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
148 hdr->nexthdr = NEXTHDR_ROUTING;
149
150 isrh = (void *)hdr + sizeof(*hdr);
151 memcpy(isrh, osrh, hdrlen);
152
153 isrh->nexthdr = proto;
154
155 hdr->daddr = isrh->segments[isrh->first_segment];
156 set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
157
158 #ifdef CONFIG_IPV6_SEG6_HMAC
159 if (sr_has_hmac(isrh)) {
160 err = seg6_push_hmac(net, &hdr->saddr, isrh);
161 if (unlikely(err))
162 return err;
163 }
164 #endif
165
166 skb_postpush_rcsum(skb, hdr, tot_len);
167
168 return 0;
169 }
170 EXPORT_SYMBOL_GPL(seg6_do_srh_encap);
171
172 /* 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)173 int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
174 {
175 struct ipv6hdr *hdr, *oldhdr;
176 struct ipv6_sr_hdr *isrh;
177 int hdrlen, err;
178
179 hdrlen = (osrh->hdrlen + 1) << 3;
180
181 err = skb_cow_head(skb, hdrlen + skb->mac_len);
182 if (unlikely(err))
183 return err;
184
185 oldhdr = ipv6_hdr(skb);
186
187 skb_pull(skb, sizeof(struct ipv6hdr));
188 skb_postpull_rcsum(skb, skb_network_header(skb),
189 sizeof(struct ipv6hdr));
190
191 skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
192 skb_reset_network_header(skb);
193 skb_mac_header_rebuild(skb);
194
195 hdr = ipv6_hdr(skb);
196
197 memmove(hdr, oldhdr, sizeof(*hdr));
198
199 isrh = (void *)hdr + sizeof(*hdr);
200 memcpy(isrh, osrh, hdrlen);
201
202 isrh->nexthdr = hdr->nexthdr;
203 hdr->nexthdr = NEXTHDR_ROUTING;
204
205 isrh->segments[0] = hdr->daddr;
206 hdr->daddr = isrh->segments[isrh->first_segment];
207
208 #ifdef CONFIG_IPV6_SEG6_HMAC
209 if (sr_has_hmac(isrh)) {
210 struct net *net = dev_net(skb_dst(skb)->dev);
211
212 err = seg6_push_hmac(net, &hdr->saddr, isrh);
213 if (unlikely(err))
214 return err;
215 }
216 #endif
217
218 skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
219
220 return 0;
221 }
222 EXPORT_SYMBOL_GPL(seg6_do_srh_inline);
223
seg6_do_srh(struct sk_buff * skb)224 static int seg6_do_srh(struct sk_buff *skb)
225 {
226 struct dst_entry *dst = skb_dst(skb);
227 struct seg6_iptunnel_encap *tinfo;
228 int proto, err = 0;
229
230 tinfo = seg6_encap_lwtunnel(dst->lwtstate);
231
232 switch (tinfo->mode) {
233 case SEG6_IPTUN_MODE_INLINE:
234 if (skb->protocol != htons(ETH_P_IPV6))
235 return -EINVAL;
236
237 err = seg6_do_srh_inline(skb, tinfo->srh);
238 if (err)
239 return err;
240 break;
241 case SEG6_IPTUN_MODE_ENCAP:
242 err = iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6);
243 if (err)
244 return err;
245
246 if (skb->protocol == htons(ETH_P_IPV6))
247 proto = IPPROTO_IPV6;
248 else if (skb->protocol == htons(ETH_P_IP))
249 proto = IPPROTO_IPIP;
250 else
251 return -EINVAL;
252
253 err = seg6_do_srh_encap(skb, tinfo->srh, proto);
254 if (err)
255 return err;
256
257 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
258 skb_set_inner_protocol(skb, skb->protocol);
259 skb->protocol = htons(ETH_P_IPV6);
260 break;
261 case SEG6_IPTUN_MODE_L2ENCAP:
262 if (!skb_mac_header_was_set(skb))
263 return -EINVAL;
264
265 if (pskb_expand_head(skb, skb->mac_len, 0, GFP_ATOMIC) < 0)
266 return -ENOMEM;
267
268 skb_mac_header_rebuild(skb);
269 skb_push(skb, skb->mac_len);
270
271 err = seg6_do_srh_encap(skb, tinfo->srh, NEXTHDR_NONE);
272 if (err)
273 return err;
274
275 skb->protocol = htons(ETH_P_IPV6);
276 break;
277 }
278
279 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
280 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
281
282 return 0;
283 }
284
seg6_input(struct sk_buff * skb)285 static int seg6_input(struct sk_buff *skb)
286 {
287 struct dst_entry *orig_dst = skb_dst(skb);
288 struct dst_entry *dst = NULL;
289 struct seg6_lwt *slwt;
290 int err;
291
292 err = seg6_do_srh(skb);
293 if (unlikely(err)) {
294 kfree_skb(skb);
295 return err;
296 }
297
298 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
299
300 preempt_disable();
301 dst = dst_cache_get(&slwt->cache);
302 preempt_enable();
303
304 skb_dst_drop(skb);
305
306 if (!dst) {
307 ip6_route_input(skb);
308 dst = skb_dst(skb);
309 if (!dst->error) {
310 preempt_disable();
311 dst_cache_set_ip6(&slwt->cache, dst,
312 &ipv6_hdr(skb)->saddr);
313 preempt_enable();
314 }
315 } else {
316 skb_dst_set(skb, dst);
317 }
318
319 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
320 if (unlikely(err))
321 return err;
322
323 return dst_input(skb);
324 }
325
seg6_output(struct net * net,struct sock * sk,struct sk_buff * skb)326 static int seg6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
327 {
328 struct dst_entry *orig_dst = skb_dst(skb);
329 struct dst_entry *dst = NULL;
330 struct seg6_lwt *slwt;
331 int err = -EINVAL;
332
333 err = seg6_do_srh(skb);
334 if (unlikely(err))
335 goto drop;
336
337 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
338
339 preempt_disable();
340 dst = dst_cache_get(&slwt->cache);
341 preempt_enable();
342
343 if (unlikely(!dst)) {
344 struct ipv6hdr *hdr = ipv6_hdr(skb);
345 struct flowi6 fl6;
346
347 memset(&fl6, 0, sizeof(fl6));
348 fl6.daddr = hdr->daddr;
349 fl6.saddr = hdr->saddr;
350 fl6.flowlabel = ip6_flowinfo(hdr);
351 fl6.flowi6_mark = skb->mark;
352 fl6.flowi6_proto = hdr->nexthdr;
353
354 dst = ip6_route_output(net, NULL, &fl6);
355 if (dst->error) {
356 err = dst->error;
357 dst_release(dst);
358 goto drop;
359 }
360
361 preempt_disable();
362 dst_cache_set_ip6(&slwt->cache, dst, &fl6.saddr);
363 preempt_enable();
364 }
365
366 skb_dst_drop(skb);
367 skb_dst_set(skb, dst);
368
369 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
370 if (unlikely(err))
371 goto drop;
372
373 return dst_output(net, sk, skb);
374 drop:
375 kfree_skb(skb);
376 return err;
377 }
378
seg6_build_state(struct nlattr * nla,unsigned int family,const void * cfg,struct lwtunnel_state ** ts,struct netlink_ext_ack * extack)379 static int seg6_build_state(struct nlattr *nla,
380 unsigned int family, const void *cfg,
381 struct lwtunnel_state **ts,
382 struct netlink_ext_ack *extack)
383 {
384 struct nlattr *tb[SEG6_IPTUNNEL_MAX + 1];
385 struct seg6_iptunnel_encap *tuninfo;
386 struct lwtunnel_state *newts;
387 int tuninfo_len, min_size;
388 struct seg6_lwt *slwt;
389 int err;
390
391 if (family != AF_INET && family != AF_INET6)
392 return -EINVAL;
393
394 err = nla_parse_nested_deprecated(tb, SEG6_IPTUNNEL_MAX, nla,
395 seg6_iptunnel_policy, extack);
396
397 if (err < 0)
398 return err;
399
400 if (!tb[SEG6_IPTUNNEL_SRH])
401 return -EINVAL;
402
403 tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
404 tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
405
406 /* tuninfo must contain at least the iptunnel encap structure,
407 * the SRH and one segment
408 */
409 min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
410 sizeof(struct in6_addr);
411 if (tuninfo_len < min_size)
412 return -EINVAL;
413
414 switch (tuninfo->mode) {
415 case SEG6_IPTUN_MODE_INLINE:
416 if (family != AF_INET6)
417 return -EINVAL;
418
419 break;
420 case SEG6_IPTUN_MODE_ENCAP:
421 break;
422 case SEG6_IPTUN_MODE_L2ENCAP:
423 break;
424 default:
425 return -EINVAL;
426 }
427
428 /* verify that SRH is consistent */
429 if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo)))
430 return -EINVAL;
431
432 newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
433 if (!newts)
434 return -ENOMEM;
435
436 slwt = seg6_lwt_lwtunnel(newts);
437
438 err = dst_cache_init(&slwt->cache, GFP_ATOMIC);
439 if (err) {
440 kfree(newts);
441 return err;
442 }
443
444 memcpy(&slwt->tuninfo, tuninfo, tuninfo_len);
445
446 newts->type = LWTUNNEL_ENCAP_SEG6;
447 newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
448
449 if (tuninfo->mode != SEG6_IPTUN_MODE_L2ENCAP)
450 newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
451
452 newts->headroom = seg6_lwt_headroom(tuninfo);
453
454 *ts = newts;
455
456 return 0;
457 }
458
seg6_destroy_state(struct lwtunnel_state * lwt)459 static void seg6_destroy_state(struct lwtunnel_state *lwt)
460 {
461 dst_cache_destroy(&seg6_lwt_lwtunnel(lwt)->cache);
462 }
463
seg6_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)464 static int seg6_fill_encap_info(struct sk_buff *skb,
465 struct lwtunnel_state *lwtstate)
466 {
467 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
468
469 if (nla_put_srh(skb, SEG6_IPTUNNEL_SRH, tuninfo))
470 return -EMSGSIZE;
471
472 return 0;
473 }
474
seg6_encap_nlsize(struct lwtunnel_state * lwtstate)475 static int seg6_encap_nlsize(struct lwtunnel_state *lwtstate)
476 {
477 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
478
479 return nla_total_size(SEG6_IPTUN_ENCAP_SIZE(tuninfo));
480 }
481
seg6_encap_cmp(struct lwtunnel_state * a,struct lwtunnel_state * b)482 static int seg6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
483 {
484 struct seg6_iptunnel_encap *a_hdr = seg6_encap_lwtunnel(a);
485 struct seg6_iptunnel_encap *b_hdr = seg6_encap_lwtunnel(b);
486 int len = SEG6_IPTUN_ENCAP_SIZE(a_hdr);
487
488 if (len != SEG6_IPTUN_ENCAP_SIZE(b_hdr))
489 return 1;
490
491 return memcmp(a_hdr, b_hdr, len);
492 }
493
494 static const struct lwtunnel_encap_ops seg6_iptun_ops = {
495 .build_state = seg6_build_state,
496 .destroy_state = seg6_destroy_state,
497 .output = seg6_output,
498 .input = seg6_input,
499 .fill_encap = seg6_fill_encap_info,
500 .get_encap_size = seg6_encap_nlsize,
501 .cmp_encap = seg6_encap_cmp,
502 .owner = THIS_MODULE,
503 };
504
seg6_iptunnel_init(void)505 int __init seg6_iptunnel_init(void)
506 {
507 return lwtunnel_encap_add_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
508 }
509
seg6_iptunnel_exit(void)510 void seg6_iptunnel_exit(void)
511 {
512 lwtunnel_encap_del_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
513 }
514