2 * SR-IPv6 implementation
5 * David Lebrun <david.lebrun@uclouvain.be>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #include <linux/types.h>
15 #include <linux/skbuff.h>
16 #include <linux/net.h>
17 #include <linux/module.h>
19 #include <net/ip_tunnels.h>
20 #include <net/lwtunnel.h>
21 #include <net/netevent.h>
22 #include <net/netns/generic.h>
23 #include <net/ip6_fib.h>
24 #include <net/route.h>
26 #include <linux/seg6.h>
27 #include <linux/seg6_iptunnel.h>
28 #include <net/addrconf.h>
29 #include <net/ip6_route.h>
30 #include <net/dst_cache.h>
31 #ifdef CONFIG_IPV6_SEG6_HMAC
32 #include <net/seg6_hmac.h>
36 struct dst_cache cache;
37 struct seg6_iptunnel_encap tuninfo[0];
40 static inline struct seg6_lwt *seg6_lwt_lwtunnel(struct lwtunnel_state *lwt)
42 return (struct seg6_lwt *)lwt->data;
45 static inline struct seg6_iptunnel_encap *
46 seg6_encap_lwtunnel(struct lwtunnel_state *lwt)
48 return seg6_lwt_lwtunnel(lwt)->tuninfo;
51 static const struct nla_policy seg6_iptunnel_policy[SEG6_IPTUNNEL_MAX + 1] = {
52 [SEG6_IPTUNNEL_SRH] = { .type = NLA_BINARY },
55 static int nla_put_srh(struct sk_buff *skb, int attrtype,
56 struct seg6_iptunnel_encap *tuninfo)
58 struct seg6_iptunnel_encap *data;
62 len = SEG6_IPTUN_ENCAP_SIZE(tuninfo);
64 nla = nla_reserve(skb, attrtype, len);
69 memcpy(data, tuninfo, len);
74 static void set_tun_src(struct net *net, struct net_device *dev,
75 struct in6_addr *daddr, struct in6_addr *saddr)
77 struct seg6_pernet_data *sdata = seg6_pernet(net);
78 struct in6_addr *tun_src;
82 tun_src = rcu_dereference(sdata->tun_src);
84 if (!ipv6_addr_any(tun_src)) {
85 memcpy(saddr, tun_src, sizeof(struct in6_addr));
87 ipv6_dev_get_saddr(net, dev, daddr, IPV6_PREFER_SRC_PUBLIC,
94 /* encapsulate an IPv6 packet within an outer IPv6 header with a given SRH */
95 int seg6_do_srh_encap(struct sk_buff *skb, struct ipv6_sr_hdr *osrh, int proto)
97 struct dst_entry *dst = skb_dst(skb);
98 struct net *net = dev_net(dst->dev);
99 struct ipv6hdr *hdr, *inner_hdr;
100 struct ipv6_sr_hdr *isrh;
101 int hdrlen, tot_len, err;
103 hdrlen = (osrh->hdrlen + 1) << 3;
104 tot_len = hdrlen + sizeof(*hdr);
106 err = skb_cow_head(skb, tot_len + skb->mac_len);
110 inner_hdr = ipv6_hdr(skb);
112 skb_push(skb, tot_len);
113 skb_reset_network_header(skb);
114 skb_mac_header_rebuild(skb);
117 /* inherit tc, flowlabel and hlim
118 * hlim will be decremented in ip6_forward() afterwards and
119 * decapsulation will overwrite inner hlim with outer hlim
122 if (skb->protocol == htons(ETH_P_IPV6)) {
123 ip6_flow_hdr(hdr, ip6_tclass(ip6_flowinfo(inner_hdr)),
124 ip6_flowlabel(inner_hdr));
125 hdr->hop_limit = inner_hdr->hop_limit;
127 ip6_flow_hdr(hdr, 0, 0);
128 hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
130 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
132 /* the control block has been erased, so we have to set the
134 * We read the receiving interface index directly from the
135 * skb->skb_iif as it is done in the IPv4 receiving path (i.e.:
138 IP6CB(skb)->iif = skb->skb_iif;
141 hdr->nexthdr = NEXTHDR_ROUTING;
143 isrh = (void *)hdr + sizeof(*hdr);
144 memcpy(isrh, osrh, hdrlen);
146 isrh->nexthdr = proto;
148 hdr->daddr = isrh->segments[isrh->first_segment];
149 set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
151 #ifdef CONFIG_IPV6_SEG6_HMAC
152 if (sr_has_hmac(isrh)) {
153 err = seg6_push_hmac(net, &hdr->saddr, isrh);
159 hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
161 skb_postpush_rcsum(skb, hdr, tot_len);
165 EXPORT_SYMBOL_GPL(seg6_do_srh_encap);
167 /* insert an SRH within an IPv6 packet, just after the IPv6 header */
168 int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
170 struct ipv6hdr *hdr, *oldhdr;
171 struct ipv6_sr_hdr *isrh;
174 hdrlen = (osrh->hdrlen + 1) << 3;
176 err = skb_cow_head(skb, hdrlen + skb->mac_len);
180 oldhdr = ipv6_hdr(skb);
182 skb_pull(skb, sizeof(struct ipv6hdr));
183 skb_postpull_rcsum(skb, skb_network_header(skb),
184 sizeof(struct ipv6hdr));
186 skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
187 skb_reset_network_header(skb);
188 skb_mac_header_rebuild(skb);
192 memmove(hdr, oldhdr, sizeof(*hdr));
194 isrh = (void *)hdr + sizeof(*hdr);
195 memcpy(isrh, osrh, hdrlen);
197 isrh->nexthdr = hdr->nexthdr;
198 hdr->nexthdr = NEXTHDR_ROUTING;
200 isrh->segments[0] = hdr->daddr;
201 hdr->daddr = isrh->segments[isrh->first_segment];
203 #ifdef CONFIG_IPV6_SEG6_HMAC
204 if (sr_has_hmac(isrh)) {
205 struct net *net = dev_net(skb_dst(skb)->dev);
207 err = seg6_push_hmac(net, &hdr->saddr, isrh);
213 hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
215 skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
219 EXPORT_SYMBOL_GPL(seg6_do_srh_inline);
221 static int seg6_do_srh(struct sk_buff *skb)
223 struct dst_entry *dst = skb_dst(skb);
224 struct seg6_iptunnel_encap *tinfo;
227 tinfo = seg6_encap_lwtunnel(dst->lwtstate);
229 switch (tinfo->mode) {
230 case SEG6_IPTUN_MODE_INLINE:
231 if (skb->protocol != htons(ETH_P_IPV6))
234 err = seg6_do_srh_inline(skb, tinfo->srh);
238 case SEG6_IPTUN_MODE_ENCAP:
239 err = iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6);
243 if (skb->protocol == htons(ETH_P_IPV6))
244 proto = IPPROTO_IPV6;
245 else if (skb->protocol == htons(ETH_P_IP))
246 proto = IPPROTO_IPIP;
250 err = seg6_do_srh_encap(skb, tinfo->srh, proto);
254 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
255 skb_set_inner_protocol(skb, skb->protocol);
256 skb->protocol = htons(ETH_P_IPV6);
258 case SEG6_IPTUN_MODE_L2ENCAP:
259 if (!skb_mac_header_was_set(skb))
262 if (pskb_expand_head(skb, skb->mac_len, 0, GFP_ATOMIC) < 0)
265 skb_mac_header_rebuild(skb);
266 skb_push(skb, skb->mac_len);
268 err = seg6_do_srh_encap(skb, tinfo->srh, NEXTHDR_NONE);
272 skb->protocol = htons(ETH_P_IPV6);
276 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
281 static int seg6_input(struct sk_buff *skb)
283 struct dst_entry *orig_dst = skb_dst(skb);
284 struct dst_entry *dst = NULL;
285 struct seg6_lwt *slwt;
288 err = seg6_do_srh(skb);
294 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
297 dst = dst_cache_get(&slwt->cache);
303 ip6_route_input(skb);
307 dst_cache_set_ip6(&slwt->cache, dst,
308 &ipv6_hdr(skb)->saddr);
312 skb_dst_set(skb, dst);
315 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
319 return dst_input(skb);
322 static int seg6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
324 struct dst_entry *orig_dst = skb_dst(skb);
325 struct dst_entry *dst = NULL;
326 struct seg6_lwt *slwt;
329 err = seg6_do_srh(skb);
333 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
336 dst = dst_cache_get(&slwt->cache);
339 if (unlikely(!dst)) {
340 struct ipv6hdr *hdr = ipv6_hdr(skb);
343 memset(&fl6, 0, sizeof(fl6));
344 fl6.daddr = hdr->daddr;
345 fl6.saddr = hdr->saddr;
346 fl6.flowlabel = ip6_flowinfo(hdr);
347 fl6.flowi6_mark = skb->mark;
348 fl6.flowi6_proto = hdr->nexthdr;
350 dst = ip6_route_output(net, NULL, &fl6);
358 dst_cache_set_ip6(&slwt->cache, dst, &fl6.saddr);
363 skb_dst_set(skb, dst);
365 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
369 return dst_output(net, sk, skb);
375 static int seg6_build_state(struct nlattr *nla,
376 unsigned int family, const void *cfg,
377 struct lwtunnel_state **ts,
378 struct netlink_ext_ack *extack)
380 struct nlattr *tb[SEG6_IPTUNNEL_MAX + 1];
381 struct seg6_iptunnel_encap *tuninfo;
382 struct lwtunnel_state *newts;
383 int tuninfo_len, min_size;
384 struct seg6_lwt *slwt;
387 if (family != AF_INET && family != AF_INET6)
390 err = nla_parse_nested(tb, SEG6_IPTUNNEL_MAX, nla,
391 seg6_iptunnel_policy, extack);
396 if (!tb[SEG6_IPTUNNEL_SRH])
399 tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
400 tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
402 /* tuninfo must contain at least the iptunnel encap structure,
403 * the SRH and one segment
405 min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
406 sizeof(struct in6_addr);
407 if (tuninfo_len < min_size)
410 switch (tuninfo->mode) {
411 case SEG6_IPTUN_MODE_INLINE:
412 if (family != AF_INET6)
416 case SEG6_IPTUN_MODE_ENCAP:
418 case SEG6_IPTUN_MODE_L2ENCAP:
424 /* verify that SRH is consistent */
425 if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo)))
428 newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
432 slwt = seg6_lwt_lwtunnel(newts);
434 err = dst_cache_init(&slwt->cache, GFP_ATOMIC);
440 memcpy(&slwt->tuninfo, tuninfo, tuninfo_len);
442 newts->type = LWTUNNEL_ENCAP_SEG6;
443 newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
445 if (tuninfo->mode != SEG6_IPTUN_MODE_L2ENCAP)
446 newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
448 newts->headroom = seg6_lwt_headroom(tuninfo);
455 static void seg6_destroy_state(struct lwtunnel_state *lwt)
457 dst_cache_destroy(&seg6_lwt_lwtunnel(lwt)->cache);
460 static int seg6_fill_encap_info(struct sk_buff *skb,
461 struct lwtunnel_state *lwtstate)
463 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
465 if (nla_put_srh(skb, SEG6_IPTUNNEL_SRH, tuninfo))
471 static int seg6_encap_nlsize(struct lwtunnel_state *lwtstate)
473 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
475 return nla_total_size(SEG6_IPTUN_ENCAP_SIZE(tuninfo));
478 static int seg6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
480 struct seg6_iptunnel_encap *a_hdr = seg6_encap_lwtunnel(a);
481 struct seg6_iptunnel_encap *b_hdr = seg6_encap_lwtunnel(b);
482 int len = SEG6_IPTUN_ENCAP_SIZE(a_hdr);
484 if (len != SEG6_IPTUN_ENCAP_SIZE(b_hdr))
487 return memcmp(a_hdr, b_hdr, len);
490 static const struct lwtunnel_encap_ops seg6_iptun_ops = {
491 .build_state = seg6_build_state,
492 .destroy_state = seg6_destroy_state,
493 .output = seg6_output,
495 .fill_encap = seg6_fill_encap_info,
496 .get_encap_size = seg6_encap_nlsize,
497 .cmp_encap = seg6_encap_cmp,
498 .owner = THIS_MODULE,
501 int __init seg6_iptunnel_init(void)
503 return lwtunnel_encap_add_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
506 void seg6_iptunnel_exit(void)
508 lwtunnel_encap_del_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);