1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * SR-IPv6 implementation
4 *
5 * Author:
6 * David Lebrun <david.lebrun@uclouvain.be>
7 */
8
9 #include <linux/types.h>
10 #include <linux/skbuff.h>
11 #include <linux/net.h>
12 #include <linux/module.h>
13 #include <net/ip.h>
14 #include <net/ip_tunnels.h>
15 #include <net/lwtunnel.h>
16 #include <net/netevent.h>
17 #include <net/netns/generic.h>
18 #include <net/ip6_fib.h>
19 #include <net/route.h>
20 #include <net/seg6.h>
21 #include <linux/seg6.h>
22 #include <linux/seg6_iptunnel.h>
23 #include <net/addrconf.h>
24 #include <net/ip6_route.h>
25 #include <net/dst_cache.h>
26 #ifdef CONFIG_IPV6_SEG6_HMAC
27 #include <net/seg6_hmac.h>
28 #endif
29
30 struct seg6_lwt {
31 struct dst_cache cache;
32 struct seg6_iptunnel_encap tuninfo[0];
33 };
34
seg6_lwt_lwtunnel(struct lwtunnel_state * lwt)35 static inline struct seg6_lwt *seg6_lwt_lwtunnel(struct lwtunnel_state *lwt)
36 {
37 return (struct seg6_lwt *)lwt->data;
38 }
39
40 static inline struct seg6_iptunnel_encap *
seg6_encap_lwtunnel(struct lwtunnel_state * lwt)41 seg6_encap_lwtunnel(struct lwtunnel_state *lwt)
42 {
43 return seg6_lwt_lwtunnel(lwt)->tuninfo;
44 }
45
46 static const struct nla_policy seg6_iptunnel_policy[SEG6_IPTUNNEL_MAX + 1] = {
47 [SEG6_IPTUNNEL_SRH] = { .type = NLA_BINARY },
48 };
49
nla_put_srh(struct sk_buff * skb,int attrtype,struct seg6_iptunnel_encap * tuninfo)50 static int nla_put_srh(struct sk_buff *skb, int attrtype,
51 struct seg6_iptunnel_encap *tuninfo)
52 {
53 struct seg6_iptunnel_encap *data;
54 struct nlattr *nla;
55 int len;
56
57 len = SEG6_IPTUN_ENCAP_SIZE(tuninfo);
58
59 nla = nla_reserve(skb, attrtype, len);
60 if (!nla)
61 return -EMSGSIZE;
62
63 data = nla_data(nla);
64 memcpy(data, tuninfo, len);
65
66 return 0;
67 }
68
set_tun_src(struct net * net,struct net_device * dev,struct in6_addr * daddr,struct in6_addr * saddr)69 static void set_tun_src(struct net *net, struct net_device *dev,
70 struct in6_addr *daddr, struct in6_addr *saddr)
71 {
72 struct seg6_pernet_data *sdata = seg6_pernet(net);
73 struct in6_addr *tun_src;
74
75 rcu_read_lock();
76
77 tun_src = rcu_dereference(sdata->tun_src);
78
79 if (!ipv6_addr_any(tun_src)) {
80 memcpy(saddr, tun_src, sizeof(struct in6_addr));
81 } else {
82 ipv6_dev_get_saddr(net, dev, daddr, IPV6_PREFER_SRC_PUBLIC,
83 saddr);
84 }
85
86 rcu_read_unlock();
87 }
88
89 /* Compute flowlabel for outer IPv6 header */
seg6_make_flowlabel(struct net * net,struct sk_buff * skb,struct ipv6hdr * inner_hdr)90 static __be32 seg6_make_flowlabel(struct net *net, struct sk_buff *skb,
91 struct ipv6hdr *inner_hdr)
92 {
93 int do_flowlabel = net->ipv6.sysctl.seg6_flowlabel;
94 __be32 flowlabel = 0;
95 u32 hash;
96
97 if (do_flowlabel > 0) {
98 hash = skb_get_hash(skb);
99 hash = rol32(hash, 16);
100 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
101 } else if (!do_flowlabel && skb->protocol == htons(ETH_P_IPV6)) {
102 flowlabel = ip6_flowlabel(inner_hdr);
103 }
104 return flowlabel;
105 }
106
107 /* 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)108 int seg6_do_srh_encap(struct sk_buff *skb, struct ipv6_sr_hdr *osrh, int proto)
109 {
110 struct dst_entry *dst = skb_dst(skb);
111 struct net *net = dev_net(dst->dev);
112 struct ipv6hdr *hdr, *inner_hdr;
113 struct ipv6_sr_hdr *isrh;
114 int hdrlen, tot_len, err;
115 __be32 flowlabel;
116
117 hdrlen = (osrh->hdrlen + 1) << 3;
118 tot_len = hdrlen + sizeof(*hdr);
119
120 err = skb_cow_head(skb, tot_len + skb->mac_len);
121 if (unlikely(err))
122 return err;
123
124 inner_hdr = ipv6_hdr(skb);
125 flowlabel = seg6_make_flowlabel(net, skb, inner_hdr);
126
127 skb_push(skb, tot_len);
128 skb_reset_network_header(skb);
129 skb_mac_header_rebuild(skb);
130 hdr = ipv6_hdr(skb);
131
132 /* inherit tc, flowlabel and hlim
133 * hlim will be decremented in ip6_forward() afterwards and
134 * decapsulation will overwrite inner hlim with outer hlim
135 */
136
137 if (skb->protocol == htons(ETH_P_IPV6)) {
138 ip6_flow_hdr(hdr, ip6_tclass(ip6_flowinfo(inner_hdr)),
139 flowlabel);
140 hdr->hop_limit = inner_hdr->hop_limit;
141 } else {
142 ip6_flow_hdr(hdr, 0, flowlabel);
143 hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
144
145 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
146
147 /* the control block has been erased, so we have to set the
148 * iif once again.
149 * We read the receiving interface index directly from the
150 * skb->skb_iif as it is done in the IPv4 receiving path (i.e.:
151 * ip_rcv_core(...)).
152 */
153 IP6CB(skb)->iif = skb->skb_iif;
154 }
155
156 hdr->nexthdr = NEXTHDR_ROUTING;
157
158 isrh = (void *)hdr + sizeof(*hdr);
159 memcpy(isrh, osrh, hdrlen);
160
161 isrh->nexthdr = proto;
162
163 hdr->daddr = isrh->segments[isrh->first_segment];
164 set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
165
166 #ifdef CONFIG_IPV6_SEG6_HMAC
167 if (sr_has_hmac(isrh)) {
168 err = seg6_push_hmac(net, &hdr->saddr, isrh);
169 if (unlikely(err))
170 return err;
171 }
172 #endif
173
174 hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
175
176 skb_postpush_rcsum(skb, hdr, tot_len);
177
178 return 0;
179 }
180 EXPORT_SYMBOL_GPL(seg6_do_srh_encap);
181
182 /* 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)183 int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
184 {
185 struct ipv6hdr *hdr, *oldhdr;
186 struct ipv6_sr_hdr *isrh;
187 int hdrlen, err;
188
189 hdrlen = (osrh->hdrlen + 1) << 3;
190
191 err = skb_cow_head(skb, hdrlen + skb->mac_len);
192 if (unlikely(err))
193 return err;
194
195 oldhdr = ipv6_hdr(skb);
196
197 skb_pull(skb, sizeof(struct ipv6hdr));
198 skb_postpull_rcsum(skb, skb_network_header(skb),
199 sizeof(struct ipv6hdr));
200
201 skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
202 skb_reset_network_header(skb);
203 skb_mac_header_rebuild(skb);
204
205 hdr = ipv6_hdr(skb);
206
207 memmove(hdr, oldhdr, sizeof(*hdr));
208
209 isrh = (void *)hdr + sizeof(*hdr);
210 memcpy(isrh, osrh, hdrlen);
211
212 isrh->nexthdr = hdr->nexthdr;
213 hdr->nexthdr = NEXTHDR_ROUTING;
214
215 isrh->segments[0] = hdr->daddr;
216 hdr->daddr = isrh->segments[isrh->first_segment];
217
218 #ifdef CONFIG_IPV6_SEG6_HMAC
219 if (sr_has_hmac(isrh)) {
220 struct net *net = dev_net(skb_dst(skb)->dev);
221
222 err = seg6_push_hmac(net, &hdr->saddr, isrh);
223 if (unlikely(err))
224 return err;
225 }
226 #endif
227
228 hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
229
230 skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
231
232 return 0;
233 }
234 EXPORT_SYMBOL_GPL(seg6_do_srh_inline);
235
seg6_do_srh(struct sk_buff * skb)236 static int seg6_do_srh(struct sk_buff *skb)
237 {
238 struct dst_entry *dst = skb_dst(skb);
239 struct seg6_iptunnel_encap *tinfo;
240 int proto, err = 0;
241
242 tinfo = seg6_encap_lwtunnel(dst->lwtstate);
243
244 switch (tinfo->mode) {
245 case SEG6_IPTUN_MODE_INLINE:
246 if (skb->protocol != htons(ETH_P_IPV6))
247 return -EINVAL;
248
249 err = seg6_do_srh_inline(skb, tinfo->srh);
250 if (err)
251 return err;
252 break;
253 case SEG6_IPTUN_MODE_ENCAP:
254 err = iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6);
255 if (err)
256 return err;
257
258 if (skb->protocol == htons(ETH_P_IPV6))
259 proto = IPPROTO_IPV6;
260 else if (skb->protocol == htons(ETH_P_IP))
261 proto = IPPROTO_IPIP;
262 else
263 return -EINVAL;
264
265 err = seg6_do_srh_encap(skb, tinfo->srh, proto);
266 if (err)
267 return err;
268
269 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
270 skb_set_inner_protocol(skb, skb->protocol);
271 skb->protocol = htons(ETH_P_IPV6);
272 break;
273 case SEG6_IPTUN_MODE_L2ENCAP:
274 if (!skb_mac_header_was_set(skb))
275 return -EINVAL;
276
277 if (pskb_expand_head(skb, skb->mac_len, 0, GFP_ATOMIC) < 0)
278 return -ENOMEM;
279
280 skb_mac_header_rebuild(skb);
281 skb_push(skb, skb->mac_len);
282
283 err = seg6_do_srh_encap(skb, tinfo->srh, NEXTHDR_NONE);
284 if (err)
285 return err;
286
287 skb->protocol = htons(ETH_P_IPV6);
288 break;
289 }
290
291 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
292
293 return 0;
294 }
295
seg6_input(struct sk_buff * skb)296 static int seg6_input(struct sk_buff *skb)
297 {
298 struct dst_entry *orig_dst = skb_dst(skb);
299 struct dst_entry *dst = NULL;
300 struct seg6_lwt *slwt;
301 int err;
302
303 err = seg6_do_srh(skb);
304 if (unlikely(err)) {
305 kfree_skb(skb);
306 return err;
307 }
308
309 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
310
311 preempt_disable();
312 dst = dst_cache_get(&slwt->cache);
313 preempt_enable();
314
315 skb_dst_drop(skb);
316
317 if (!dst) {
318 ip6_route_input(skb);
319 dst = skb_dst(skb);
320 if (!dst->error) {
321 preempt_disable();
322 dst_cache_set_ip6(&slwt->cache, dst,
323 &ipv6_hdr(skb)->saddr);
324 preempt_enable();
325 }
326 } else {
327 skb_dst_set(skb, dst);
328 }
329
330 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
331 if (unlikely(err))
332 return err;
333
334 return dst_input(skb);
335 }
336
seg6_output(struct net * net,struct sock * sk,struct sk_buff * skb)337 static int seg6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
338 {
339 struct dst_entry *orig_dst = skb_dst(skb);
340 struct dst_entry *dst = NULL;
341 struct seg6_lwt *slwt;
342 int err = -EINVAL;
343
344 err = seg6_do_srh(skb);
345 if (unlikely(err))
346 goto drop;
347
348 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
349
350 preempt_disable();
351 dst = dst_cache_get(&slwt->cache);
352 preempt_enable();
353
354 if (unlikely(!dst)) {
355 struct ipv6hdr *hdr = ipv6_hdr(skb);
356 struct flowi6 fl6;
357
358 memset(&fl6, 0, sizeof(fl6));
359 fl6.daddr = hdr->daddr;
360 fl6.saddr = hdr->saddr;
361 fl6.flowlabel = ip6_flowinfo(hdr);
362 fl6.flowi6_mark = skb->mark;
363 fl6.flowi6_proto = hdr->nexthdr;
364
365 dst = ip6_route_output(net, NULL, &fl6);
366 if (dst->error) {
367 err = dst->error;
368 dst_release(dst);
369 goto drop;
370 }
371
372 preempt_disable();
373 dst_cache_set_ip6(&slwt->cache, dst, &fl6.saddr);
374 preempt_enable();
375 }
376
377 skb_dst_drop(skb);
378 skb_dst_set(skb, dst);
379
380 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
381 if (unlikely(err))
382 goto drop;
383
384 return dst_output(net, sk, skb);
385 drop:
386 kfree_skb(skb);
387 return err;
388 }
389
seg6_build_state(struct nlattr * nla,unsigned int family,const void * cfg,struct lwtunnel_state ** ts,struct netlink_ext_ack * extack)390 static int seg6_build_state(struct nlattr *nla,
391 unsigned int family, const void *cfg,
392 struct lwtunnel_state **ts,
393 struct netlink_ext_ack *extack)
394 {
395 struct nlattr *tb[SEG6_IPTUNNEL_MAX + 1];
396 struct seg6_iptunnel_encap *tuninfo;
397 struct lwtunnel_state *newts;
398 int tuninfo_len, min_size;
399 struct seg6_lwt *slwt;
400 int err;
401
402 if (family != AF_INET && family != AF_INET6)
403 return -EINVAL;
404
405 err = nla_parse_nested_deprecated(tb, SEG6_IPTUNNEL_MAX, nla,
406 seg6_iptunnel_policy, extack);
407
408 if (err < 0)
409 return err;
410
411 if (!tb[SEG6_IPTUNNEL_SRH])
412 return -EINVAL;
413
414 tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
415 tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
416
417 /* tuninfo must contain at least the iptunnel encap structure,
418 * the SRH and one segment
419 */
420 min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
421 sizeof(struct in6_addr);
422 if (tuninfo_len < min_size)
423 return -EINVAL;
424
425 switch (tuninfo->mode) {
426 case SEG6_IPTUN_MODE_INLINE:
427 if (family != AF_INET6)
428 return -EINVAL;
429
430 break;
431 case SEG6_IPTUN_MODE_ENCAP:
432 break;
433 case SEG6_IPTUN_MODE_L2ENCAP:
434 break;
435 default:
436 return -EINVAL;
437 }
438
439 /* verify that SRH is consistent */
440 if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo)))
441 return -EINVAL;
442
443 newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
444 if (!newts)
445 return -ENOMEM;
446
447 slwt = seg6_lwt_lwtunnel(newts);
448
449 err = dst_cache_init(&slwt->cache, GFP_ATOMIC);
450 if (err) {
451 kfree(newts);
452 return err;
453 }
454
455 memcpy(&slwt->tuninfo, tuninfo, tuninfo_len);
456
457 newts->type = LWTUNNEL_ENCAP_SEG6;
458 newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
459
460 if (tuninfo->mode != SEG6_IPTUN_MODE_L2ENCAP)
461 newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
462
463 newts->headroom = seg6_lwt_headroom(tuninfo);
464
465 *ts = newts;
466
467 return 0;
468 }
469
seg6_destroy_state(struct lwtunnel_state * lwt)470 static void seg6_destroy_state(struct lwtunnel_state *lwt)
471 {
472 dst_cache_destroy(&seg6_lwt_lwtunnel(lwt)->cache);
473 }
474
seg6_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)475 static int seg6_fill_encap_info(struct sk_buff *skb,
476 struct lwtunnel_state *lwtstate)
477 {
478 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
479
480 if (nla_put_srh(skb, SEG6_IPTUNNEL_SRH, tuninfo))
481 return -EMSGSIZE;
482
483 return 0;
484 }
485
seg6_encap_nlsize(struct lwtunnel_state * lwtstate)486 static int seg6_encap_nlsize(struct lwtunnel_state *lwtstate)
487 {
488 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
489
490 return nla_total_size(SEG6_IPTUN_ENCAP_SIZE(tuninfo));
491 }
492
seg6_encap_cmp(struct lwtunnel_state * a,struct lwtunnel_state * b)493 static int seg6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
494 {
495 struct seg6_iptunnel_encap *a_hdr = seg6_encap_lwtunnel(a);
496 struct seg6_iptunnel_encap *b_hdr = seg6_encap_lwtunnel(b);
497 int len = SEG6_IPTUN_ENCAP_SIZE(a_hdr);
498
499 if (len != SEG6_IPTUN_ENCAP_SIZE(b_hdr))
500 return 1;
501
502 return memcmp(a_hdr, b_hdr, len);
503 }
504
505 static const struct lwtunnel_encap_ops seg6_iptun_ops = {
506 .build_state = seg6_build_state,
507 .destroy_state = seg6_destroy_state,
508 .output = seg6_output,
509 .input = seg6_input,
510 .fill_encap = seg6_fill_encap_info,
511 .get_encap_size = seg6_encap_nlsize,
512 .cmp_encap = seg6_encap_cmp,
513 .owner = THIS_MODULE,
514 };
515
seg6_iptunnel_init(void)516 int __init seg6_iptunnel_init(void)
517 {
518 return lwtunnel_encap_add_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
519 }
520
seg6_iptunnel_exit(void)521 void seg6_iptunnel_exit(void)
522 {
523 lwtunnel_encap_del_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
524 }
525