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 void gre_build_header(const struct vport *vport,
119 const struct tnl_mutable_config *mutable,
120 const struct ovs_key_ipv4_tunnel *tun_key,
123 struct gre_base_hdr *greh = header;
124 __be32 *options = (__be32 *)(greh + 1);
129 get_gre_param(mutable, tun_key, &flags, &tunnel_type, &out_key);
131 greh->protocol = htons(ETH_P_TEB);
134 if (flags & TNL_F_CSUM) {
135 greh->flags |= GRE_CSUM;
140 if (flags & TNL_F_OUT_KEY_ACTION) {
141 greh->flags |= GRE_KEY;
142 if (tunnel_type & TNL_T_PROTO_GRE64)
143 greh->flags |= GRE_SEQ;
145 } else if (out_key ||
146 tunnel_type & TNL_T_PROTO_GRE64) {
147 greh->flags |= GRE_KEY;
148 *options = be64_get_low32(out_key);
149 if (tunnel_type & TNL_T_PROTO_GRE64) {
151 *options = be64_get_high32(out_key);
152 greh->flags |= GRE_SEQ;
157 static struct sk_buff *gre_update_header(const struct vport *vport,
158 const struct tnl_mutable_config *mutable,
159 struct dst_entry *dst,
166 const struct ovs_key_ipv4_tunnel *tun_key = OVS_CB(skb)->tun_key;
167 __be32 *options = (__be32 *)(skb_network_header(skb) + tunnel_hlen
168 - GRE_HEADER_SECTION);
170 get_gre_param(mutable, tun_key, &flags, &tunnel_type, &out_key);
172 /* Work backwards over the options so the checksum is last. */
173 if (flags & TNL_F_OUT_KEY_ACTION) {
174 if (tunnel_type & TNL_T_PROTO_GRE64) {
175 /* Set higher 32 bits to seq. */
176 *options = be64_get_high32(out_key);
179 *options = be64_get_low32(out_key);
181 } else if (out_key || tunnel_type & TNL_T_PROTO_GRE64) {
183 if (tunnel_type & TNL_T_PROTO_GRE64)
187 if (flags & TNL_F_CSUM)
188 *(__sum16 *)options = csum_fold(skb_checksum(skb,
189 skb_transport_offset(skb),
190 skb->len - skb_transport_offset(skb),
193 * Allow our local IP stack to fragment the outer packet even if the
194 * DF bit is set as a last resort. We also need to force selection of
195 * an IP ID here because Linux will otherwise leave it at 0 if the
196 * packet originally had DF set.
199 __ip_select_ident(ip_hdr(skb), dst, 0);
204 static __be64 key_to_tunnel_id(__be32 key, __be32 seq)
207 return (__force __be64)((__force u64)seq << 32 | (__force u32)key);
209 return (__force __be64)((__force u64)key << 32 | (__force u32)seq);
213 static int parse_header(struct iphdr *iph, __be16 *flags, __be64 *tun_id,
216 /* IP and ICMP protocol handlers check that the IHL is valid. */
217 struct gre_base_hdr *greh = (struct gre_base_hdr *)((u8 *)iph + (iph->ihl << 2));
218 __be32 *options = (__be32 *)(greh + 1);
221 *flags = greh->flags;
223 if (unlikely(greh->flags & (GRE_VERSION | GRE_ROUTING)))
226 if (unlikely(greh->protocol != htons(ETH_P_TEB)))
229 hdr_len = GRE_HEADER_SECTION;
231 if (greh->flags & GRE_CSUM) {
232 hdr_len += GRE_HEADER_SECTION;
236 if (greh->flags & GRE_KEY) {
241 hdr_len += GRE_HEADER_SECTION;
244 if (greh->flags & GRE_SEQ) {
246 *tunnel_type = TNL_T_PROTO_GRE64;
249 *tunnel_type = TNL_T_PROTO_GRE;
251 *tun_id = key_to_tunnel_id(gre_key, seq);
254 /* Ignore GRE seq if there is no key present. */
255 *tunnel_type = TNL_T_PROTO_GRE;
258 if (greh->flags & GRE_SEQ)
259 hdr_len += GRE_HEADER_SECTION;
264 /* Called with rcu_read_lock and BH disabled. */
265 static void gre_err(struct sk_buff *skb, u32 info)
268 const struct tnl_mutable_config *mutable;
269 const int type = icmp_hdr(skb)->type;
270 const int code = icmp_hdr(skb)->code;
271 int mtu = ntohs(icmp_hdr(skb)->un.frag.mtu);
277 int tunnel_hdr_len, tot_hdr_len;
278 unsigned int orig_mac_header;
279 unsigned int orig_nw_header;
281 if (type != ICMP_DEST_UNREACH || code != ICMP_FRAG_NEEDED)
285 * The mimimum size packet that we would actually be able to process:
286 * encapsulating IP header, minimum GRE header, Ethernet header,
289 if (!pskb_may_pull(skb, sizeof(struct iphdr) + GRE_HEADER_SECTION +
290 ETH_HLEN + sizeof(struct iphdr)))
293 iph = (struct iphdr *)skb->data;
294 if (ipv4_is_multicast(iph->daddr))
297 tunnel_hdr_len = parse_header(iph, &flags, &key, &tunnel_type);
298 if (tunnel_hdr_len < 0)
301 vport = ovs_tnl_find_port(dev_net(skb->dev), iph->saddr, iph->daddr, key,
302 tunnel_type, &mutable);
307 * Packets received by this function were previously sent by us, so
308 * any comparisons should be to the output values, not the input.
309 * However, it's not really worth it to have a hash table based on
310 * output keys (especially since ICMP error handling of tunneled packets
311 * isn't that reliable anyways). Therefore, we do a lookup based on the
312 * out key as if it were the in key and then check to see if the input
313 * and output keys are the same.
315 if (mutable->key.in_key != mutable->out_key)
318 if (!!(mutable->flags & TNL_F_IN_KEY_MATCH) !=
319 !!(mutable->flags & TNL_F_OUT_KEY_ACTION))
322 if ((mutable->flags & TNL_F_CSUM) && !(flags & GRE_CSUM))
325 tunnel_hdr_len += iph->ihl << 2;
327 orig_mac_header = skb_mac_header(skb) - skb->data;
328 orig_nw_header = skb_network_header(skb) - skb->data;
329 skb_set_mac_header(skb, tunnel_hdr_len);
331 tot_hdr_len = tunnel_hdr_len + ETH_HLEN;
333 skb->protocol = eth_hdr(skb)->h_proto;
334 if (skb->protocol == htons(ETH_P_8021Q)) {
335 tot_hdr_len += VLAN_HLEN;
336 skb->protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
339 skb_set_network_header(skb, tot_hdr_len);
342 if (skb->protocol == htons(ETH_P_IP))
343 tot_hdr_len += sizeof(struct iphdr);
344 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
345 else if (skb->protocol == htons(ETH_P_IPV6))
346 tot_hdr_len += sizeof(struct ipv6hdr);
351 if (!pskb_may_pull(skb, tot_hdr_len))
354 if (skb->protocol == htons(ETH_P_IP)) {
355 if (mtu < IP_MIN_MTU) {
356 if (ntohs(ip_hdr(skb)->tot_len) >= IP_MIN_MTU)
363 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
364 else if (skb->protocol == htons(ETH_P_IPV6)) {
365 if (mtu < IPV6_MIN_MTU) {
366 unsigned int packet_length = sizeof(struct ipv6hdr) +
367 ntohs(ipv6_hdr(skb)->payload_len);
369 if (packet_length >= IPV6_MIN_MTU
370 || ntohs(ipv6_hdr(skb)->payload_len) == 0)
378 __skb_pull(skb, tunnel_hdr_len);
379 ovs_tnl_frag_needed(vport, mutable, skb, mtu);
380 __skb_push(skb, tunnel_hdr_len);
383 skb_set_mac_header(skb, orig_mac_header);
384 skb_set_network_header(skb, orig_nw_header);
385 skb->protocol = htons(ETH_P_IP);
388 static bool check_checksum(struct sk_buff *skb)
390 struct iphdr *iph = ip_hdr(skb);
391 struct gre_base_hdr *greh = (struct gre_base_hdr *)(iph + 1);
394 if (greh->flags & GRE_CSUM) {
395 switch (skb->ip_summed) {
396 case CHECKSUM_COMPLETE:
397 csum = csum_fold(skb->csum);
405 csum = __skb_checksum_complete(skb);
406 skb->ip_summed = CHECKSUM_COMPLETE;
414 static u32 gre_flags_to_tunnel_flags(const struct tnl_mutable_config *mutable,
417 u32 tunnel_flags = 0;
419 if (gre_flags & GRE_KEY) {
420 if (mutable->key.daddr && (mutable->flags & TNL_F_IN_KEY_MATCH))
421 tunnel_flags = OVS_FLOW_TNL_F_KEY;
422 else if (!mutable->key.daddr)
423 tunnel_flags = OVS_FLOW_TNL_F_KEY;
426 if (gre_flags & GRE_CSUM)
427 tunnel_flags |= OVS_FLOW_TNL_F_CSUM;
432 /* Called with rcu_read_lock and BH disabled. */
433 static int gre_rcv(struct sk_buff *skb)
436 const struct tnl_mutable_config *mutable;
439 struct ovs_key_ipv4_tunnel tun_key;
444 if (unlikely(!pskb_may_pull(skb, sizeof(struct gre_base_hdr) + ETH_HLEN)))
446 if (unlikely(!check_checksum(skb)))
449 hdr_len = parse_header(ip_hdr(skb), &flags, &key, &tunnel_type);
450 if (unlikely(hdr_len < 0))
453 if (unlikely(!pskb_may_pull(skb, hdr_len + ETH_HLEN)))
457 vport = ovs_tnl_find_port(dev_net(skb->dev), iph->daddr, iph->saddr, key,
458 tunnel_type, &mutable);
459 if (unlikely(!vport)) {
460 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
464 tnl_tun_key_init(&tun_key, iph, key, gre_flags_to_tunnel_flags(mutable, flags));
465 OVS_CB(skb)->tun_key = &tun_key;
467 __skb_pull(skb, hdr_len);
468 skb_postpull_rcsum(skb, skb_transport_header(skb), hdr_len + ETH_HLEN);
470 ovs_tnl_rcv(vport, skb);
478 static const struct tnl_ops gre_tnl_ops = {
479 .tunnel_type = TNL_T_PROTO_GRE,
480 .ipproto = IPPROTO_GRE,
481 .hdr_len = gre_hdr_len,
482 .build_header = gre_build_header,
483 .update_header = gre_update_header,
486 static struct vport *gre_create(const struct vport_parms *parms)
488 return ovs_tnl_create(parms, &ovs_gre_vport_ops, &gre_tnl_ops);
491 static const struct tnl_ops gre64_tnl_ops = {
492 .tunnel_type = TNL_T_PROTO_GRE64,
493 .ipproto = IPPROTO_GRE,
494 .hdr_len = gre_hdr_len,
495 .build_header = gre_build_header,
496 .update_header = gre_update_header,
499 static struct vport *gre_create64(const struct vport_parms *parms)
501 return ovs_tnl_create(parms, &ovs_gre64_vport_ops, &gre64_tnl_ops);
504 static const struct net_protocol gre_protocol_handlers = {
506 .err_handler = gre_err,
507 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32)
514 static int gre_init(void)
522 err = inet_add_protocol(&gre_protocol_handlers, IPPROTO_GRE);
524 pr_warn("cannot register gre protocol handler\n");
529 static void gre_exit(void)
536 inet_del_protocol(&gre_protocol_handlers, IPPROTO_GRE);
539 const struct vport_ops ovs_gre_vport_ops = {
540 .type = OVS_VPORT_TYPE_GRE,
541 .flags = VPORT_F_TUN_ID,
544 .create = gre_create,
545 .destroy = ovs_tnl_destroy,
546 .set_addr = ovs_tnl_set_addr,
547 .get_name = ovs_tnl_get_name,
548 .get_addr = ovs_tnl_get_addr,
549 .get_options = ovs_tnl_get_options,
550 .set_options = ovs_tnl_set_options,
551 .get_dev_flags = ovs_vport_gen_get_dev_flags,
552 .is_running = ovs_vport_gen_is_running,
553 .get_operstate = ovs_vport_gen_get_operstate,
554 .send = ovs_tnl_send,
557 const struct vport_ops ovs_gre64_vport_ops = {
558 .type = OVS_VPORT_TYPE_GRE64,
559 .flags = VPORT_F_TUN_ID,
562 .create = gre_create64,
563 .destroy = ovs_tnl_destroy,
564 .set_addr = ovs_tnl_set_addr,
565 .get_name = ovs_tnl_get_name,
566 .get_addr = ovs_tnl_get_addr,
567 .get_options = ovs_tnl_get_options,
568 .set_options = ovs_tnl_set_options,
569 .get_dev_flags = ovs_vport_gen_get_dev_flags,
570 .is_running = ovs_vport_gen_is_running,
571 .get_operstate = ovs_vport_gen_get_operstate,
572 .send = ovs_tnl_send,