Loading...
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
35struct seg6_lwt {
36 struct dst_cache cache;
37 struct seg6_iptunnel_encap tuninfo[0];
38};
39
40static inline struct seg6_lwt *seg6_lwt_lwtunnel(struct lwtunnel_state *lwt)
41{
42 return (struct seg6_lwt *)lwt->data;
43}
44
45static inline struct seg6_iptunnel_encap *
46seg6_encap_lwtunnel(struct lwtunnel_state *lwt)
47{
48 return seg6_lwt_lwtunnel(lwt)->tuninfo;
49}
50
51static const struct nla_policy seg6_iptunnel_policy[SEG6_IPTUNNEL_MAX + 1] = {
52 [SEG6_IPTUNNEL_SRH] = { .type = NLA_BINARY },
53};
54
55static 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
74static 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 */
95int 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
131 hdr->nexthdr = NEXTHDR_ROUTING;
132
133 isrh = (void *)hdr + sizeof(*hdr);
134 memcpy(isrh, osrh, hdrlen);
135
136 isrh->nexthdr = proto;
137
138 hdr->daddr = isrh->segments[isrh->first_segment];
139 set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
140
141#ifdef CONFIG_IPV6_SEG6_HMAC
142 if (sr_has_hmac(isrh)) {
143 err = seg6_push_hmac(net, &hdr->saddr, isrh);
144 if (unlikely(err))
145 return err;
146 }
147#endif
148
149 skb_postpush_rcsum(skb, hdr, tot_len);
150
151 return 0;
152}
153EXPORT_SYMBOL_GPL(seg6_do_srh_encap);
154
155/* insert an SRH within an IPv6 packet, just after the IPv6 header */
156int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
157{
158 struct ipv6hdr *hdr, *oldhdr;
159 struct ipv6_sr_hdr *isrh;
160 int hdrlen, err;
161
162 hdrlen = (osrh->hdrlen + 1) << 3;
163
164 err = skb_cow_head(skb, hdrlen + skb->mac_len);
165 if (unlikely(err))
166 return err;
167
168 oldhdr = ipv6_hdr(skb);
169
170 skb_pull(skb, sizeof(struct ipv6hdr));
171 skb_postpull_rcsum(skb, skb_network_header(skb),
172 sizeof(struct ipv6hdr));
173
174 skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
175 skb_reset_network_header(skb);
176 skb_mac_header_rebuild(skb);
177
178 hdr = ipv6_hdr(skb);
179
180 memmove(hdr, oldhdr, sizeof(*hdr));
181
182 isrh = (void *)hdr + sizeof(*hdr);
183 memcpy(isrh, osrh, hdrlen);
184
185 isrh->nexthdr = hdr->nexthdr;
186 hdr->nexthdr = NEXTHDR_ROUTING;
187
188 isrh->segments[0] = hdr->daddr;
189 hdr->daddr = isrh->segments[isrh->first_segment];
190
191#ifdef CONFIG_IPV6_SEG6_HMAC
192 if (sr_has_hmac(isrh)) {
193 struct net *net = dev_net(skb_dst(skb)->dev);
194
195 err = seg6_push_hmac(net, &hdr->saddr, isrh);
196 if (unlikely(err))
197 return err;
198 }
199#endif
200
201 skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
202
203 return 0;
204}
205EXPORT_SYMBOL_GPL(seg6_do_srh_inline);
206
207static int seg6_do_srh(struct sk_buff *skb)
208{
209 struct dst_entry *dst = skb_dst(skb);
210 struct seg6_iptunnel_encap *tinfo;
211 int proto, err = 0;
212
213 tinfo = seg6_encap_lwtunnel(dst->lwtstate);
214
215 switch (tinfo->mode) {
216 case SEG6_IPTUN_MODE_INLINE:
217 if (skb->protocol != htons(ETH_P_IPV6))
218 return -EINVAL;
219
220 err = seg6_do_srh_inline(skb, tinfo->srh);
221 if (err)
222 return err;
223 break;
224 case SEG6_IPTUN_MODE_ENCAP:
225 err = iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6);
226 if (err)
227 return err;
228
229 if (skb->protocol == htons(ETH_P_IPV6))
230 proto = IPPROTO_IPV6;
231 else if (skb->protocol == htons(ETH_P_IP))
232 proto = IPPROTO_IPIP;
233 else
234 return -EINVAL;
235
236 err = seg6_do_srh_encap(skb, tinfo->srh, proto);
237 if (err)
238 return err;
239
240 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
241 skb_set_inner_protocol(skb, skb->protocol);
242 skb->protocol = htons(ETH_P_IPV6);
243 break;
244 case SEG6_IPTUN_MODE_L2ENCAP:
245 if (!skb_mac_header_was_set(skb))
246 return -EINVAL;
247
248 if (pskb_expand_head(skb, skb->mac_len, 0, GFP_ATOMIC) < 0)
249 return -ENOMEM;
250
251 skb_mac_header_rebuild(skb);
252 skb_push(skb, skb->mac_len);
253
254 err = seg6_do_srh_encap(skb, tinfo->srh, NEXTHDR_NONE);
255 if (err)
256 return err;
257
258 skb->protocol = htons(ETH_P_IPV6);
259 break;
260 }
261
262 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
263 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
264
265 return 0;
266}
267
268static int seg6_input(struct sk_buff *skb)
269{
270 struct dst_entry *orig_dst = skb_dst(skb);
271 struct dst_entry *dst = NULL;
272 struct seg6_lwt *slwt;
273 int err;
274
275 err = seg6_do_srh(skb);
276 if (unlikely(err)) {
277 kfree_skb(skb);
278 return err;
279 }
280
281 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
282
283 preempt_disable();
284 dst = dst_cache_get(&slwt->cache);
285 preempt_enable();
286
287 skb_dst_drop(skb);
288
289 if (!dst) {
290 ip6_route_input(skb);
291 dst = skb_dst(skb);
292 if (!dst->error) {
293 preempt_disable();
294 dst_cache_set_ip6(&slwt->cache, dst,
295 &ipv6_hdr(skb)->saddr);
296 preempt_enable();
297 }
298 } else {
299 skb_dst_set(skb, dst);
300 }
301
302 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
303 if (unlikely(err))
304 return err;
305
306 return dst_input(skb);
307}
308
309static int seg6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
310{
311 struct dst_entry *orig_dst = skb_dst(skb);
312 struct dst_entry *dst = NULL;
313 struct seg6_lwt *slwt;
314 int err = -EINVAL;
315
316 err = seg6_do_srh(skb);
317 if (unlikely(err))
318 goto drop;
319
320 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
321
322 preempt_disable();
323 dst = dst_cache_get(&slwt->cache);
324 preempt_enable();
325
326 if (unlikely(!dst)) {
327 struct ipv6hdr *hdr = ipv6_hdr(skb);
328 struct flowi6 fl6;
329
330 fl6.daddr = hdr->daddr;
331 fl6.saddr = hdr->saddr;
332 fl6.flowlabel = ip6_flowinfo(hdr);
333 fl6.flowi6_mark = skb->mark;
334 fl6.flowi6_proto = hdr->nexthdr;
335
336 dst = ip6_route_output(net, NULL, &fl6);
337 if (dst->error) {
338 err = dst->error;
339 dst_release(dst);
340 goto drop;
341 }
342
343 preempt_disable();
344 dst_cache_set_ip6(&slwt->cache, dst, &fl6.saddr);
345 preempt_enable();
346 }
347
348 skb_dst_drop(skb);
349 skb_dst_set(skb, dst);
350
351 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
352 if (unlikely(err))
353 goto drop;
354
355 return dst_output(net, sk, skb);
356drop:
357 kfree_skb(skb);
358 return err;
359}
360
361static int seg6_build_state(struct nlattr *nla,
362 unsigned int family, const void *cfg,
363 struct lwtunnel_state **ts,
364 struct netlink_ext_ack *extack)
365{
366 struct nlattr *tb[SEG6_IPTUNNEL_MAX + 1];
367 struct seg6_iptunnel_encap *tuninfo;
368 struct lwtunnel_state *newts;
369 int tuninfo_len, min_size;
370 struct seg6_lwt *slwt;
371 int err;
372
373 if (family != AF_INET && family != AF_INET6)
374 return -EINVAL;
375
376 err = nla_parse_nested(tb, SEG6_IPTUNNEL_MAX, nla,
377 seg6_iptunnel_policy, extack);
378
379 if (err < 0)
380 return err;
381
382 if (!tb[SEG6_IPTUNNEL_SRH])
383 return -EINVAL;
384
385 tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
386 tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
387
388 /* tuninfo must contain at least the iptunnel encap structure,
389 * the SRH and one segment
390 */
391 min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
392 sizeof(struct in6_addr);
393 if (tuninfo_len < min_size)
394 return -EINVAL;
395
396 switch (tuninfo->mode) {
397 case SEG6_IPTUN_MODE_INLINE:
398 if (family != AF_INET6)
399 return -EINVAL;
400
401 break;
402 case SEG6_IPTUN_MODE_ENCAP:
403 break;
404 case SEG6_IPTUN_MODE_L2ENCAP:
405 break;
406 default:
407 return -EINVAL;
408 }
409
410 /* verify that SRH is consistent */
411 if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo)))
412 return -EINVAL;
413
414 newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
415 if (!newts)
416 return -ENOMEM;
417
418 slwt = seg6_lwt_lwtunnel(newts);
419
420 err = dst_cache_init(&slwt->cache, GFP_ATOMIC);
421 if (err) {
422 kfree(newts);
423 return err;
424 }
425
426 memcpy(&slwt->tuninfo, tuninfo, tuninfo_len);
427
428 newts->type = LWTUNNEL_ENCAP_SEG6;
429 newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
430
431 if (tuninfo->mode != SEG6_IPTUN_MODE_L2ENCAP)
432 newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
433
434 newts->headroom = seg6_lwt_headroom(tuninfo);
435
436 *ts = newts;
437
438 return 0;
439}
440
441static void seg6_destroy_state(struct lwtunnel_state *lwt)
442{
443 dst_cache_destroy(&seg6_lwt_lwtunnel(lwt)->cache);
444}
445
446static int seg6_fill_encap_info(struct sk_buff *skb,
447 struct lwtunnel_state *lwtstate)
448{
449 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
450
451 if (nla_put_srh(skb, SEG6_IPTUNNEL_SRH, tuninfo))
452 return -EMSGSIZE;
453
454 return 0;
455}
456
457static int seg6_encap_nlsize(struct lwtunnel_state *lwtstate)
458{
459 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
460
461 return nla_total_size(SEG6_IPTUN_ENCAP_SIZE(tuninfo));
462}
463
464static int seg6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
465{
466 struct seg6_iptunnel_encap *a_hdr = seg6_encap_lwtunnel(a);
467 struct seg6_iptunnel_encap *b_hdr = seg6_encap_lwtunnel(b);
468 int len = SEG6_IPTUN_ENCAP_SIZE(a_hdr);
469
470 if (len != SEG6_IPTUN_ENCAP_SIZE(b_hdr))
471 return 1;
472
473 return memcmp(a_hdr, b_hdr, len);
474}
475
476static const struct lwtunnel_encap_ops seg6_iptun_ops = {
477 .build_state = seg6_build_state,
478 .destroy_state = seg6_destroy_state,
479 .output = seg6_output,
480 .input = seg6_input,
481 .fill_encap = seg6_fill_encap_info,
482 .get_encap_size = seg6_encap_nlsize,
483 .cmp_encap = seg6_encap_cmp,
484 .owner = THIS_MODULE,
485};
486
487int __init seg6_iptunnel_init(void)
488{
489 return lwtunnel_encap_add_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
490}
491
492void seg6_iptunnel_exit(void)
493{
494 lwtunnel_encap_del_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
495}
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#include <linux/netfilter.h>
30
31static size_t seg6_lwt_headroom(struct seg6_iptunnel_encap *tuninfo)
32{
33 int head = 0;
34
35 switch (tuninfo->mode) {
36 case SEG6_IPTUN_MODE_INLINE:
37 break;
38 case SEG6_IPTUN_MODE_ENCAP:
39 case SEG6_IPTUN_MODE_ENCAP_RED:
40 head = sizeof(struct ipv6hdr);
41 break;
42 case SEG6_IPTUN_MODE_L2ENCAP:
43 case SEG6_IPTUN_MODE_L2ENCAP_RED:
44 return 0;
45 }
46
47 return ((tuninfo->srh->hdrlen + 1) << 3) + head;
48}
49
50struct seg6_lwt {
51 struct dst_cache cache;
52 struct seg6_iptunnel_encap tuninfo[];
53};
54
55static inline struct seg6_lwt *seg6_lwt_lwtunnel(struct lwtunnel_state *lwt)
56{
57 return (struct seg6_lwt *)lwt->data;
58}
59
60static inline struct seg6_iptunnel_encap *
61seg6_encap_lwtunnel(struct lwtunnel_state *lwt)
62{
63 return seg6_lwt_lwtunnel(lwt)->tuninfo;
64}
65
66static const struct nla_policy seg6_iptunnel_policy[SEG6_IPTUNNEL_MAX + 1] = {
67 [SEG6_IPTUNNEL_SRH] = { .type = NLA_BINARY },
68};
69
70static int nla_put_srh(struct sk_buff *skb, int attrtype,
71 struct seg6_iptunnel_encap *tuninfo)
72{
73 struct seg6_iptunnel_encap *data;
74 struct nlattr *nla;
75 int len;
76
77 len = SEG6_IPTUN_ENCAP_SIZE(tuninfo);
78
79 nla = nla_reserve(skb, attrtype, len);
80 if (!nla)
81 return -EMSGSIZE;
82
83 data = nla_data(nla);
84 memcpy(data, tuninfo, len);
85
86 return 0;
87}
88
89static void set_tun_src(struct net *net, struct net_device *dev,
90 struct in6_addr *daddr, struct in6_addr *saddr)
91{
92 struct seg6_pernet_data *sdata = seg6_pernet(net);
93 struct in6_addr *tun_src;
94
95 rcu_read_lock();
96
97 tun_src = rcu_dereference(sdata->tun_src);
98
99 if (!ipv6_addr_any(tun_src)) {
100 memcpy(saddr, tun_src, sizeof(struct in6_addr));
101 } else {
102 ipv6_dev_get_saddr(net, dev, daddr, IPV6_PREFER_SRC_PUBLIC,
103 saddr);
104 }
105
106 rcu_read_unlock();
107}
108
109/* Compute flowlabel for outer IPv6 header */
110static __be32 seg6_make_flowlabel(struct net *net, struct sk_buff *skb,
111 struct ipv6hdr *inner_hdr)
112{
113 int do_flowlabel = net->ipv6.sysctl.seg6_flowlabel;
114 __be32 flowlabel = 0;
115 u32 hash;
116
117 if (do_flowlabel > 0) {
118 hash = skb_get_hash(skb);
119 hash = rol32(hash, 16);
120 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
121 } else if (!do_flowlabel && skb->protocol == htons(ETH_P_IPV6)) {
122 flowlabel = ip6_flowlabel(inner_hdr);
123 }
124 return flowlabel;
125}
126
127/* encapsulate an IPv6 packet within an outer IPv6 header with a given SRH */
128int seg6_do_srh_encap(struct sk_buff *skb, struct ipv6_sr_hdr *osrh, int proto)
129{
130 struct dst_entry *dst = skb_dst(skb);
131 struct net *net = dev_net(dst->dev);
132 struct ipv6hdr *hdr, *inner_hdr;
133 struct ipv6_sr_hdr *isrh;
134 int hdrlen, tot_len, err;
135 __be32 flowlabel;
136
137 hdrlen = (osrh->hdrlen + 1) << 3;
138 tot_len = hdrlen + sizeof(*hdr);
139
140 err = skb_cow_head(skb, tot_len + skb->mac_len);
141 if (unlikely(err))
142 return err;
143
144 inner_hdr = ipv6_hdr(skb);
145 flowlabel = seg6_make_flowlabel(net, skb, inner_hdr);
146
147 skb_push(skb, tot_len);
148 skb_reset_network_header(skb);
149 skb_mac_header_rebuild(skb);
150 hdr = ipv6_hdr(skb);
151
152 /* inherit tc, flowlabel and hlim
153 * hlim will be decremented in ip6_forward() afterwards and
154 * decapsulation will overwrite inner hlim with outer hlim
155 */
156
157 if (skb->protocol == htons(ETH_P_IPV6)) {
158 ip6_flow_hdr(hdr, ip6_tclass(ip6_flowinfo(inner_hdr)),
159 flowlabel);
160 hdr->hop_limit = inner_hdr->hop_limit;
161 } else {
162 ip6_flow_hdr(hdr, 0, flowlabel);
163 hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
164
165 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
166
167 /* the control block has been erased, so we have to set the
168 * iif once again.
169 * We read the receiving interface index directly from the
170 * skb->skb_iif as it is done in the IPv4 receiving path (i.e.:
171 * ip_rcv_core(...)).
172 */
173 IP6CB(skb)->iif = skb->skb_iif;
174 }
175
176 hdr->nexthdr = NEXTHDR_ROUTING;
177
178 isrh = (void *)hdr + sizeof(*hdr);
179 memcpy(isrh, osrh, hdrlen);
180
181 isrh->nexthdr = proto;
182
183 hdr->daddr = isrh->segments[isrh->first_segment];
184 set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
185
186#ifdef CONFIG_IPV6_SEG6_HMAC
187 if (sr_has_hmac(isrh)) {
188 err = seg6_push_hmac(net, &hdr->saddr, isrh);
189 if (unlikely(err))
190 return err;
191 }
192#endif
193
194 hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
195
196 skb_postpush_rcsum(skb, hdr, tot_len);
197
198 return 0;
199}
200EXPORT_SYMBOL_GPL(seg6_do_srh_encap);
201
202/* encapsulate an IPv6 packet within an outer IPv6 header with reduced SRH */
203static int seg6_do_srh_encap_red(struct sk_buff *skb,
204 struct ipv6_sr_hdr *osrh, int proto)
205{
206 __u8 first_seg = osrh->first_segment;
207 struct dst_entry *dst = skb_dst(skb);
208 struct net *net = dev_net(dst->dev);
209 struct ipv6hdr *hdr, *inner_hdr;
210 int hdrlen = ipv6_optlen(osrh);
211 int red_tlv_offset, tlv_offset;
212 struct ipv6_sr_hdr *isrh;
213 bool skip_srh = false;
214 __be32 flowlabel;
215 int tot_len, err;
216 int red_hdrlen;
217 int tlvs_len;
218
219 if (first_seg > 0) {
220 red_hdrlen = hdrlen - sizeof(struct in6_addr);
221 } else {
222 /* NOTE: if tag/flags and/or other TLVs are introduced in the
223 * seg6_iptunnel infrastructure, they should be considered when
224 * deciding to skip the SRH.
225 */
226 skip_srh = !sr_has_hmac(osrh);
227
228 red_hdrlen = skip_srh ? 0 : hdrlen;
229 }
230
231 tot_len = red_hdrlen + sizeof(struct ipv6hdr);
232
233 err = skb_cow_head(skb, tot_len + skb->mac_len);
234 if (unlikely(err))
235 return err;
236
237 inner_hdr = ipv6_hdr(skb);
238 flowlabel = seg6_make_flowlabel(net, skb, inner_hdr);
239
240 skb_push(skb, tot_len);
241 skb_reset_network_header(skb);
242 skb_mac_header_rebuild(skb);
243 hdr = ipv6_hdr(skb);
244
245 /* based on seg6_do_srh_encap() */
246 if (skb->protocol == htons(ETH_P_IPV6)) {
247 ip6_flow_hdr(hdr, ip6_tclass(ip6_flowinfo(inner_hdr)),
248 flowlabel);
249 hdr->hop_limit = inner_hdr->hop_limit;
250 } else {
251 ip6_flow_hdr(hdr, 0, flowlabel);
252 hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
253
254 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
255 IP6CB(skb)->iif = skb->skb_iif;
256 }
257
258 /* no matter if we have to skip the SRH or not, the first segment
259 * always comes in the pushed IPv6 header.
260 */
261 hdr->daddr = osrh->segments[first_seg];
262
263 if (skip_srh) {
264 hdr->nexthdr = proto;
265
266 set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
267 goto out;
268 }
269
270 /* we cannot skip the SRH, slow path */
271
272 hdr->nexthdr = NEXTHDR_ROUTING;
273 isrh = (void *)hdr + sizeof(struct ipv6hdr);
274
275 if (unlikely(!first_seg)) {
276 /* this is a very rare case; we have only one SID but
277 * we cannot skip the SRH since we are carrying some
278 * other info.
279 */
280 memcpy(isrh, osrh, hdrlen);
281 goto srcaddr;
282 }
283
284 tlv_offset = sizeof(*osrh) + (first_seg + 1) * sizeof(struct in6_addr);
285 red_tlv_offset = tlv_offset - sizeof(struct in6_addr);
286
287 memcpy(isrh, osrh, red_tlv_offset);
288
289 tlvs_len = hdrlen - tlv_offset;
290 if (unlikely(tlvs_len > 0)) {
291 const void *s = (const void *)osrh + tlv_offset;
292 void *d = (void *)isrh + red_tlv_offset;
293
294 memcpy(d, s, tlvs_len);
295 }
296
297 --isrh->first_segment;
298 isrh->hdrlen -= 2;
299
300srcaddr:
301 isrh->nexthdr = proto;
302 set_tun_src(net, dst->dev, &hdr->daddr, &hdr->saddr);
303
304#ifdef CONFIG_IPV6_SEG6_HMAC
305 if (unlikely(!skip_srh && sr_has_hmac(isrh))) {
306 err = seg6_push_hmac(net, &hdr->saddr, isrh);
307 if (unlikely(err))
308 return err;
309 }
310#endif
311
312out:
313 hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
314
315 skb_postpush_rcsum(skb, hdr, tot_len);
316
317 return 0;
318}
319
320/* insert an SRH within an IPv6 packet, just after the IPv6 header */
321int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
322{
323 struct ipv6hdr *hdr, *oldhdr;
324 struct ipv6_sr_hdr *isrh;
325 int hdrlen, err;
326
327 hdrlen = (osrh->hdrlen + 1) << 3;
328
329 err = skb_cow_head(skb, hdrlen + skb->mac_len);
330 if (unlikely(err))
331 return err;
332
333 oldhdr = ipv6_hdr(skb);
334
335 skb_pull(skb, sizeof(struct ipv6hdr));
336 skb_postpull_rcsum(skb, skb_network_header(skb),
337 sizeof(struct ipv6hdr));
338
339 skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
340 skb_reset_network_header(skb);
341 skb_mac_header_rebuild(skb);
342
343 hdr = ipv6_hdr(skb);
344
345 memmove(hdr, oldhdr, sizeof(*hdr));
346
347 isrh = (void *)hdr + sizeof(*hdr);
348 memcpy(isrh, osrh, hdrlen);
349
350 isrh->nexthdr = hdr->nexthdr;
351 hdr->nexthdr = NEXTHDR_ROUTING;
352
353 isrh->segments[0] = hdr->daddr;
354 hdr->daddr = isrh->segments[isrh->first_segment];
355
356#ifdef CONFIG_IPV6_SEG6_HMAC
357 if (sr_has_hmac(isrh)) {
358 struct net *net = dev_net(skb_dst(skb)->dev);
359
360 err = seg6_push_hmac(net, &hdr->saddr, isrh);
361 if (unlikely(err))
362 return err;
363 }
364#endif
365
366 hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
367
368 skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
369
370 return 0;
371}
372EXPORT_SYMBOL_GPL(seg6_do_srh_inline);
373
374static int seg6_do_srh(struct sk_buff *skb)
375{
376 struct dst_entry *dst = skb_dst(skb);
377 struct seg6_iptunnel_encap *tinfo;
378 int proto, err = 0;
379
380 tinfo = seg6_encap_lwtunnel(dst->lwtstate);
381
382 switch (tinfo->mode) {
383 case SEG6_IPTUN_MODE_INLINE:
384 if (skb->protocol != htons(ETH_P_IPV6))
385 return -EINVAL;
386
387 err = seg6_do_srh_inline(skb, tinfo->srh);
388 if (err)
389 return err;
390 break;
391 case SEG6_IPTUN_MODE_ENCAP:
392 case SEG6_IPTUN_MODE_ENCAP_RED:
393 err = iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6);
394 if (err)
395 return err;
396
397 if (skb->protocol == htons(ETH_P_IPV6))
398 proto = IPPROTO_IPV6;
399 else if (skb->protocol == htons(ETH_P_IP))
400 proto = IPPROTO_IPIP;
401 else
402 return -EINVAL;
403
404 if (tinfo->mode == SEG6_IPTUN_MODE_ENCAP)
405 err = seg6_do_srh_encap(skb, tinfo->srh, proto);
406 else
407 err = seg6_do_srh_encap_red(skb, tinfo->srh, proto);
408
409 if (err)
410 return err;
411
412 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
413 skb_set_inner_protocol(skb, skb->protocol);
414 skb->protocol = htons(ETH_P_IPV6);
415 break;
416 case SEG6_IPTUN_MODE_L2ENCAP:
417 case SEG6_IPTUN_MODE_L2ENCAP_RED:
418 if (!skb_mac_header_was_set(skb))
419 return -EINVAL;
420
421 if (pskb_expand_head(skb, skb->mac_len, 0, GFP_ATOMIC) < 0)
422 return -ENOMEM;
423
424 skb_mac_header_rebuild(skb);
425 skb_push(skb, skb->mac_len);
426
427 if (tinfo->mode == SEG6_IPTUN_MODE_L2ENCAP)
428 err = seg6_do_srh_encap(skb, tinfo->srh,
429 IPPROTO_ETHERNET);
430 else
431 err = seg6_do_srh_encap_red(skb, tinfo->srh,
432 IPPROTO_ETHERNET);
433
434 if (err)
435 return err;
436
437 skb->protocol = htons(ETH_P_IPV6);
438 break;
439 }
440
441 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
442 nf_reset_ct(skb);
443
444 return 0;
445}
446
447static int seg6_input_finish(struct net *net, struct sock *sk,
448 struct sk_buff *skb)
449{
450 return dst_input(skb);
451}
452
453static int seg6_input_core(struct net *net, struct sock *sk,
454 struct sk_buff *skb)
455{
456 struct dst_entry *orig_dst = skb_dst(skb);
457 struct dst_entry *dst = NULL;
458 struct seg6_lwt *slwt;
459 int err;
460
461 err = seg6_do_srh(skb);
462 if (unlikely(err)) {
463 kfree_skb(skb);
464 return err;
465 }
466
467 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
468
469 preempt_disable();
470 dst = dst_cache_get(&slwt->cache);
471 preempt_enable();
472
473 if (!dst) {
474 ip6_route_input(skb);
475 dst = skb_dst(skb);
476 if (!dst->error) {
477 preempt_disable();
478 dst_cache_set_ip6(&slwt->cache, dst,
479 &ipv6_hdr(skb)->saddr);
480 preempt_enable();
481 }
482 } else {
483 skb_dst_drop(skb);
484 skb_dst_set(skb, dst);
485 }
486
487 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
488 if (unlikely(err))
489 return err;
490
491 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
492 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
493 dev_net(skb->dev), NULL, skb, NULL,
494 skb_dst(skb)->dev, seg6_input_finish);
495
496 return seg6_input_finish(dev_net(skb->dev), NULL, skb);
497}
498
499static int seg6_input_nf(struct sk_buff *skb)
500{
501 struct net_device *dev = skb_dst(skb)->dev;
502 struct net *net = dev_net(skb->dev);
503
504 switch (skb->protocol) {
505 case htons(ETH_P_IP):
506 return NF_HOOK(NFPROTO_IPV4, NF_INET_POST_ROUTING, net, NULL,
507 skb, NULL, dev, seg6_input_core);
508 case htons(ETH_P_IPV6):
509 return NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, net, NULL,
510 skb, NULL, dev, seg6_input_core);
511 }
512
513 return -EINVAL;
514}
515
516static int seg6_input(struct sk_buff *skb)
517{
518 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
519 return seg6_input_nf(skb);
520
521 return seg6_input_core(dev_net(skb->dev), NULL, skb);
522}
523
524static int seg6_output_core(struct net *net, struct sock *sk,
525 struct sk_buff *skb)
526{
527 struct dst_entry *orig_dst = skb_dst(skb);
528 struct dst_entry *dst = NULL;
529 struct seg6_lwt *slwt;
530 int err;
531
532 err = seg6_do_srh(skb);
533 if (unlikely(err))
534 goto drop;
535
536 slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
537
538 preempt_disable();
539 dst = dst_cache_get(&slwt->cache);
540 preempt_enable();
541
542 if (unlikely(!dst)) {
543 struct ipv6hdr *hdr = ipv6_hdr(skb);
544 struct flowi6 fl6;
545
546 memset(&fl6, 0, sizeof(fl6));
547 fl6.daddr = hdr->daddr;
548 fl6.saddr = hdr->saddr;
549 fl6.flowlabel = ip6_flowinfo(hdr);
550 fl6.flowi6_mark = skb->mark;
551 fl6.flowi6_proto = hdr->nexthdr;
552
553 dst = ip6_route_output(net, NULL, &fl6);
554 if (dst->error) {
555 err = dst->error;
556 dst_release(dst);
557 goto drop;
558 }
559
560 preempt_disable();
561 dst_cache_set_ip6(&slwt->cache, dst, &fl6.saddr);
562 preempt_enable();
563 }
564
565 skb_dst_drop(skb);
566 skb_dst_set(skb, dst);
567
568 err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
569 if (unlikely(err))
570 goto drop;
571
572 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
573 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
574 NULL, skb_dst(skb)->dev, dst_output);
575
576 return dst_output(net, sk, skb);
577drop:
578 kfree_skb(skb);
579 return err;
580}
581
582static int seg6_output_nf(struct net *net, struct sock *sk, struct sk_buff *skb)
583{
584 struct net_device *dev = skb_dst(skb)->dev;
585
586 switch (skb->protocol) {
587 case htons(ETH_P_IP):
588 return NF_HOOK(NFPROTO_IPV4, NF_INET_POST_ROUTING, net, sk, skb,
589 NULL, dev, seg6_output_core);
590 case htons(ETH_P_IPV6):
591 return NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, net, sk, skb,
592 NULL, dev, seg6_output_core);
593 }
594
595 return -EINVAL;
596}
597
598static int seg6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
599{
600 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
601 return seg6_output_nf(net, sk, skb);
602
603 return seg6_output_core(net, sk, skb);
604}
605
606static int seg6_build_state(struct net *net, struct nlattr *nla,
607 unsigned int family, const void *cfg,
608 struct lwtunnel_state **ts,
609 struct netlink_ext_ack *extack)
610{
611 struct nlattr *tb[SEG6_IPTUNNEL_MAX + 1];
612 struct seg6_iptunnel_encap *tuninfo;
613 struct lwtunnel_state *newts;
614 int tuninfo_len, min_size;
615 struct seg6_lwt *slwt;
616 int err;
617
618 if (family != AF_INET && family != AF_INET6)
619 return -EINVAL;
620
621 err = nla_parse_nested_deprecated(tb, SEG6_IPTUNNEL_MAX, nla,
622 seg6_iptunnel_policy, extack);
623
624 if (err < 0)
625 return err;
626
627 if (!tb[SEG6_IPTUNNEL_SRH])
628 return -EINVAL;
629
630 tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
631 tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
632
633 /* tuninfo must contain at least the iptunnel encap structure,
634 * the SRH and one segment
635 */
636 min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
637 sizeof(struct in6_addr);
638 if (tuninfo_len < min_size)
639 return -EINVAL;
640
641 switch (tuninfo->mode) {
642 case SEG6_IPTUN_MODE_INLINE:
643 if (family != AF_INET6)
644 return -EINVAL;
645
646 break;
647 case SEG6_IPTUN_MODE_ENCAP:
648 break;
649 case SEG6_IPTUN_MODE_L2ENCAP:
650 break;
651 case SEG6_IPTUN_MODE_ENCAP_RED:
652 break;
653 case SEG6_IPTUN_MODE_L2ENCAP_RED:
654 break;
655 default:
656 return -EINVAL;
657 }
658
659 /* verify that SRH is consistent */
660 if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo), false))
661 return -EINVAL;
662
663 newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
664 if (!newts)
665 return -ENOMEM;
666
667 slwt = seg6_lwt_lwtunnel(newts);
668
669 err = dst_cache_init(&slwt->cache, GFP_ATOMIC);
670 if (err) {
671 kfree(newts);
672 return err;
673 }
674
675 memcpy(&slwt->tuninfo, tuninfo, tuninfo_len);
676
677 newts->type = LWTUNNEL_ENCAP_SEG6;
678 newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
679
680 if (tuninfo->mode != SEG6_IPTUN_MODE_L2ENCAP)
681 newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
682
683 newts->headroom = seg6_lwt_headroom(tuninfo);
684
685 *ts = newts;
686
687 return 0;
688}
689
690static void seg6_destroy_state(struct lwtunnel_state *lwt)
691{
692 dst_cache_destroy(&seg6_lwt_lwtunnel(lwt)->cache);
693}
694
695static int seg6_fill_encap_info(struct sk_buff *skb,
696 struct lwtunnel_state *lwtstate)
697{
698 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
699
700 if (nla_put_srh(skb, SEG6_IPTUNNEL_SRH, tuninfo))
701 return -EMSGSIZE;
702
703 return 0;
704}
705
706static int seg6_encap_nlsize(struct lwtunnel_state *lwtstate)
707{
708 struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
709
710 return nla_total_size(SEG6_IPTUN_ENCAP_SIZE(tuninfo));
711}
712
713static int seg6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
714{
715 struct seg6_iptunnel_encap *a_hdr = seg6_encap_lwtunnel(a);
716 struct seg6_iptunnel_encap *b_hdr = seg6_encap_lwtunnel(b);
717 int len = SEG6_IPTUN_ENCAP_SIZE(a_hdr);
718
719 if (len != SEG6_IPTUN_ENCAP_SIZE(b_hdr))
720 return 1;
721
722 return memcmp(a_hdr, b_hdr, len);
723}
724
725static const struct lwtunnel_encap_ops seg6_iptun_ops = {
726 .build_state = seg6_build_state,
727 .destroy_state = seg6_destroy_state,
728 .output = seg6_output,
729 .input = seg6_input,
730 .fill_encap = seg6_fill_encap_info,
731 .get_encap_size = seg6_encap_nlsize,
732 .cmp_encap = seg6_encap_cmp,
733 .owner = THIS_MODULE,
734};
735
736int __init seg6_iptunnel_init(void)
737{
738 return lwtunnel_encap_add_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
739}
740
741void seg6_iptunnel_exit(void)
742{
743 lwtunnel_encap_del_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
744}