1
2
3
4
5
6#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
7
8#include <linux/types.h>
9#include <linux/kernel.h>
10#include <linux/skbuff.h>
11#include <linux/netdevice.h>
12#include <linux/in.h>
13#include <linux/if_arp.h>
14#include <linux/init.h>
15#include <linux/in6.h>
16#include <linux/inetdevice.h>
17#include <linux/netfilter_ipv4.h>
18#include <linux/etherdevice.h>
19#include <linux/if_ether.h>
20#include <linux/if_vlan.h>
21#include <linux/static_key.h>
22
23#include <net/ip.h>
24#include <net/icmp.h>
25#include <net/protocol.h>
26#include <net/ip_tunnels.h>
27#include <net/ip6_tunnel.h>
28#include <net/ip6_checksum.h>
29#include <net/arp.h>
30#include <net/checksum.h>
31#include <net/dsfield.h>
32#include <net/inet_ecn.h>
33#include <net/xfrm.h>
34#include <net/net_namespace.h>
35#include <net/netns/generic.h>
36#include <net/rtnetlink.h>
37#include <net/dst_metadata.h>
38#include <net/geneve.h>
39#include <net/vxlan.h>
40#include <net/erspan.h>
41
42const struct ip_tunnel_encap_ops __rcu *
43 iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
44EXPORT_SYMBOL(iptun_encaps);
45
46const struct ip6_tnl_encap_ops __rcu *
47 ip6tun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
48EXPORT_SYMBOL(ip6tun_encaps);
49
50void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
51 __be32 src, __be32 dst, __u8 proto,
52 __u8 tos, __u8 ttl, __be16 df, bool xnet)
53{
54 int pkt_len = skb->len - skb_inner_network_offset(skb);
55 struct net *net = dev_net(rt->dst.dev);
56 struct net_device *dev = skb->dev;
57 struct iphdr *iph;
58 int err;
59
60 skb_scrub_packet(skb, xnet);
61
62 skb_clear_hash_if_not_l4(skb);
63 skb_dst_set(skb, &rt->dst);
64 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
65
66
67 skb_push(skb, sizeof(struct iphdr));
68 skb_reset_network_header(skb);
69
70 iph = ip_hdr(skb);
71
72 iph->version = 4;
73 iph->ihl = sizeof(struct iphdr) >> 2;
74 iph->frag_off = ip_mtu_locked(&rt->dst) ? 0 : df;
75 iph->protocol = proto;
76 iph->tos = tos;
77 iph->daddr = dst;
78 iph->saddr = src;
79 iph->ttl = ttl;
80 __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
81
82 err = ip_local_out(net, sk, skb);
83
84 if (dev) {
85 if (unlikely(net_xmit_eval(err)))
86 pkt_len = 0;
87 iptunnel_xmit_stats(dev, pkt_len);
88 }
89}
90EXPORT_SYMBOL_GPL(iptunnel_xmit);
91
92int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
93 __be16 inner_proto, bool raw_proto, bool xnet)
94{
95 if (unlikely(!pskb_may_pull(skb, hdr_len)))
96 return -ENOMEM;
97
98 skb_pull_rcsum(skb, hdr_len);
99
100 if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
101 struct ethhdr *eh;
102
103 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
104 return -ENOMEM;
105
106 eh = (struct ethhdr *)skb->data;
107 if (likely(eth_proto_is_802_3(eh->h_proto)))
108 skb->protocol = eh->h_proto;
109 else
110 skb->protocol = htons(ETH_P_802_2);
111
112 } else {
113 skb->protocol = inner_proto;
114 }
115
116 skb_clear_hash_if_not_l4(skb);
117 __vlan_hwaccel_clear_tag(skb);
118 skb_set_queue_mapping(skb, 0);
119 skb_scrub_packet(skb, xnet);
120
121 return iptunnel_pull_offloads(skb);
122}
123EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
124
125struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
126 gfp_t flags)
127{
128 struct metadata_dst *res;
129 struct ip_tunnel_info *dst, *src;
130
131 if (!md || md->type != METADATA_IP_TUNNEL ||
132 md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
133 return NULL;
134
135 src = &md->u.tun_info;
136 res = metadata_dst_alloc(src->options_len, METADATA_IP_TUNNEL, flags);
137 if (!res)
138 return NULL;
139
140 dst = &res->u.tun_info;
141 dst->key.tun_id = src->key.tun_id;
142 if (src->mode & IP_TUNNEL_INFO_IPV6)
143 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
144 sizeof(struct in6_addr));
145 else
146 dst->key.u.ipv4.dst = src->key.u.ipv4.src;
147 dst->key.tun_flags = src->key.tun_flags;
148 dst->mode = src->mode | IP_TUNNEL_INFO_TX;
149 ip_tunnel_info_opts_set(dst, ip_tunnel_info_opts(src),
150 src->options_len, 0);
151
152 return res;
153}
154EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
155
156int iptunnel_handle_offloads(struct sk_buff *skb,
157 int gso_type_mask)
158{
159 int err;
160
161 if (likely(!skb->encapsulation)) {
162 skb_reset_inner_headers(skb);
163 skb->encapsulation = 1;
164 }
165
166 if (skb_is_gso(skb)) {
167 err = skb_header_unclone(skb, GFP_ATOMIC);
168 if (unlikely(err))
169 return err;
170 skb_shinfo(skb)->gso_type |= gso_type_mask;
171 return 0;
172 }
173
174 if (skb->ip_summed != CHECKSUM_PARTIAL) {
175 skb->ip_summed = CHECKSUM_NONE;
176
177
178
179
180
181 skb->encapsulation = 0;
182 }
183
184 return 0;
185}
186EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
187
188
189
190
191
192
193
194
195static int iptunnel_pmtud_build_icmp(struct sk_buff *skb, int mtu)
196{
197 const struct iphdr *iph = ip_hdr(skb);
198 struct icmphdr *icmph;
199 struct iphdr *niph;
200 struct ethhdr eh;
201 int len, err;
202
203 if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct iphdr)))
204 return -EINVAL;
205
206 skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
207 pskb_pull(skb, ETH_HLEN);
208 skb_reset_network_header(skb);
209
210 err = pskb_trim(skb, 576 - sizeof(*niph) - sizeof(*icmph));
211 if (err)
212 return err;
213
214 len = skb->len + sizeof(*icmph);
215 err = skb_cow(skb, sizeof(*niph) + sizeof(*icmph) + ETH_HLEN);
216 if (err)
217 return err;
218
219 icmph = skb_push(skb, sizeof(*icmph));
220 *icmph = (struct icmphdr) {
221 .type = ICMP_DEST_UNREACH,
222 .code = ICMP_FRAG_NEEDED,
223 .checksum = 0,
224 .un.frag.__unused = 0,
225 .un.frag.mtu = ntohs(mtu),
226 };
227 icmph->checksum = ip_compute_csum(icmph, len);
228 skb_reset_transport_header(skb);
229
230 niph = skb_push(skb, sizeof(*niph));
231 *niph = (struct iphdr) {
232 .ihl = sizeof(*niph) / 4u,
233 .version = 4,
234 .tos = 0,
235 .tot_len = htons(len + sizeof(*niph)),
236 .id = 0,
237 .frag_off = htons(IP_DF),
238 .ttl = iph->ttl,
239 .protocol = IPPROTO_ICMP,
240 .saddr = iph->daddr,
241 .daddr = iph->saddr,
242 };
243 ip_send_check(niph);
244 skb_reset_network_header(skb);
245
246 skb->ip_summed = CHECKSUM_NONE;
247
248 eth_header(skb, skb->dev, htons(eh.h_proto), eh.h_source, eh.h_dest, 0);
249 skb_reset_mac_header(skb);
250
251 return skb->len;
252}
253
254
255
256
257
258
259
260
261static int iptunnel_pmtud_check_icmp(struct sk_buff *skb, int mtu)
262{
263 const struct icmphdr *icmph = icmp_hdr(skb);
264 const struct iphdr *iph = ip_hdr(skb);
265
266 if (mtu < 576 || iph->frag_off != htons(IP_DF))
267 return 0;
268
269 if (ipv4_is_lbcast(iph->daddr) || ipv4_is_multicast(iph->daddr) ||
270 ipv4_is_zeronet(iph->saddr) || ipv4_is_loopback(iph->saddr) ||
271 ipv4_is_lbcast(iph->saddr) || ipv4_is_multicast(iph->saddr))
272 return 0;
273
274 if (iph->protocol == IPPROTO_ICMP && icmp_is_err(icmph->type))
275 return 0;
276
277 return iptunnel_pmtud_build_icmp(skb, mtu);
278}
279
280#if IS_ENABLED(CONFIG_IPV6)
281
282
283
284
285
286
287
288static int iptunnel_pmtud_build_icmpv6(struct sk_buff *skb, int mtu)
289{
290 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
291 struct icmp6hdr *icmp6h;
292 struct ipv6hdr *nip6h;
293 struct ethhdr eh;
294 int len, err;
295 __wsum csum;
296
297 if (!pskb_may_pull(skb, ETH_HLEN + sizeof(struct ipv6hdr)))
298 return -EINVAL;
299
300 skb_copy_bits(skb, skb_mac_offset(skb), &eh, ETH_HLEN);
301 pskb_pull(skb, ETH_HLEN);
302 skb_reset_network_header(skb);
303
304 err = pskb_trim(skb, IPV6_MIN_MTU - sizeof(*nip6h) - sizeof(*icmp6h));
305 if (err)
306 return err;
307
308 len = skb->len + sizeof(*icmp6h);
309 err = skb_cow(skb, sizeof(*nip6h) + sizeof(*icmp6h) + ETH_HLEN);
310 if (err)
311 return err;
312
313 icmp6h = skb_push(skb, sizeof(*icmp6h));
314 *icmp6h = (struct icmp6hdr) {
315 .icmp6_type = ICMPV6_PKT_TOOBIG,
316 .icmp6_code = 0,
317 .icmp6_cksum = 0,
318 .icmp6_mtu = htonl(mtu),
319 };
320 skb_reset_transport_header(skb);
321
322 nip6h = skb_push(skb, sizeof(*nip6h));
323 *nip6h = (struct ipv6hdr) {
324 .priority = 0,
325 .version = 6,
326 .flow_lbl = { 0 },
327 .payload_len = htons(len),
328 .nexthdr = IPPROTO_ICMPV6,
329 .hop_limit = ip6h->hop_limit,
330 .saddr = ip6h->daddr,
331 .daddr = ip6h->saddr,
332 };
333 skb_reset_network_header(skb);
334
335 csum = csum_partial(icmp6h, len, 0);
336 icmp6h->icmp6_cksum = csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr, len,
337 IPPROTO_ICMPV6, csum);
338
339 skb->ip_summed = CHECKSUM_NONE;
340
341 eth_header(skb, skb->dev, htons(eh.h_proto), eh.h_source, eh.h_dest, 0);
342 skb_reset_mac_header(skb);
343
344 return skb->len;
345}
346
347
348
349
350
351
352
353
354static int iptunnel_pmtud_check_icmpv6(struct sk_buff *skb, int mtu)
355{
356 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
357 int stype = ipv6_addr_type(&ip6h->saddr);
358 u8 proto = ip6h->nexthdr;
359 __be16 frag_off;
360 int offset;
361
362 if (mtu < IPV6_MIN_MTU)
363 return 0;
364
365 if (stype == IPV6_ADDR_ANY || stype == IPV6_ADDR_MULTICAST ||
366 stype == IPV6_ADDR_LOOPBACK)
367 return 0;
368
369 offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &proto,
370 &frag_off);
371 if (offset < 0 || (frag_off & htons(~0x7)))
372 return 0;
373
374 if (proto == IPPROTO_ICMPV6) {
375 struct icmp6hdr *icmp6h;
376
377 if (!pskb_may_pull(skb, skb_network_header(skb) +
378 offset + 1 - skb->data))
379 return 0;
380
381 icmp6h = (struct icmp6hdr *)(skb_network_header(skb) + offset);
382 if (icmpv6_is_err(icmp6h->icmp6_type) ||
383 icmp6h->icmp6_type == NDISC_REDIRECT)
384 return 0;
385 }
386
387 return iptunnel_pmtud_build_icmpv6(skb, mtu);
388}
389#endif
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
408 int headroom, bool reply)
409{
410 u32 mtu = dst_mtu(encap_dst) - headroom;
411
412 if ((skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu)) ||
413 (!skb_is_gso(skb) && (skb->len - skb_mac_header_len(skb)) <= mtu))
414 return 0;
415
416 skb_dst_update_pmtu_no_confirm(skb, mtu);
417
418 if (!reply || skb->pkt_type == PACKET_HOST)
419 return 0;
420
421 if (skb->protocol == htons(ETH_P_IP))
422 return iptunnel_pmtud_check_icmp(skb, mtu);
423
424#if IS_ENABLED(CONFIG_IPV6)
425 if (skb->protocol == htons(ETH_P_IPV6))
426 return iptunnel_pmtud_check_icmpv6(skb, mtu);
427#endif
428 return 0;
429}
430EXPORT_SYMBOL(skb_tunnel_check_pmtu);
431
432static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
433 [LWTUNNEL_IP_UNSPEC] = { .strict_start_type = LWTUNNEL_IP_OPTS },
434 [LWTUNNEL_IP_ID] = { .type = NLA_U64 },
435 [LWTUNNEL_IP_DST] = { .type = NLA_U32 },
436 [LWTUNNEL_IP_SRC] = { .type = NLA_U32 },
437 [LWTUNNEL_IP_TTL] = { .type = NLA_U8 },
438 [LWTUNNEL_IP_TOS] = { .type = NLA_U8 },
439 [LWTUNNEL_IP_FLAGS] = { .type = NLA_U16 },
440 [LWTUNNEL_IP_OPTS] = { .type = NLA_NESTED },
441};
442
443static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = {
444 [LWTUNNEL_IP_OPTS_GENEVE] = { .type = NLA_NESTED },
445 [LWTUNNEL_IP_OPTS_VXLAN] = { .type = NLA_NESTED },
446 [LWTUNNEL_IP_OPTS_ERSPAN] = { .type = NLA_NESTED },
447};
448
449static const struct nla_policy
450geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = {
451 [LWTUNNEL_IP_OPT_GENEVE_CLASS] = { .type = NLA_U16 },
452 [LWTUNNEL_IP_OPT_GENEVE_TYPE] = { .type = NLA_U8 },
453 [LWTUNNEL_IP_OPT_GENEVE_DATA] = { .type = NLA_BINARY, .len = 128 },
454};
455
456static const struct nla_policy
457vxlan_opt_policy[LWTUNNEL_IP_OPT_VXLAN_MAX + 1] = {
458 [LWTUNNEL_IP_OPT_VXLAN_GBP] = { .type = NLA_U32 },
459};
460
461static const struct nla_policy
462erspan_opt_policy[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1] = {
463 [LWTUNNEL_IP_OPT_ERSPAN_VER] = { .type = NLA_U8 },
464 [LWTUNNEL_IP_OPT_ERSPAN_INDEX] = { .type = NLA_U32 },
465 [LWTUNNEL_IP_OPT_ERSPAN_DIR] = { .type = NLA_U8 },
466 [LWTUNNEL_IP_OPT_ERSPAN_HWID] = { .type = NLA_U8 },
467};
468
469static int ip_tun_parse_opts_geneve(struct nlattr *attr,
470 struct ip_tunnel_info *info, int opts_len,
471 struct netlink_ext_ack *extack)
472{
473 struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1];
474 int data_len, err;
475
476 err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_GENEVE_MAX, attr,
477 geneve_opt_policy, extack);
478 if (err)
479 return err;
480
481 if (!tb[LWTUNNEL_IP_OPT_GENEVE_CLASS] ||
482 !tb[LWTUNNEL_IP_OPT_GENEVE_TYPE] ||
483 !tb[LWTUNNEL_IP_OPT_GENEVE_DATA])
484 return -EINVAL;
485
486 attr = tb[LWTUNNEL_IP_OPT_GENEVE_DATA];
487 data_len = nla_len(attr);
488 if (data_len % 4)
489 return -EINVAL;
490
491 if (info) {
492 struct geneve_opt *opt = ip_tunnel_info_opts(info) + opts_len;
493
494 memcpy(opt->opt_data, nla_data(attr), data_len);
495 opt->length = data_len / 4;
496 attr = tb[LWTUNNEL_IP_OPT_GENEVE_CLASS];
497 opt->opt_class = nla_get_be16(attr);
498 attr = tb[LWTUNNEL_IP_OPT_GENEVE_TYPE];
499 opt->type = nla_get_u8(attr);
500 info->key.tun_flags |= TUNNEL_GENEVE_OPT;
501 }
502
503 return sizeof(struct geneve_opt) + data_len;
504}
505
506static int ip_tun_parse_opts_vxlan(struct nlattr *attr,
507 struct ip_tunnel_info *info, int opts_len,
508 struct netlink_ext_ack *extack)
509{
510 struct nlattr *tb[LWTUNNEL_IP_OPT_VXLAN_MAX + 1];
511 int err;
512
513 err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_VXLAN_MAX, attr,
514 vxlan_opt_policy, extack);
515 if (err)
516 return err;
517
518 if (!tb[LWTUNNEL_IP_OPT_VXLAN_GBP])
519 return -EINVAL;
520
521 if (info) {
522 struct vxlan_metadata *md =
523 ip_tunnel_info_opts(info) + opts_len;
524
525 attr = tb[LWTUNNEL_IP_OPT_VXLAN_GBP];
526 md->gbp = nla_get_u32(attr);
527 md->gbp &= VXLAN_GBP_MASK;
528 info->key.tun_flags |= TUNNEL_VXLAN_OPT;
529 }
530
531 return sizeof(struct vxlan_metadata);
532}
533
534static int ip_tun_parse_opts_erspan(struct nlattr *attr,
535 struct ip_tunnel_info *info, int opts_len,
536 struct netlink_ext_ack *extack)
537{
538 struct nlattr *tb[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1];
539 int err;
540 u8 ver;
541
542 err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_ERSPAN_MAX, attr,
543 erspan_opt_policy, extack);
544 if (err)
545 return err;
546
547 if (!tb[LWTUNNEL_IP_OPT_ERSPAN_VER])
548 return -EINVAL;
549
550 ver = nla_get_u8(tb[LWTUNNEL_IP_OPT_ERSPAN_VER]);
551 if (ver == 1) {
552 if (!tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX])
553 return -EINVAL;
554 } else if (ver == 2) {
555 if (!tb[LWTUNNEL_IP_OPT_ERSPAN_DIR] ||
556 !tb[LWTUNNEL_IP_OPT_ERSPAN_HWID])
557 return -EINVAL;
558 } else {
559 return -EINVAL;
560 }
561
562 if (info) {
563 struct erspan_metadata *md =
564 ip_tunnel_info_opts(info) + opts_len;
565
566 md->version = ver;
567 if (ver == 1) {
568 attr = tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX];
569 md->u.index = nla_get_be32(attr);
570 } else {
571 attr = tb[LWTUNNEL_IP_OPT_ERSPAN_DIR];
572 md->u.md2.dir = nla_get_u8(attr);
573 attr = tb[LWTUNNEL_IP_OPT_ERSPAN_HWID];
574 set_hwid(&md->u.md2, nla_get_u8(attr));
575 }
576
577 info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
578 }
579
580 return sizeof(struct erspan_metadata);
581}
582
583static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info,
584 struct netlink_ext_ack *extack)
585{
586 int err, rem, opt_len, opts_len = 0, type = 0;
587 struct nlattr *nla;
588
589 if (!attr)
590 return 0;
591
592 err = nla_validate(nla_data(attr), nla_len(attr), LWTUNNEL_IP_OPTS_MAX,
593 ip_opts_policy, extack);
594 if (err)
595 return err;
596
597 nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
598 switch (nla_type(nla)) {
599 case LWTUNNEL_IP_OPTS_GENEVE:
600 if (type && type != TUNNEL_GENEVE_OPT)
601 return -EINVAL;
602 opt_len = ip_tun_parse_opts_geneve(nla, info, opts_len,
603 extack);
604 if (opt_len < 0)
605 return opt_len;
606 opts_len += opt_len;
607 if (opts_len > IP_TUNNEL_OPTS_MAX)
608 return -EINVAL;
609 type = TUNNEL_GENEVE_OPT;
610 break;
611 case LWTUNNEL_IP_OPTS_VXLAN:
612 if (type)
613 return -EINVAL;
614 opt_len = ip_tun_parse_opts_vxlan(nla, info, opts_len,
615 extack);
616 if (opt_len < 0)
617 return opt_len;
618 opts_len += opt_len;
619 type = TUNNEL_VXLAN_OPT;
620 break;
621 case LWTUNNEL_IP_OPTS_ERSPAN:
622 if (type)
623 return -EINVAL;
624 opt_len = ip_tun_parse_opts_erspan(nla, info, opts_len,
625 extack);
626 if (opt_len < 0)
627 return opt_len;
628 opts_len += opt_len;
629 type = TUNNEL_ERSPAN_OPT;
630 break;
631 default:
632 return -EINVAL;
633 }
634 }
635
636 return opts_len;
637}
638
639static int ip_tun_get_optlen(struct nlattr *attr,
640 struct netlink_ext_ack *extack)
641{
642 return ip_tun_parse_opts(attr, NULL, extack);
643}
644
645static int ip_tun_set_opts(struct nlattr *attr, struct ip_tunnel_info *info,
646 struct netlink_ext_ack *extack)
647{
648 return ip_tun_parse_opts(attr, info, extack);
649}
650
651static int ip_tun_build_state(struct net *net, struct nlattr *attr,
652 unsigned int family, const void *cfg,
653 struct lwtunnel_state **ts,
654 struct netlink_ext_ack *extack)
655{
656 struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
657 struct lwtunnel_state *new_state;
658 struct ip_tunnel_info *tun_info;
659 int err, opt_len;
660
661 err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_MAX, attr,
662 ip_tun_policy, extack);
663 if (err < 0)
664 return err;
665
666 opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP_OPTS], extack);
667 if (opt_len < 0)
668 return opt_len;
669
670 new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
671 if (!new_state)
672 return -ENOMEM;
673
674 new_state->type = LWTUNNEL_ENCAP_IP;
675
676 tun_info = lwt_tun_info(new_state);
677
678 err = ip_tun_set_opts(tb[LWTUNNEL_IP_OPTS], tun_info, extack);
679 if (err < 0) {
680 lwtstate_free(new_state);
681 return err;
682 }
683
684#ifdef CONFIG_DST_CACHE
685 err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL);
686 if (err) {
687 lwtstate_free(new_state);
688 return err;
689 }
690#endif
691
692 if (tb[LWTUNNEL_IP_ID])
693 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
694
695 if (tb[LWTUNNEL_IP_DST])
696 tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
697
698 if (tb[LWTUNNEL_IP_SRC])
699 tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
700
701 if (tb[LWTUNNEL_IP_TTL])
702 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
703
704 if (tb[LWTUNNEL_IP_TOS])
705 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
706
707 if (tb[LWTUNNEL_IP_FLAGS])
708 tun_info->key.tun_flags |=
709 (nla_get_be16(tb[LWTUNNEL_IP_FLAGS]) &
710 ~TUNNEL_OPTIONS_PRESENT);
711
712 tun_info->mode = IP_TUNNEL_INFO_TX;
713 tun_info->options_len = opt_len;
714
715 *ts = new_state;
716
717 return 0;
718}
719
720static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate)
721{
722#ifdef CONFIG_DST_CACHE
723 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
724
725 dst_cache_destroy(&tun_info->dst_cache);
726#endif
727}
728
729static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
730 struct ip_tunnel_info *tun_info)
731{
732 struct geneve_opt *opt;
733 struct nlattr *nest;
734 int offset = 0;
735
736 nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE);
737 if (!nest)
738 return -ENOMEM;
739
740 while (tun_info->options_len > offset) {
741 opt = ip_tunnel_info_opts(tun_info) + offset;
742 if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS,
743 opt->opt_class) ||
744 nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
745 nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
746 opt->opt_data)) {
747 nla_nest_cancel(skb, nest);
748 return -ENOMEM;
749 }
750 offset += sizeof(*opt) + opt->length * 4;
751 }
752
753 nla_nest_end(skb, nest);
754 return 0;
755}
756
757static int ip_tun_fill_encap_opts_vxlan(struct sk_buff *skb,
758 struct ip_tunnel_info *tun_info)
759{
760 struct vxlan_metadata *md;
761 struct nlattr *nest;
762
763 nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_VXLAN);
764 if (!nest)
765 return -ENOMEM;
766
767 md = ip_tunnel_info_opts(tun_info);
768 if (nla_put_u32(skb, LWTUNNEL_IP_OPT_VXLAN_GBP, md->gbp)) {
769 nla_nest_cancel(skb, nest);
770 return -ENOMEM;
771 }
772
773 nla_nest_end(skb, nest);
774 return 0;
775}
776
777static int ip_tun_fill_encap_opts_erspan(struct sk_buff *skb,
778 struct ip_tunnel_info *tun_info)
779{
780 struct erspan_metadata *md;
781 struct nlattr *nest;
782
783 nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_ERSPAN);
784 if (!nest)
785 return -ENOMEM;
786
787 md = ip_tunnel_info_opts(tun_info);
788 if (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_VER, md->version))
789 goto err;
790
791 if (md->version == 1 &&
792 nla_put_be32(skb, LWTUNNEL_IP_OPT_ERSPAN_INDEX, md->u.index))
793 goto err;
794
795 if (md->version == 2 &&
796 (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_DIR, md->u.md2.dir) ||
797 nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_HWID,
798 get_hwid(&md->u.md2))))
799 goto err;
800
801 nla_nest_end(skb, nest);
802 return 0;
803err:
804 nla_nest_cancel(skb, nest);
805 return -ENOMEM;
806}
807
808static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
809 struct ip_tunnel_info *tun_info)
810{
811 struct nlattr *nest;
812 int err = 0;
813
814 if (!(tun_info->key.tun_flags & TUNNEL_OPTIONS_PRESENT))
815 return 0;
816
817 nest = nla_nest_start_noflag(skb, type);
818 if (!nest)
819 return -ENOMEM;
820
821 if (tun_info->key.tun_flags & TUNNEL_GENEVE_OPT)
822 err = ip_tun_fill_encap_opts_geneve(skb, tun_info);
823 else if (tun_info->key.tun_flags & TUNNEL_VXLAN_OPT)
824 err = ip_tun_fill_encap_opts_vxlan(skb, tun_info);
825 else if (tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)
826 err = ip_tun_fill_encap_opts_erspan(skb, tun_info);
827
828 if (err) {
829 nla_nest_cancel(skb, nest);
830 return err;
831 }
832
833 nla_nest_end(skb, nest);
834 return 0;
835}
836
837static int ip_tun_fill_encap_info(struct sk_buff *skb,
838 struct lwtunnel_state *lwtstate)
839{
840 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
841
842 if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
843 LWTUNNEL_IP_PAD) ||
844 nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
845 nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
846 nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
847 nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
848 nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags) ||
849 ip_tun_fill_encap_opts(skb, LWTUNNEL_IP_OPTS, tun_info))
850 return -ENOMEM;
851
852 return 0;
853}
854
855static int ip_tun_opts_nlsize(struct ip_tunnel_info *info)
856{
857 int opt_len;
858
859 if (!(info->key.tun_flags & TUNNEL_OPTIONS_PRESENT))
860 return 0;
861
862 opt_len = nla_total_size(0);
863 if (info->key.tun_flags & TUNNEL_GENEVE_OPT) {
864 struct geneve_opt *opt;
865 int offset = 0;
866
867 opt_len += nla_total_size(0);
868 while (info->options_len > offset) {
869 opt = ip_tunnel_info_opts(info) + offset;
870 opt_len += nla_total_size(2)
871 + nla_total_size(1)
872 + nla_total_size(opt->length * 4);
873
874 offset += sizeof(*opt) + opt->length * 4;
875 }
876 } else if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
877 opt_len += nla_total_size(0)
878 + nla_total_size(4);
879 } else if (info->key.tun_flags & TUNNEL_ERSPAN_OPT) {
880 struct erspan_metadata *md = ip_tunnel_info_opts(info);
881
882 opt_len += nla_total_size(0)
883 + nla_total_size(1)
884 + (md->version == 1 ? nla_total_size(4)
885
886 : nla_total_size(1) +
887 nla_total_size(1));
888
889 }
890
891 return opt_len;
892}
893
894static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
895{
896 return nla_total_size_64bit(8)
897 + nla_total_size(4)
898 + nla_total_size(4)
899 + nla_total_size(1)
900 + nla_total_size(1)
901 + nla_total_size(2)
902 + ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
903
904}
905
906static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
907{
908 struct ip_tunnel_info *info_a = lwt_tun_info(a);
909 struct ip_tunnel_info *info_b = lwt_tun_info(b);
910
911 return memcmp(info_a, info_b, sizeof(info_a->key)) ||
912 info_a->mode != info_b->mode ||
913 info_a->options_len != info_b->options_len ||
914 memcmp(ip_tunnel_info_opts(info_a),
915 ip_tunnel_info_opts(info_b), info_a->options_len);
916}
917
918static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
919 .build_state = ip_tun_build_state,
920 .destroy_state = ip_tun_destroy_state,
921 .fill_encap = ip_tun_fill_encap_info,
922 .get_encap_size = ip_tun_encap_nlsize,
923 .cmp_encap = ip_tun_cmp_encap,
924 .owner = THIS_MODULE,
925};
926
927static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
928 [LWTUNNEL_IP6_UNSPEC] = { .strict_start_type = LWTUNNEL_IP6_OPTS },
929 [LWTUNNEL_IP6_ID] = { .type = NLA_U64 },
930 [LWTUNNEL_IP6_DST] = { .len = sizeof(struct in6_addr) },
931 [LWTUNNEL_IP6_SRC] = { .len = sizeof(struct in6_addr) },
932 [LWTUNNEL_IP6_HOPLIMIT] = { .type = NLA_U8 },
933 [LWTUNNEL_IP6_TC] = { .type = NLA_U8 },
934 [LWTUNNEL_IP6_FLAGS] = { .type = NLA_U16 },
935 [LWTUNNEL_IP6_OPTS] = { .type = NLA_NESTED },
936};
937
938static int ip6_tun_build_state(struct net *net, struct nlattr *attr,
939 unsigned int family, const void *cfg,
940 struct lwtunnel_state **ts,
941 struct netlink_ext_ack *extack)
942{
943 struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
944 struct lwtunnel_state *new_state;
945 struct ip_tunnel_info *tun_info;
946 int err, opt_len;
947
948 err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP6_MAX, attr,
949 ip6_tun_policy, extack);
950 if (err < 0)
951 return err;
952
953 opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP6_OPTS], extack);
954 if (opt_len < 0)
955 return opt_len;
956
957 new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
958 if (!new_state)
959 return -ENOMEM;
960
961 new_state->type = LWTUNNEL_ENCAP_IP6;
962
963 tun_info = lwt_tun_info(new_state);
964
965 err = ip_tun_set_opts(tb[LWTUNNEL_IP6_OPTS], tun_info, extack);
966 if (err < 0) {
967 lwtstate_free(new_state);
968 return err;
969 }
970
971 if (tb[LWTUNNEL_IP6_ID])
972 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
973
974 if (tb[LWTUNNEL_IP6_DST])
975 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
976
977 if (tb[LWTUNNEL_IP6_SRC])
978 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
979
980 if (tb[LWTUNNEL_IP6_HOPLIMIT])
981 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
982
983 if (tb[LWTUNNEL_IP6_TC])
984 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
985
986 if (tb[LWTUNNEL_IP6_FLAGS])
987 tun_info->key.tun_flags |=
988 (nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]) &
989 ~TUNNEL_OPTIONS_PRESENT);
990
991 tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
992 tun_info->options_len = opt_len;
993
994 *ts = new_state;
995
996 return 0;
997}
998
999static int ip6_tun_fill_encap_info(struct sk_buff *skb,
1000 struct lwtunnel_state *lwtstate)
1001{
1002 struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
1003
1004 if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
1005 LWTUNNEL_IP6_PAD) ||
1006 nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
1007 nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
1008 nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
1009 nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
1010 nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags) ||
1011 ip_tun_fill_encap_opts(skb, LWTUNNEL_IP6_OPTS, tun_info))
1012 return -ENOMEM;
1013
1014 return 0;
1015}
1016
1017static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
1018{
1019 return nla_total_size_64bit(8)
1020 + nla_total_size(16)
1021 + nla_total_size(16)
1022 + nla_total_size(1)
1023 + nla_total_size(1)
1024 + nla_total_size(2)
1025 + ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
1026
1027}
1028
1029static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
1030 .build_state = ip6_tun_build_state,
1031 .fill_encap = ip6_tun_fill_encap_info,
1032 .get_encap_size = ip6_tun_encap_nlsize,
1033 .cmp_encap = ip_tun_cmp_encap,
1034 .owner = THIS_MODULE,
1035};
1036
1037void __init ip_tunnel_core_init(void)
1038{
1039
1040
1041
1042
1043 BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
1044
1045 lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
1046 lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
1047}
1048
1049DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
1050EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
1051
1052void ip_tunnel_need_metadata(void)
1053{
1054 static_branch_inc(&ip_tunnel_metadata_cnt);
1055}
1056EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
1057
1058void ip_tunnel_unneed_metadata(void)
1059{
1060 static_branch_dec(&ip_tunnel_metadata_cnt);
1061}
1062EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);
1063
1064
1065__be16 ip_tunnel_parse_protocol(const struct sk_buff *skb)
1066{
1067 if (skb_network_header(skb) >= skb->head &&
1068 (skb_network_header(skb) + sizeof(struct iphdr)) <= skb_tail_pointer(skb) &&
1069 ip_hdr(skb)->version == 4)
1070 return htons(ETH_P_IP);
1071 if (skb_network_header(skb) >= skb->head &&
1072 (skb_network_header(skb) + sizeof(struct ipv6hdr)) <= skb_tail_pointer(skb) &&
1073 ipv6_hdr(skb)->version == 6)
1074 return htons(ETH_P_IPV6);
1075 return 0;
1076}
1077EXPORT_SYMBOL(ip_tunnel_parse_protocol);
1078
1079const struct header_ops ip_tunnel_header_ops = { .parse_protocol = ip_tunnel_parse_protocol };
1080EXPORT_SYMBOL(ip_tunnel_header_ops);
1081
