1 /*
2 * SR-IPv6 implementation
3 *
4 * Author:
5 * David Lebrun <david.lebrun@uclouvain.be>
6 *
7 *
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.
12 */
13
14 #include <linux/types.h>
15 #include <linux/skbuff.h>
16 #include <linux/net.h>
17 #include <linux/module.h>
18 #include <net/ip.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>
25 #include <net/seg6.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>
33 #endif
34
35 struct seg6_lwt {
36 struct dst_cache cache;
37 struct seg6_iptunnel_encap tuninfo[0];
38 };
39
seg6_lwt_lwtunnel(struct lwtunnel_state * lwt)40 static inline struct seg6_lwt *seg6_lwt_lwtunnel(struct lwtunnel_state *lwt)
41 {
42 return (struct seg6_lwt *)lwt->data;
43 }
44
45 static inline struct seg6_iptunnel_encap *
seg6_encap_lwtunnel(struct lwtunnel_state * lwt)46 seg6_encap_lwtunnel(struct lwtunnel_state *lwt)
47 {
48 return seg6_lwt_lwtunnel(lwt)->tuninfo;
49 }
50
51 static const struct nla_policy seg6_iptunnel_policy[SEG6_IPTUNNEL_MAX + 1] = {
52 [SEG6_IPTUNNEL_SRH] = { .type = NLA_BINARY },
53 };
54
nla_put_srh(struct sk_buff * skb,int attrtype,struct seg6_iptunnel_encap * tuninfo)55 static int nla_put_srh(struct sk_buff *skb, int attrtype,
56 struct seg6_iptunnel_encap *tuninfo)
57 {
58 struct seg6_iptunnel_encap *data;
59 struct nlattr *nla;
60 int len;
61
62 len = SEG6_IPTUN_ENCAP_SIZE(tuninfo);
63
64 nla = nla_reserve(skb, attrtype, len);
65 if (!nla)
66 return -EMSGSIZE;
67
68 data = nla_data(nla);
69 memcpy(data, tuninfo, len);
70
71 return 0;
72 }
73
set_tun_src(struct net * net,struct net_device * dev,struct in6_addr * daddr,struct in6_addr * saddr)74 static void set_tun_src(struct net *net, struct net_device *dev,
75 struct in6_addr *daddr, struct in6_addr *saddr)
76 {
77 struct seg6_pernet_data *sdata = seg6_pernet(net);
78 struct in6_addr *tun_src;
79
80 rcu_read_lock();
81
82 tun_src = rcu_dereference(sdata->tun_src);
83
84 if (!ipv6_addr_any(tun_src)) {
85 memcpy(saddr, tun_src, sizeof(struct in6_addr));
86 } else {
87 ipv6_dev_get_saddr(net, dev, daddr, IPV6_PREFER_SRC_PUBLIC,
88 saddr);
89 }
90
91 rcu_read_unlock();
92 }
93
94 /* encapsulate an IPv6 packet within an outer IPv6 header with a given SRH */
seg6_do_srh_encap(struct sk_buff * skb,struct ipv6_sr_hdr * osrh,int proto)95 int seg6_do_srh_encap(struct sk_buff *skb, struct ipv6_sr_hdr *osrh, int proto)
96 {
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;
102
103 hdrlen = (osrh->hdrlen + 1) << 3;
104 tot_len = hdrlen + sizeof(*hdr);
105
106 err = skb_cow_head(skb, tot_len + skb->mac_len);
107 if (unlikely(err))
108 return err;
109
110 inner_hdr = ipv6_hdr(skb);
111
112 skb_push(skb, tot_len);
113 skb_reset_network_header(skb);
114 skb_mac_header_rebuild(skb);
115 hdr = ipv6_hdr(skb);
116
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
120 */
121
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;
126 } else {
127 ip6_flow_hdr(hdr, 0, 0);
128 hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
129
130 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
131 }
132
133 hdr->nexthdr = NEXTHDR_ROUTING;
134
135 isrh = (void *)hdr + sizeof(*hdr);
136 memcpy(isrh, osrh, hdrlen);
137
138 isrh->nexthdr = proto;
139
140 hdr->daddr = isrh->segments[isrh->first_segment];
141 set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
142
143 #ifdef CONFIG_IPV6_SEG6_HMAC
144 if (sr_has_hmac(isrh)) {
145 err = seg6_push_hmac(net, &hdr->saddr, isrh);
146 if (unlikely(err))
147 return err;
148 }
149 #endif
150
151 skb_postpush_rcsum(skb, hdr, tot_len);
152
153 return 0;
154 }
155 EXPORT_SYMBOL_GPL(seg6_do_srh_encap);
156
157 /* insert an SRH within an IPv6 packet, just after the IPv6 header */
seg6_do_srh_inline(struct sk_buff * skb,struct ipv6_sr_hdr * osrh)158 int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
159 {
160 struct ipv6hdr *hdr, *oldhdr;
161 struct ipv6_sr_hdr *isrh;
162 int hdrlen, err;
163
164 hdrlen = (osrh->hdrlen + 1) << 3;
165
166 err = skb_cow_head(skb, hdrlen + skb->mac_len);
167 if (unlikely(err))
168 return err;
169
170 oldhdr = ipv6_hdr(skb);
171
172 skb_pull(skb, sizeof(struct ipv6hdr));
173 skb_postpull_rcsum(skb, skb_network_header(skb),
174 sizeof(struct ipv6hdr));
175
176 skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
177 skb_reset_network_header(skb);
178 skb_mac_header_rebuild(skb);
179
180 hdr = ipv6_hdr(skb);
181
182 memmove(hdr, oldhdr, sizeof(*hdr));
183
184 isrh = (void *)hdr + sizeof(*hdr);
185 memcpy(isrh, osrh, hdrlen);
186
187 isrh->nexthdr = hdr->nexthdr;
188 hdr->nexthdr = NEXTHDR_ROUTING;
189
190 isrh->segments[0] = hdr->daddr;
191 hdr->daddr = isrh->segments[isrh->first_segment];
192
193 #ifdef CONFIG_IPV6_SEG6_HMAC
194 if (sr_has_hmac(isrh)) {
195 struct net *net = dev_net(skb_dst(skb)->dev);
196
197 err = seg6_push_hmac(net, &hdr->saddr, isrh);
198 if (unlikely(err))
199 return err;
200 }
201 #endif
202
203 skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
204
205 return 0;
206 }
207 EXPORT_SYMBOL_GPL(seg6_do_srh_inline);
208
seg6_do_srh(struct sk_buff * skb)209 static int seg6_do_srh(struct sk_buff *skb)
210 {
211 struct dst_entry *dst = skb_dst(skb);
212 struct seg6_iptunnel_encap *tinfo;
213 int proto, err = 0;
214
215 tinfo = seg6_encap_lwtunnel(dst->lwtstate);
216
217 switch (tinfo->mode) {
218 case SEG6_IPTUN_MODE_INLINE:
219 if (skb->protocol != htons(ETH_P_IPV6))
220 return -EINVAL;
221
222 err = seg6_do_srh_inline(skb, tinfo->srh);
223 if (err)
224 return err;
225 break;
226 case SEG6_IPTUN_MODE_ENCAP:
227 err = iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6);
228 if (err)
229 return err;
230
231 if (skb->protocol == htons(ETH_P_IPV6))
232 proto = IPPROTO_IPV6;
233 else if (skb->protocol == htons(ETH_P_IP))
234 proto = IPPROTO_IPIP;
235 else
236 return -EINVAL;
237
238 err = seg6_do_srh_encap(skb, tinfo->srh, proto);
239 if (err)
240 return err;
241
242 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
243 skb_set_inner_protocol(skb, skb->protocol);
244 skb->protocol = htons(ETH_P_IPV6);
245 break;
246 case SEG6_IPTUN_MODE_L2ENCAP:
247 if (!skb_mac_header_was_set(skb))
248 return -EINVAL;
249
250 if (pskb_expand_head(skb, skb->mac_len, 0, GFP_ATOMIC) < 0)
251 return -ENOMEM;
252
253 skb_mac_header_rebuild(skb);
254 skb_push(skb, skb->mac_len);
255
256 err = seg6_do_srh_encap(skb, tinfo->srh, NEXTHDR_NONE);
257 if (err)
258 return err;
259
260 skb->protocol = htons(ETH_P_IPV6);
261 break;
262 }
263
264 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
265 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
266
267 return 0;
268 }
269
seg6_input(struct sk_buff * skb)270 static int seg6_input(struct sk_buff *skb)
271 {
272 struct dst_entry *orig_dst = skb_dst(skb);
273 struct dst_entry *dst = NULL;
274 struct seg6_lwt *slwt;
275 int err;
276
277 err = seg6_do_srh(skb);
278 if (unlikely(err)) {
279 kfree_skb(skb);
280 return err;
281 }
282
283 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
284
285 preempt_disable();
286 dst = dst_cache_get(&slwt->cache);
287 preempt_enable();
288
289 skb_dst_drop(skb);
290
291 if (!dst) {
292 ip6_route_input(skb);
293 dst = skb_dst(skb);
294 if (!dst->error) {
295 preempt_disable();
296 dst_cache_set_ip6(&slwt->cache, dst,
297 &ipv6_hdr(skb)->saddr);
298 preempt_enable();
299 }
300 } else {
301 skb_dst_set(skb, dst);
302 }
303
304 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
305 if (unlikely(err))
306 return err;
307
308 return dst_input(skb);
309 }
310
seg6_output(struct net * net,struct sock * sk,struct sk_buff * skb)311 static int seg6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
312 {
313 struct dst_entry *orig_dst = skb_dst(skb);
314 struct dst_entry *dst = NULL;
315 struct seg6_lwt *slwt;
316 int err = -EINVAL;
317
318 err = seg6_do_srh(skb);
319 if (unlikely(err))
320 goto drop;
321
322 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
323
324 preempt_disable();
325 dst = dst_cache_get(&slwt->cache);
326 preempt_enable();
327
328 if (unlikely(!dst)) {
329 struct ipv6hdr *hdr = ipv6_hdr(skb);
330 struct flowi6 fl6;
331
332 memset(&fl6, 0, sizeof(fl6));
333 fl6.daddr = hdr->daddr;
334 fl6.saddr = hdr->saddr;
335 fl6.flowlabel = ip6_flowinfo(hdr);
336 fl6.flowi6_mark = skb->mark;
337 fl6.flowi6_proto = hdr->nexthdr;
338
339 dst = ip6_route_output(net, NULL, &fl6);
340 if (dst->error) {
341 err = dst->error;
342 dst_release(dst);
343 goto drop;
344 }
345
346 preempt_disable();
347 dst_cache_set_ip6(&slwt->cache, dst, &fl6.saddr);
348 preempt_enable();
349 }
350
351 skb_dst_drop(skb);
352 skb_dst_set(skb, dst);
353
354 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
355 if (unlikely(err))
356 goto drop;
357
358 return dst_output(net, sk, skb);
359 drop:
360 kfree_skb(skb);
361 return err;
362 }
363
seg6_build_state(struct nlattr * nla,unsigned int family,const void * cfg,struct lwtunnel_state ** ts,struct netlink_ext_ack * extack)364 static int seg6_build_state(struct nlattr *nla,
365 unsigned int family, const void *cfg,
366 struct lwtunnel_state **ts,
367 struct netlink_ext_ack *extack)
368 {
369 struct nlattr *tb[SEG6_IPTUNNEL_MAX + 1];
370 struct seg6_iptunnel_encap *tuninfo;
371 struct lwtunnel_state *newts;
372 int tuninfo_len, min_size;
373 struct seg6_lwt *slwt;
374 int err;
375
376 if (family != AF_INET && family != AF_INET6)
377 return -EINVAL;
378
379 err = nla_parse_nested(tb, SEG6_IPTUNNEL_MAX, nla,
380 seg6_iptunnel_policy, extack);
381
382 if (err < 0)
383 return err;
384
385 if (!tb[SEG6_IPTUNNEL_SRH])
386 return -EINVAL;
387
388 tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
389 tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
390
391 /* tuninfo must contain at least the iptunnel encap structure,
392 * the SRH and one segment
393 */
394 min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
395 sizeof(struct in6_addr);
396 if (tuninfo_len < min_size)
397 return -EINVAL;
398
399 switch (tuninfo->mode) {
400 case SEG6_IPTUN_MODE_INLINE:
401 if (family != AF_INET6)
402 return -EINVAL;
403
404 break;
405 case SEG6_IPTUN_MODE_ENCAP:
406 break;
407 case SEG6_IPTUN_MODE_L2ENCAP:
408 break;
409 default:
410 return -EINVAL;
411 }
412
413 /* verify that SRH is consistent */
414 if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo)))
415 return -EINVAL;
416
417 newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
418 if (!newts)
419 return -ENOMEM;
420
421 slwt = seg6_lwt_lwtunnel(newts);
422
423 err = dst_cache_init(&slwt->cache, GFP_ATOMIC);
424 if (err) {
425 kfree(newts);
426 return err;
427 }
428
429 memcpy(&slwt->tuninfo, tuninfo, tuninfo_len);
430
431 newts->type = LWTUNNEL_ENCAP_SEG6;
432 newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
433
434 if (tuninfo->mode != SEG6_IPTUN_MODE_L2ENCAP)
435 newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
436
437 newts->headroom = seg6_lwt_headroom(tuninfo);
438
439 *ts = newts;
440
441 return 0;
442 }
443
seg6_destroy_state(struct lwtunnel_state * lwt)444 static void seg6_destroy_state(struct lwtunnel_state *lwt)
445 {
446 dst_cache_destroy(&seg6_lwt_lwtunnel(lwt)->cache);
447 }
448
seg6_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)449 static int seg6_fill_encap_info(struct sk_buff *skb,
450 struct lwtunnel_state *lwtstate)
451 {
452 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
453
454 if (nla_put_srh(skb, SEG6_IPTUNNEL_SRH, tuninfo))
455 return -EMSGSIZE;
456
457 return 0;
458 }
459
seg6_encap_nlsize(struct lwtunnel_state * lwtstate)460 static int seg6_encap_nlsize(struct lwtunnel_state *lwtstate)
461 {
462 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
463
464 return nla_total_size(SEG6_IPTUN_ENCAP_SIZE(tuninfo));
465 }
466
seg6_encap_cmp(struct lwtunnel_state * a,struct lwtunnel_state * b)467 static int seg6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
468 {
469 struct seg6_iptunnel_encap *a_hdr = seg6_encap_lwtunnel(a);
470 struct seg6_iptunnel_encap *b_hdr = seg6_encap_lwtunnel(b);
471 int len = SEG6_IPTUN_ENCAP_SIZE(a_hdr);
472
473 if (len != SEG6_IPTUN_ENCAP_SIZE(b_hdr))
474 return 1;
475
476 return memcmp(a_hdr, b_hdr, len);
477 }
478
479 static const struct lwtunnel_encap_ops seg6_iptun_ops = {
480 .build_state = seg6_build_state,
481 .destroy_state = seg6_destroy_state,
482 .output = seg6_output,
483 .input = seg6_input,
484 .fill_encap = seg6_fill_encap_info,
485 .get_encap_size = seg6_encap_nlsize,
486 .cmp_encap = seg6_encap_cmp,
487 .owner = THIS_MODULE,
488 };
489
seg6_iptunnel_init(void)490 int __init seg6_iptunnel_init(void)
491 {
492 return lwtunnel_encap_add_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
493 }
494
seg6_iptunnel_exit(void)495 void seg6_iptunnel_exit(void)
496 {
497 lwtunnel_encap_del_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
498 }
499