gso;
208 if (skb->ip_summed != CHECKSUM_PARTIAL)
209 skb->ip_summed = CHECKSUM_NONE;
214 int vxlan_xmit_skb(struct net *net, struct vxlan_handler *vh,
215 struct rtable *rt, struct sk_buff *skb,
216 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
217 __be16 src_port, __be16 dst_port, __be32 vni)
219 struct vxlanhdr *vxh;
224 skb_reset_inner_headers(skb);
226 min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
227 + VXLAN_HLEN + sizeof(struct iphdr)
228 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
230 /* Need space for new headers (invalidates iph ptr) */
231 err = skb_cow_head(skb, min_headroom);
235 if (unlikely(vlan_deaccel_tag(skb)))
238 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
239 vxh->vx_flags = htonl(VXLAN_FLAGS);
242 __skb_push(skb, sizeof(*uh));
243 skb_reset_transport_header(skb);
247 uh->source = src_port;
249 uh->len = htons(skb->len);
252 vxlan_set_owner(vh->vs->sock->sk, skb);
254 err = handle_offloads(skb);
258 return iptunnel_xmit(net, rt, skb, src, dst,
259 IPPROTO_UDP, tos, ttl, df);
262 static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port)
264 struct vxlan_sock *vs;
266 struct sockaddr_in vxlan_addr = {
267 .sin_family = AF_INET,
268 .sin_addr.s_addr = htonl(INADDR_ANY),
273 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
275 return ERR_PTR(-ENOMEM);
277 /* Create UDP socket for encapsulation receive. */
278 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &vs->sock);
280 pr_debug("UDP socket create failed\n");
285 /* Put in proper namespace */
287 sk_change_net(sk, net);
289 rc = kernel_bind(vs->sock, (struct sockaddr *) &vxlan_addr,
292 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
293 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
294 sk_release_kernel(sk);
299 /* Disable multicast loopback */
300 inet_sk(sk)->mc_loop = 0;
301 INIT_LIST_HEAD(&vs->handler_list);
302 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
304 /* Mark socket as an encapsulation socket. */
305 udp_sk(sk)->encap_type = 1;
306 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
312 static void rcu_free_vs_callback(struct rcu_head *rcu)
314 struct vxlan_sock *vs = container_of(rcu, struct vxlan_sock, rcu);
319 static void vxlan_socket_del(struct vxlan_sock *vs)
321 if (list_empty(&vs->handler_list)) {
322 hlist_del_rcu(&vs->hlist);
324 sk_release_kernel(vs->sock->sk);
325 call_rcu(&vs->rcu, rcu_free_vs_callback);
329 static int vxlan_init_module(void);
330 static void vxlan_cleanup_module(void);
332 static void rcu_free_vh_callback(struct rcu_head *rcu)
334 struct vxlan_handler *vh = container_of(rcu, struct vxlan_handler, rcu);
339 static void vh_del_work(struct work_struct *work)
341 struct vxlan_handler *vh = container_of(work, struct vxlan_handler, del_work);
342 struct vxlan_sock *vs = vh->vs;
343 struct net *net = sock_net(vs->sock->sk);
344 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
346 mutex_lock(&vn->sock_lock);
348 list_del_rcu(&vh->node);
349 call_rcu(&vh->rcu, rcu_free_vh_callback);
350 vxlan_socket_del(vs);
352 mutex_unlock(&vn->sock_lock);
354 vxlan_cleanup_module();
357 struct vxlan_handler *vxlan_handler_add(struct net *net,
358 __be16 portno, vxlan_rcv_t *rcv,
359 void *data, int priority, bool create)
361 struct vxlan_net *vn;
362 struct vxlan_sock *vs;
363 struct vxlan_handler *vh;
364 struct vxlan_handler *new;
367 err = vxlan_init_module();
371 vn = net_generic(net, vxlan_net_id);
372 mutex_lock(&vn->sock_lock);
373 /* Look to see if can reuse socket */
374 vs = vxlan_find_port(net, portno);
376 vs = vxlan_socket_create(net, portno);
383 /* Try existing vxlan hanlders for this socket. */
384 list_for_each_entry(vh, &vs->handler_list, node) {
385 if (vh->rcv == rcv) {
387 vxlan_socket_del(vs);
388 new = ERR_PTR(-EEXIST);
391 atomic_inc(&vh->refcnt);
397 new = kzalloc(sizeof(*new), GFP_KERNEL);
399 vxlan_socket_del(vs);
400 new = ERR_PTR(-ENOMEM);
406 atomic_set(&new->refcnt, 1);
407 INIT_WORK(&new->del_work, vh_del_work);
409 new->priority = priority;
411 list_for_each_entry(vh, &vs->handler_list, node) {
412 if (vh->priority > priority) {
413 list_add_tail_rcu(&new->node, &vh->node);
418 list_add_tail_rcu(&new->node, &vs->handler_list);
420 mutex_unlock(&vn->sock_lock);
424 void vxlan_handler_put(struct vxlan_handler *vh)
428 if (atomic_dec_and_test(&vh->refcnt))
429 queue_work(&vh->del_work);
432 static int vxlan_init_net(struct net *net)
434 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
437 mutex_init(&vn->sock_lock);
439 for (h = 0; h < PORT_HASH_SIZE; ++h)
440 INIT_HLIST_HEAD(&vn->sock_list[h]);
445 static struct pernet_operations vxlan_net_ops = {
446 .init = vxlan_init_net,
448 .size = sizeof(struct vxlan_net),
452 static DEFINE_MUTEX(init_lock);
453 DEFINE_COMPAT_PNET_REG_FUNC(device);
455 static int vxlan_init_module(void)
459 mutex_lock(&init_lock);
462 err = register_pernet_device(&vxlan_net_ops);
466 mutex_unlock(&init_lock);
470 static void vxlan_cleanup_module(void)
472 mutex_lock(&init_lock);
476 unregister_pernet_device(&vxlan_net_ops);
478 mutex_unlock(&init_lock);