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
seg6_lwt_headroom(struct seg6_iptunnel_encap * tuninfo)30 static size_t seg6_lwt_headroom(struct seg6_iptunnel_encap *tuninfo)
31 {
32 int head = 0;
33
34 switch (tuninfo->mode) {
35 case SEG6_IPTUN_MODE_INLINE:
36 break;
37 case SEG6_IPTUN_MODE_ENCAP:
38 head = sizeof(struct ipv6hdr);
39 break;
40 case SEG6_IPTUN_MODE_L2ENCAP:
41 return 0;
42 }
43
44 return ((tuninfo->srh->hdrlen + 1) << 3) + head;
45 }
46
47 struct seg6_lwt {
48 struct dst_cache cache;
49 struct seg6_iptunnel_encap tuninfo[];
50 };
51
seg6_lwt_lwtunnel(struct lwtunnel_state * lwt)52 static inline struct seg6_lwt *seg6_lwt_lwtunnel(struct lwtunnel_state *lwt)
53 {
54 return (struct seg6_lwt *)lwt->data;
55 }
56
57 static inline struct seg6_iptunnel_encap *
seg6_encap_lwtunnel(struct lwtunnel_state * lwt)58 seg6_encap_lwtunnel(struct lwtunnel_state *lwt)
59 {
60 return seg6_lwt_lwtunnel(lwt)->tuninfo;
61 }
62
63 static const struct nla_policy seg6_iptunnel_policy[SEG6_IPTUNNEL_MAX + 1] = {
64 [SEG6_IPTUNNEL_SRH] = { .type = NLA_BINARY },
65 };
66
nla_put_srh(struct sk_buff * skb,int attrtype,struct seg6_iptunnel_encap * tuninfo)67 static int nla_put_srh(struct sk_buff *skb, int attrtype,
68 struct seg6_iptunnel_encap *tuninfo)
69 {
70 struct seg6_iptunnel_encap *data;
71 struct nlattr *nla;
72 int len;
73
74 len = SEG6_IPTUN_ENCAP_SIZE(tuninfo);
75
76 nla = nla_reserve(skb, attrtype, len);
77 if (!nla)
78 return -EMSGSIZE;
79
80 data = nla_data(nla);
81 memcpy(data, tuninfo, len);
82
83 return 0;
84 }
85
set_tun_src(struct net * net,struct net_device * dev,struct in6_addr * daddr,struct in6_addr * saddr)86 static void set_tun_src(struct net *net, struct net_device *dev,
87 struct in6_addr *daddr, struct in6_addr *saddr)
88 {
89 struct seg6_pernet_data *sdata = seg6_pernet(net);
90 struct in6_addr *tun_src;
91
92 rcu_read_lock();
93
94 tun_src = rcu_dereference(sdata->tun_src);
95
96 if (!ipv6_addr_any(tun_src)) {
97 memcpy(saddr, tun_src, sizeof(struct in6_addr));
98 } else {
99 ipv6_dev_get_saddr(net, dev, daddr, IPV6_PREFER_SRC_PUBLIC,
100 saddr);
101 }
102
103 rcu_read_unlock();
104 }
105
106 /* Compute flowlabel for outer IPv6 header */
seg6_make_flowlabel(struct net * net,struct sk_buff * skb,struct ipv6hdr * inner_hdr)107 static __be32 seg6_make_flowlabel(struct net *net, struct sk_buff *skb,
108 struct ipv6hdr *inner_hdr)
109 {
110 int do_flowlabel = net->ipv6.sysctl.seg6_flowlabel;
111 __be32 flowlabel = 0;
112 u32 hash;
113
114 if (do_flowlabel > 0) {
115 hash = skb_get_hash(skb);
116 hash = rol32(hash, 16);
117 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
118 } else if (!do_flowlabel && skb->protocol == htons(ETH_P_IPV6)) {
119 flowlabel = ip6_flowlabel(inner_hdr);
120 }
121 return flowlabel;
122 }
123
124 /* 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)125 int seg6_do_srh_encap(struct sk_buff *skb, struct ipv6_sr_hdr *osrh, int proto)
126 {
127 struct dst_entry *dst = skb_dst(skb);
128 struct net *net = dev_net(dst->dev);
129 struct ipv6hdr *hdr, *inner_hdr;
130 struct ipv6_sr_hdr *isrh;
131 int hdrlen, tot_len, err;
132 __be32 flowlabel;
133
134 hdrlen = (osrh->hdrlen + 1) << 3;
135 tot_len = hdrlen + sizeof(*hdr);
136
137 err = skb_cow_head(skb, tot_len + skb->mac_len);
138 if (unlikely(err))
139 return err;
140
141 inner_hdr = ipv6_hdr(skb);
142 flowlabel = seg6_make_flowlabel(net, skb, inner_hdr);
143
144 skb_push(skb, tot_len);
145 skb_reset_network_header(skb);
146 skb_mac_header_rebuild(skb);
147 hdr = ipv6_hdr(skb);
148
149 /* inherit tc, flowlabel and hlim
150 * hlim will be decremented in ip6_forward() afterwards and
151 * decapsulation will overwrite inner hlim with outer hlim
152 */
153
154 if (skb->protocol == htons(ETH_P_IPV6)) {
155 ip6_flow_hdr(hdr, ip6_tclass(ip6_flowinfo(inner_hdr)),
156 flowlabel);
157 hdr->hop_limit = inner_hdr->hop_limit;
158 } else {
159 ip6_flow_hdr(hdr, 0, flowlabel);
160 hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
161
162 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
163
164 /* the control block has been erased, so we have to set the
165 * iif once again.
166 * We read the receiving interface index directly from the
167 * skb->skb_iif as it is done in the IPv4 receiving path (i.e.:
168 * ip_rcv_core(...)).
169 */
170 IP6CB(skb)->iif = skb->skb_iif;
171 }
172
173 hdr->nexthdr = NEXTHDR_ROUTING;
174
175 isrh = (void *)hdr + sizeof(*hdr);
176 memcpy(isrh, osrh, hdrlen);
177
178 isrh->nexthdr = proto;
179
180 hdr->daddr = isrh->segments[isrh->first_segment];
181 set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
182
183 #ifdef CONFIG_IPV6_SEG6_HMAC
184 if (sr_has_hmac(isrh)) {
185 err = seg6_push_hmac(net, &hdr->saddr, isrh);
186 if (unlikely(err))
187 return err;
188 }
189 #endif
190
191 skb_postpush_rcsum(skb, hdr, tot_len);
192
193 return 0;
194 }
195 EXPORT_SYMBOL_GPL(seg6_do_srh_encap);
196
197 /* 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)198 int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
199 {
200 struct ipv6hdr *hdr, *oldhdr;
201 struct ipv6_sr_hdr *isrh;
202 int hdrlen, err;
203
204 hdrlen = (osrh->hdrlen + 1) << 3;
205
206 err = skb_cow_head(skb, hdrlen + skb->mac_len);
207 if (unlikely(err))
208 return err;
209
210 oldhdr = ipv6_hdr(skb);
211
212 skb_pull(skb, sizeof(struct ipv6hdr));
213 skb_postpull_rcsum(skb, skb_network_header(skb),
214 sizeof(struct ipv6hdr));
215
216 skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
217 skb_reset_network_header(skb);
218 skb_mac_header_rebuild(skb);
219
220 hdr = ipv6_hdr(skb);
221
222 memmove(hdr, oldhdr, sizeof(*hdr));
223
224 isrh = (void *)hdr + sizeof(*hdr);
225 memcpy(isrh, osrh, hdrlen);
226
227 isrh->nexthdr = hdr->nexthdr;
228 hdr->nexthdr = NEXTHDR_ROUTING;
229
230 isrh->segments[0] = hdr->daddr;
231 hdr->daddr = isrh->segments[isrh->first_segment];
232
233 #ifdef CONFIG_IPV6_SEG6_HMAC
234 if (sr_has_hmac(isrh)) {
235 struct net *net = dev_net(skb_dst(skb)->dev);
236
237 err = seg6_push_hmac(net, &hdr->saddr, isrh);
238 if (unlikely(err))
239 return err;
240 }
241 #endif
242
243 skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
244
245 return 0;
246 }
247 EXPORT_SYMBOL_GPL(seg6_do_srh_inline);
248
seg6_do_srh(struct sk_buff * skb)249 static int seg6_do_srh(struct sk_buff *skb)
250 {
251 struct dst_entry *dst = skb_dst(skb);
252 struct seg6_iptunnel_encap *tinfo;
253 int proto, err = 0;
254
255 tinfo = seg6_encap_lwtunnel(dst->lwtstate);
256
257 switch (tinfo->mode) {
258 case SEG6_IPTUN_MODE_INLINE:
259 if (skb->protocol != htons(ETH_P_IPV6))
260 return -EINVAL;
261
262 err = seg6_do_srh_inline(skb, tinfo->srh);
263 if (err)
264 return err;
265 break;
266 case SEG6_IPTUN_MODE_ENCAP:
267 err = iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6);
268 if (err)
269 return err;
270
271 if (skb->protocol == htons(ETH_P_IPV6))
272 proto = IPPROTO_IPV6;
273 else if (skb->protocol == htons(ETH_P_IP))
274 proto = IPPROTO_IPIP;
275 else
276 return -EINVAL;
277
278 err = seg6_do_srh_encap(skb, tinfo->srh, proto);
279 if (err)
280 return err;
281
282 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
283 skb_set_inner_protocol(skb, skb->protocol);
284 skb->protocol = htons(ETH_P_IPV6);
285 break;
286 case SEG6_IPTUN_MODE_L2ENCAP:
287 if (!skb_mac_header_was_set(skb))
288 return -EINVAL;
289
290 if (pskb_expand_head(skb, skb->mac_len, 0, GFP_ATOMIC) < 0)
291 return -ENOMEM;
292
293 skb_mac_header_rebuild(skb);
294 skb_push(skb, skb->mac_len);
295
296 err = seg6_do_srh_encap(skb, tinfo->srh, IPPROTO_ETHERNET);
297 if (err)
298 return err;
299
300 skb->protocol = htons(ETH_P_IPV6);
301 break;
302 }
303
304 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
305 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
306
307 return 0;
308 }
309
seg6_input(struct sk_buff * skb)310 static int seg6_input(struct sk_buff *skb)
311 {
312 struct dst_entry *orig_dst = skb_dst(skb);
313 struct dst_entry *dst = NULL;
314 struct seg6_lwt *slwt;
315 int err;
316
317 err = seg6_do_srh(skb);
318 if (unlikely(err)) {
319 kfree_skb(skb);
320 return err;
321 }
322
323 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
324
325 preempt_disable();
326 dst = dst_cache_get(&slwt->cache);
327 preempt_enable();
328
329 skb_dst_drop(skb);
330
331 if (!dst) {
332 ip6_route_input(skb);
333 dst = skb_dst(skb);
334 if (!dst->error) {
335 preempt_disable();
336 dst_cache_set_ip6(&slwt->cache, dst,
337 &ipv6_hdr(skb)->saddr);
338 preempt_enable();
339 }
340 } else {
341 skb_dst_set(skb, dst);
342 }
343
344 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
345 if (unlikely(err))
346 return err;
347
348 return dst_input(skb);
349 }
350
seg6_output(struct net * net,struct sock * sk,struct sk_buff * skb)351 static int seg6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
352 {
353 struct dst_entry *orig_dst = skb_dst(skb);
354 struct dst_entry *dst = NULL;
355 struct seg6_lwt *slwt;
356 int err = -EINVAL;
357
358 err = seg6_do_srh(skb);
359 if (unlikely(err))
360 goto drop;
361
362 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
363
364 preempt_disable();
365 dst = dst_cache_get(&slwt->cache);
366 preempt_enable();
367
368 if (unlikely(!dst)) {
369 struct ipv6hdr *hdr = ipv6_hdr(skb);
370 struct flowi6 fl6;
371
372 memset(&fl6, 0, sizeof(fl6));
373 fl6.daddr = hdr->daddr;
374 fl6.saddr = hdr->saddr;
375 fl6.flowlabel = ip6_flowinfo(hdr);
376 fl6.flowi6_mark = skb->mark;
377 fl6.flowi6_proto = hdr->nexthdr;
378
379 dst = ip6_route_output(net, NULL, &fl6);
380 if (dst->error) {
381 err = dst->error;
382 dst_release(dst);
383 goto drop;
384 }
385
386 preempt_disable();
387 dst_cache_set_ip6(&slwt->cache, dst, &fl6.saddr);
388 preempt_enable();
389 }
390
391 skb_dst_drop(skb);
392 skb_dst_set(skb, dst);
393
394 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
395 if (unlikely(err))
396 goto drop;
397
398 return dst_output(net, sk, skb);
399 drop:
400 kfree_skb(skb);
401 return err;
402 }
403
seg6_build_state(struct net * net,struct nlattr * nla,unsigned int family,const void * cfg,struct lwtunnel_state ** ts,struct netlink_ext_ack * extack)404 static int seg6_build_state(struct net *net, struct nlattr *nla,
405 unsigned int family, const void *cfg,
406 struct lwtunnel_state **ts,
407 struct netlink_ext_ack *extack)
408 {
409 struct nlattr *tb[SEG6_IPTUNNEL_MAX + 1];
410 struct seg6_iptunnel_encap *tuninfo;
411 struct lwtunnel_state *newts;
412 int tuninfo_len, min_size;
413 struct seg6_lwt *slwt;
414 int err;
415
416 if (family != AF_INET && family != AF_INET6)
417 return -EINVAL;
418
419 err = nla_parse_nested_deprecated(tb, SEG6_IPTUNNEL_MAX, nla,
420 seg6_iptunnel_policy, extack);
421
422 if (err < 0)
423 return err;
424
425 if (!tb[SEG6_IPTUNNEL_SRH])
426 return -EINVAL;
427
428 tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
429 tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
430
431 /* tuninfo must contain at least the iptunnel encap structure,
432 * the SRH and one segment
433 */
434 min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
435 sizeof(struct in6_addr);
436 if (tuninfo_len < min_size)
437 return -EINVAL;
438
439 switch (tuninfo->mode) {
440 case SEG6_IPTUN_MODE_INLINE:
441 if (family != AF_INET6)
442 return -EINVAL;
443
444 break;
445 case SEG6_IPTUN_MODE_ENCAP:
446 break;
447 case SEG6_IPTUN_MODE_L2ENCAP:
448 break;
449 default:
450 return -EINVAL;
451 }
452
453 /* verify that SRH is consistent */
454 if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo), false))
455 return -EINVAL;
456
457 newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
458 if (!newts)
459 return -ENOMEM;
460
461 slwt = seg6_lwt_lwtunnel(newts);
462
463 err = dst_cache_init(&slwt->cache, GFP_ATOMIC);
464 if (err) {
465 kfree(newts);
466 return err;
467 }
468
469 memcpy(&slwt->tuninfo, tuninfo, tuninfo_len);
470
471 newts->type = LWTUNNEL_ENCAP_SEG6;
472 newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
473
474 if (tuninfo->mode != SEG6_IPTUN_MODE_L2ENCAP)
475 newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
476
477 newts->headroom = seg6_lwt_headroom(tuninfo);
478
479 *ts = newts;
480
481 return 0;
482 }
483
seg6_destroy_state(struct lwtunnel_state * lwt)484 static void seg6_destroy_state(struct lwtunnel_state *lwt)
485 {
486 dst_cache_destroy(&seg6_lwt_lwtunnel(lwt)->cache);
487 }
488
seg6_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)489 static int seg6_fill_encap_info(struct sk_buff *skb,
490 struct lwtunnel_state *lwtstate)
491 {
492 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
493
494 if (nla_put_srh(skb, SEG6_IPTUNNEL_SRH, tuninfo))
495 return -EMSGSIZE;
496
497 return 0;
498 }
499
seg6_encap_nlsize(struct lwtunnel_state * lwtstate)500 static int seg6_encap_nlsize(struct lwtunnel_state *lwtstate)
501 {
502 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
503
504 return nla_total_size(SEG6_IPTUN_ENCAP_SIZE(tuninfo));
505 }
506
seg6_encap_cmp(struct lwtunnel_state * a,struct lwtunnel_state * b)507 static int seg6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
508 {
509 struct seg6_iptunnel_encap *a_hdr = seg6_encap_lwtunnel(a);
510 struct seg6_iptunnel_encap *b_hdr = seg6_encap_lwtunnel(b);
511 int len = SEG6_IPTUN_ENCAP_SIZE(a_hdr);
512
513 if (len != SEG6_IPTUN_ENCAP_SIZE(b_hdr))
514 return 1;
515
516 return memcmp(a_hdr, b_hdr, len);
517 }
518
519 static const struct lwtunnel_encap_ops seg6_iptun_ops = {
520 .build_state = seg6_build_state,
521 .destroy_state = seg6_destroy_state,
522 .output = seg6_output,
523 .input = seg6_input,
524 .fill_encap = seg6_fill_encap_info,
525 .get_encap_size = seg6_encap_nlsize,
526 .cmp_encap = seg6_encap_cmp,
527 .owner = THIS_MODULE,
528 };
529
seg6_iptunnel_init(void)530 int __init seg6_iptunnel_init(void)
531 {
532 return lwtunnel_encap_add_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
533 }
534
seg6_iptunnel_exit(void)535 void seg6_iptunnel_exit(void)
536 {
537 lwtunnel_encap_del_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
538 }
539