2 * Copyright (c) 2007-2012 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
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 #include <linux/skbuff.h>
24 #include <linux/if_tunnel.h>
25 #include <linux/if_vlan.h>
30 #include <net/protocol.h>
35 #include "vport-generic.h"
38 * The GRE header is composed of a series of sections: a base and then a variable
41 #define GRE_HEADER_SECTION 4
48 static void get_gre_param(const struct tnl_mutable_config *mutable,
49 const struct ovs_key_ipv4_tunnel *tun_key,
50 u32 *flags, u32 *tunnel_type, __be64 *out_key)
52 if (tun_key->ipv4_dst) {
55 if (tun_key->tun_flags & OVS_FLOW_TNL_F_KEY)
56 *flags = TNL_F_OUT_KEY_ACTION;
57 if (tun_key->tun_flags & OVS_FLOW_TNL_F_CSUM)
59 *tunnel_type = TNL_T_PROTO_GRE;
60 *out_key = tun_key->tun_id;
62 *flags = mutable->flags;
63 *tunnel_type = mutable->key.tunnel_type;
64 if (mutable->flags & TNL_F_OUT_KEY_ACTION)
65 *out_key = tun_key->tun_id;
67 *out_key = mutable->out_key;
72 static int gre_hdr_len(const struct tnl_mutable_config *mutable,
73 const struct ovs_key_ipv4_tunnel *tun_key)
80 get_gre_param(mutable, tun_key, &flags, &tunnel_type, &out_key);
81 len = GRE_HEADER_SECTION;
83 if (flags & TNL_F_CSUM)
84 len += GRE_HEADER_SECTION;
86 /* Set key for GRE64 tunnels, even when key if is zero. */
88 tunnel_type & TNL_T_PROTO_GRE64 ||
89 flags & TNL_F_OUT_KEY_ACTION) {
91 len += GRE_HEADER_SECTION;
92 if (tunnel_type & TNL_T_PROTO_GRE64)
93 len += GRE_HEADER_SECTION;
99 /* Returns the least-significant 32 bits of a __be64. */
100 static __be32 be64_get_low32(__be64 x)
103 return (__force __be32)x;
105 return (__force __be32)((__force u64)x >> 32);
109 static __be32 be64_get_high32(__be64 x)
112 return (__force __be32)((__force u64)x >> 32);
114 return (__force __be32)x;
118 static struct sk_buff *gre_build_header(const struct vport *vport,
119 const struct tnl_mutable_config *mutable,
120 struct dst_entry *dst,
127 const struct ovs_key_ipv4_tunnel *tun_key = OVS_CB(skb)->tun_key;
128 __be32 *options = (__be32 *)(skb_network_header(skb) + tunnel_hlen
129 - GRE_HEADER_SECTION);
130 struct gre_base_hdr *greh = (struct gre_base_hdr *) skb_transport_header(skb);
132 get_gre_param(mutable, tun_key, &flags, &tunnel_type, &out_key);
134 greh->protocol = htons(ETH_P_TEB);
137 /* Work backwards over the options so the checksum is last. */
138 if (out_key || flags & TNL_F_OUT_KEY_ACTION || tunnel_type & TNL_T_PROTO_GRE64) {
139 greh->flags |= GRE_KEY;
140 if (tunnel_type & TNL_T_PROTO_GRE64) {
141 /* Set higher 32 bits to seq. */
142 *options = be64_get_high32(out_key);
144 greh->flags |= GRE_SEQ;
146 *options = be64_get_low32(out_key);
150 if (flags & TNL_F_CSUM) {
151 greh->flags |= GRE_CSUM;
153 *(__sum16 *)options = csum_fold(skb_checksum(skb,
154 skb_transport_offset(skb),
155 skb->len - skb_transport_offset(skb),
159 * Allow our local IP stack to fragment the outer packet even if the
160 * DF bit is set as a last resort. We also need to force selection of
161 * an IP ID here because Linux will otherwise leave it at 0 if the
162 * packet originally had DF set.
165 __ip_select_ident(ip_hdr(skb), dst, 0);
170 static __be64 key_to_tunnel_id(__be32 key, __be32 seq)
173 return (__force __be64)((__force u64)seq << 32 | (__force u32)key);
175 return (__force __be64)((__force u64)key << 32 | (__force u32)seq);
179 static int parse_header(struct iphdr *iph, __be16 *flags, __be64 *tun_id,
182 /* IP and ICMP protocol handlers check that the IHL is valid. */
183 struct gre_base_hdr *greh = (struct gre_base_hdr *)((u8 *)iph + (iph->ihl << 2));
184 __be32 *options = (__be32 *)(greh + 1);
187 *flags = greh->flags;
189 if (unlikely(greh->flags & (GRE_VERSION | GRE_ROUTING)))
192 if (unlikely(greh->protocol != htons(ETH_P_TEB)))
195 hdr_len = GRE_HEADER_SECTION;
197 if (greh->flags & GRE_CSUM) {
198 hdr_len += GRE_HEADER_SECTION;
202 if (greh->flags & GRE_KEY) {
207 hdr_len += GRE_HEADER_SECTION;
210 if (greh->flags & GRE_SEQ) {
212 *tunnel_type = TNL_T_PROTO_GRE64;
215 *tunnel_type = TNL_T_PROTO_GRE;
217 *tun_id = key_to_tunnel_id(gre_key, seq);
220 /* Ignore GRE seq if there is no key present. */
221 *tunnel_type = TNL_T_PROTO_GRE;
224 if (greh->flags & GRE_SEQ)
225 hdr_len += GRE_HEADER_SECTION;
230 /* Called with rcu_read_lock and BH disabled. */
231 static void gre_err(struct sk_buff *skb, u32 info)
234 const struct tnl_mutable_config *mutable;
235 const int type = icmp_hdr(skb)->type;
236 const int code = icmp_hdr(skb)->code;
237 int mtu = ntohs(icmp_hdr(skb)->un.frag.mtu);
243 int tunnel_hdr_len, tot_hdr_len;
244 unsigned int orig_mac_header;
245 unsigned int orig_nw_header;
247 if (type != ICMP_DEST_UNREACH || code != ICMP_FRAG_NEEDED)
251 * The mimimum size packet that we would actually be able to process:
252 * encapsulating IP header, minimum GRE header, Ethernet header,
255 if (!pskb_may_pull(skb, sizeof(struct iphdr) + GRE_HEADER_SECTION +
256 ETH_HLEN + sizeof(struct iphdr)))
259 iph = (struct iphdr *)skb->data;
260 if (ipv4_is_multicast(iph->daddr))
263 tunnel_hdr_len = parse_header(iph, &flags, &key, &tunnel_type);
264 if (tunnel_hdr_len < 0)
267 vport = ovs_tnl_find_port(dev_net(skb->dev), iph->saddr, iph->daddr, key,
268 tunnel_type, &mutable);
273 * Packets received by this function were previously sent by us, so
274 * any comparisons should be to the output values, not the input.
275 * However, it's not really worth it to have a hash table based on
276 * output keys (especially since ICMP error handling of tunneled packets
277 * isn't that reliable anyways). Therefore, we do a lookup based on the
278 * out key as if it were the in key and then check to see if the input
279 * and output keys are the same.
281 if (mutable->key.in_key != mutable->out_key)
284 if (!!(mutable->flags & TNL_F_IN_KEY_MATCH) !=
285 !!(mutable->flags & TNL_F_OUT_KEY_ACTION))
288 if ((mutable->flags & TNL_F_CSUM) && !(flags & GRE_CSUM))
291 tunnel_hdr_len += iph->ihl << 2;
293 orig_mac_header = skb_mac_header(skb) - skb->data;
294 orig_nw_header = skb_network_header(skb) - skb->data;
295 skb_set_mac_header(skb, tunnel_hdr_len);
297 tot_hdr_len = tunnel_hdr_len + ETH_HLEN;
299 skb->protocol = eth_hdr(skb)->h_proto;
300 if (skb->protocol == htons(ETH_P_8021Q)) {
301 tot_hdr_len += VLAN_HLEN;
302 skb->protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
305 skb_set_network_header(skb, tot_hdr_len);
308 if (skb->protocol == htons(ETH_P_IP))
309 tot_hdr_len += sizeof(struct iphdr);
310 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
311 else if (skb->protocol == htons(ETH_P_IPV6))
312 tot_hdr_len += sizeof(struct ipv6hdr);
317 if (!pskb_may_pull(skb, tot_hdr_len))
320 if (skb->protocol == htons(ETH_P_IP)) {
321 if (mtu < IP_MIN_MTU) {
322 if (ntohs(ip_hdr(skb)->tot_len) >= IP_MIN_MTU)
329 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
330 else if (skb->protocol == htons(ETH_P_IPV6)) {
331 if (mtu < IPV6_MIN_MTU) {
332 unsigned int packet_length = sizeof(struct ipv6hdr) +
333 ntohs(ipv6_hdr(skb)->payload_len);
335 if (packet_length >= IPV6_MIN_MTU
336 || ntohs(ipv6_hdr(skb)->payload_len) == 0)
344 __skb_pull(skb, tunnel_hdr_len);
345 ovs_tnl_frag_needed(vport, mutable, skb, mtu);
346 __skb_push(skb, tunnel_hdr_len);
349 skb_set_mac_header(skb, orig_mac_header);
350 skb_set_network_header(skb, orig_nw_header);
351 skb->protocol = htons(ETH_P_IP);
354 static bool check_checksum(struct sk_buff *skb)
356 struct iphdr *iph = ip_hdr(skb);
357 struct gre_base_hdr *greh = (struct gre_base_hdr *)(iph + 1);
360 if (greh->flags & GRE_CSUM) {
361 switch (skb->ip_summed) {
362 case CHECKSUM_COMPLETE:
363 csum = csum_fold(skb->csum);
371 csum = __skb_checksum_complete(skb);
372 skb->ip_summed = CHECKSUM_COMPLETE;
380 static u32 gre_flags_to_tunnel_flags(const struct tnl_mutable_config *mutable,
383 u32 tunnel_flags = 0;
385 if (gre_flags & GRE_KEY) {
386 if (mutable->key.daddr && (mutable->flags & TNL_F_IN_KEY_MATCH))
387 tunnel_flags = OVS_FLOW_TNL_F_KEY;
388 else if (!mutable->key.daddr)
389 tunnel_flags = OVS_FLOW_TNL_F_KEY;
392 if (gre_flags & GRE_CSUM)
393 tunnel_flags |= OVS_FLOW_TNL_F_CSUM;
398 /* Called with rcu_read_lock and BH disabled. */
399 static int gre_rcv(struct sk_buff *skb)
402 const struct tnl_mutable_config *mutable;
405 struct ovs_key_ipv4_tunnel tun_key;
410 if (unlikely(!pskb_may_pull(skb, sizeof(struct gre_base_hdr) + ETH_HLEN)))
412 if (unlikely(!check_checksum(skb)))
415 hdr_len = parse_header(ip_hdr(skb), &flags, &key, &tunnel_type);
416 if (unlikely(hdr_len < 0))
419 if (unlikely(!pskb_may_pull(skb, hdr_len + ETH_HLEN)))
423 vport = ovs_tnl_find_port(dev_net(skb->dev), iph->daddr, iph->saddr, key,
424 tunnel_type, &mutable);
425 if (unlikely(!vport)) {
426 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
430 tnl_tun_key_init(&tun_key, iph, key, gre_flags_to_tunnel_flags(mutable, flags));
431 OVS_CB(skb)->tun_key = &tun_key;
433 __skb_pull(skb, hdr_len);
434 skb_postpull_rcsum(skb, skb_transport_header(skb), hdr_len + ETH_HLEN);
436 ovs_tnl_rcv(vport, skb);
444 static const struct tnl_ops gre_tnl_ops = {
445 .tunnel_type = TNL_T_PROTO_GRE,
446 .ipproto = IPPROTO_GRE,
447 .hdr_len = gre_hdr_len,
448 .build_header = gre_build_header,
451 static struct vport *gre_create(const struct vport_parms *parms)
453 return ovs_tnl_create(parms, &ovs_gre_vport_ops, &gre_tnl_ops);
456 static struct vport *gre_create_ft(const struct vport_parms *parms)
458 return ovs_tnl_create(parms, &ovs_gre_ft_vport_ops, &gre_tnl_ops);
461 static const struct tnl_ops gre64_tnl_ops = {
462 .tunnel_type = TNL_T_PROTO_GRE64,
463 .ipproto = IPPROTO_GRE,
464 .hdr_len = gre_hdr_len,
465 .build_header = gre_build_header,
468 static struct vport *gre_create64(const struct vport_parms *parms)
470 return ovs_tnl_create(parms, &ovs_gre64_vport_ops, &gre64_tnl_ops);
473 static const struct net_protocol gre_protocol_handlers = {
475 .err_handler = gre_err,
476 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32)
483 static int gre_init(void)
491 err = inet_add_protocol(&gre_protocol_handlers, IPPROTO_GRE);
493 pr_warn("cannot register gre protocol handler\n");
498 static void gre_exit(void)
505 inet_del_protocol(&gre_protocol_handlers, IPPROTO_GRE);
508 const struct vport_ops ovs_gre_ft_vport_ops = {
509 .type = OVS_VPORT_TYPE_FT_GRE,
510 .flags = VPORT_F_TUN_ID,
513 .create = gre_create_ft,
514 .destroy = ovs_tnl_destroy,
515 .set_addr = ovs_tnl_set_addr,
516 .get_name = ovs_tnl_get_name,
517 .get_addr = ovs_tnl_get_addr,
518 .get_options = ovs_tnl_get_options,
519 .set_options = ovs_tnl_set_options,
520 .get_dev_flags = ovs_vport_gen_get_dev_flags,
521 .is_running = ovs_vport_gen_is_running,
522 .get_operstate = ovs_vport_gen_get_operstate,
523 .send = ovs_tnl_send,
526 const struct vport_ops ovs_gre_vport_ops = {
527 .type = OVS_VPORT_TYPE_GRE,
528 .flags = VPORT_F_TUN_ID,
531 .create = gre_create,
532 .destroy = ovs_tnl_destroy,
533 .set_addr = ovs_tnl_set_addr,
534 .get_name = ovs_tnl_get_name,
535 .get_addr = ovs_tnl_get_addr,
536 .get_options = ovs_tnl_get_options,
537 .set_options = ovs_tnl_set_options,
538 .get_dev_flags = ovs_vport_gen_get_dev_flags,
539 .is_running = ovs_vport_gen_is_running,
540 .get_operstate = ovs_vport_gen_get_operstate,
541 .send = ovs_tnl_send,
544 const struct vport_ops ovs_gre64_vport_ops = {
545 .type = OVS_VPORT_TYPE_GRE64,
546 .flags = VPORT_F_TUN_ID,
549 .create = gre_create64,
550 .destroy = ovs_tnl_destroy,
551 .set_addr = ovs_tnl_set_addr,
552 .get_name = ovs_tnl_get_name,
553 .get_addr = ovs_tnl_get_addr,
554 .get_options = ovs_tnl_get_options,
555 .set_options = ovs_tnl_set_options,
556 .get_dev_flags = ovs_vport_gen_get_dev_flags,
557 .is_running = ovs_vport_gen_is_running,
558 .get_operstate = ovs_vport_gen_get_operstate,
559 .send = ovs_tnl_send,