]> asedeno.scripts.mit.edu Git - linux.git/blob - net/ipv4/ip_tunnel_core.c
Merge tag 'wireless-drivers-next-2019-11-15' of git://git.kernel.org/pub/scm/linux...
[linux.git] / net / ipv4 / ip_tunnel_core.c
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (c) 2013 Nicira, Inc.
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/arp.h>
29 #include <net/checksum.h>
30 #include <net/dsfield.h>
31 #include <net/inet_ecn.h>
32 #include <net/xfrm.h>
33 #include <net/net_namespace.h>
34 #include <net/netns/generic.h>
35 #include <net/rtnetlink.h>
36 #include <net/dst_metadata.h>
37 #include <net/geneve.h>
38 #include <net/vxlan.h>
39 #include <net/erspan.h>
40
41 const struct ip_tunnel_encap_ops __rcu *
42                 iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
43 EXPORT_SYMBOL(iptun_encaps);
44
45 const struct ip6_tnl_encap_ops __rcu *
46                 ip6tun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
47 EXPORT_SYMBOL(ip6tun_encaps);
48
49 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
50                    __be32 src, __be32 dst, __u8 proto,
51                    __u8 tos, __u8 ttl, __be16 df, bool xnet)
52 {
53         int pkt_len = skb->len - skb_inner_network_offset(skb);
54         struct net *net = dev_net(rt->dst.dev);
55         struct net_device *dev = skb->dev;
56         struct iphdr *iph;
57         int err;
58
59         skb_scrub_packet(skb, xnet);
60
61         skb_clear_hash_if_not_l4(skb);
62         skb_dst_set(skb, &rt->dst);
63         memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
64
65         /* Push down and install the IP header. */
66         skb_push(skb, sizeof(struct iphdr));
67         skb_reset_network_header(skb);
68
69         iph = ip_hdr(skb);
70
71         iph->version    =       4;
72         iph->ihl        =       sizeof(struct iphdr) >> 2;
73         iph->frag_off   =       ip_mtu_locked(&rt->dst) ? 0 : df;
74         iph->protocol   =       proto;
75         iph->tos        =       tos;
76         iph->daddr      =       dst;
77         iph->saddr      =       src;
78         iph->ttl        =       ttl;
79         __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
80
81         err = ip_local_out(net, sk, skb);
82
83         if (dev) {
84                 if (unlikely(net_xmit_eval(err)))
85                         pkt_len = 0;
86                 iptunnel_xmit_stats(dev, pkt_len);
87         }
88 }
89 EXPORT_SYMBOL_GPL(iptunnel_xmit);
90
91 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
92                            __be16 inner_proto, bool raw_proto, bool xnet)
93 {
94         if (unlikely(!pskb_may_pull(skb, hdr_len)))
95                 return -ENOMEM;
96
97         skb_pull_rcsum(skb, hdr_len);
98
99         if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
100                 struct ethhdr *eh;
101
102                 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
103                         return -ENOMEM;
104
105                 eh = (struct ethhdr *)skb->data;
106                 if (likely(eth_proto_is_802_3(eh->h_proto)))
107                         skb->protocol = eh->h_proto;
108                 else
109                         skb->protocol = htons(ETH_P_802_2);
110
111         } else {
112                 skb->protocol = inner_proto;
113         }
114
115         skb_clear_hash_if_not_l4(skb);
116         __vlan_hwaccel_clear_tag(skb);
117         skb_set_queue_mapping(skb, 0);
118         skb_scrub_packet(skb, xnet);
119
120         return iptunnel_pull_offloads(skb);
121 }
122 EXPORT_SYMBOL_GPL(__iptunnel_pull_header);
123
124 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
125                                              gfp_t flags)
126 {
127         struct metadata_dst *res;
128         struct ip_tunnel_info *dst, *src;
129
130         if (!md || md->type != METADATA_IP_TUNNEL ||
131             md->u.tun_info.mode & IP_TUNNEL_INFO_TX)
132                 return NULL;
133
134         src = &md->u.tun_info;
135         res = metadata_dst_alloc(src->options_len, METADATA_IP_TUNNEL, flags);
136         if (!res)
137                 return NULL;
138
139         dst = &res->u.tun_info;
140         dst->key.tun_id = src->key.tun_id;
141         if (src->mode & IP_TUNNEL_INFO_IPV6)
142                 memcpy(&dst->key.u.ipv6.dst, &src->key.u.ipv6.src,
143                        sizeof(struct in6_addr));
144         else
145                 dst->key.u.ipv4.dst = src->key.u.ipv4.src;
146         dst->key.tun_flags = src->key.tun_flags;
147         dst->mode = src->mode | IP_TUNNEL_INFO_TX;
148         ip_tunnel_info_opts_set(dst, ip_tunnel_info_opts(src),
149                                 src->options_len, 0);
150
151         return res;
152 }
153 EXPORT_SYMBOL_GPL(iptunnel_metadata_reply);
154
155 int iptunnel_handle_offloads(struct sk_buff *skb,
156                              int gso_type_mask)
157 {
158         int err;
159
160         if (likely(!skb->encapsulation)) {
161                 skb_reset_inner_headers(skb);
162                 skb->encapsulation = 1;
163         }
164
165         if (skb_is_gso(skb)) {
166                 err = skb_header_unclone(skb, GFP_ATOMIC);
167                 if (unlikely(err))
168                         return err;
169                 skb_shinfo(skb)->gso_type |= gso_type_mask;
170                 return 0;
171         }
172
173         if (skb->ip_summed != CHECKSUM_PARTIAL) {
174                 skb->ip_summed = CHECKSUM_NONE;
175                 /* We clear encapsulation here to prevent badly-written
176                  * drivers potentially deciding to offload an inner checksum
177                  * if we set CHECKSUM_PARTIAL on the outer header.
178                  * This should go away when the drivers are all fixed.
179                  */
180                 skb->encapsulation = 0;
181         }
182
183         return 0;
184 }
185 EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
186
187 /* Often modified stats are per cpu, other are shared (netdev->stats) */
188 void ip_tunnel_get_stats64(struct net_device *dev,
189                            struct rtnl_link_stats64 *tot)
190 {
191         int i;
192
193         netdev_stats_to_stats64(tot, &dev->stats);
194
195         for_each_possible_cpu(i) {
196                 const struct pcpu_sw_netstats *tstats =
197                                                    per_cpu_ptr(dev->tstats, i);
198                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
199                 unsigned int start;
200
201                 do {
202                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
203                         rx_packets = tstats->rx_packets;
204                         tx_packets = tstats->tx_packets;
205                         rx_bytes = tstats->rx_bytes;
206                         tx_bytes = tstats->tx_bytes;
207                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
208
209                 tot->rx_packets += rx_packets;
210                 tot->tx_packets += tx_packets;
211                 tot->rx_bytes   += rx_bytes;
212                 tot->tx_bytes   += tx_bytes;
213         }
214 }
215 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
216
217 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
218         [LWTUNNEL_IP_ID]        = { .type = NLA_U64 },
219         [LWTUNNEL_IP_DST]       = { .type = NLA_U32 },
220         [LWTUNNEL_IP_SRC]       = { .type = NLA_U32 },
221         [LWTUNNEL_IP_TTL]       = { .type = NLA_U8 },
222         [LWTUNNEL_IP_TOS]       = { .type = NLA_U8 },
223         [LWTUNNEL_IP_FLAGS]     = { .type = NLA_U16 },
224         [LWTUNNEL_IP_OPTS]      = { .type = NLA_NESTED },
225 };
226
227 static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = {
228         [LWTUNNEL_IP_OPTS_GENEVE]       = { .type = NLA_NESTED },
229         [LWTUNNEL_IP_OPTS_VXLAN]        = { .type = NLA_NESTED },
230         [LWTUNNEL_IP_OPTS_ERSPAN]       = { .type = NLA_NESTED },
231 };
232
233 static const struct nla_policy
234 geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = {
235         [LWTUNNEL_IP_OPT_GENEVE_CLASS]  = { .type = NLA_U16 },
236         [LWTUNNEL_IP_OPT_GENEVE_TYPE]   = { .type = NLA_U8 },
237         [LWTUNNEL_IP_OPT_GENEVE_DATA]   = { .type = NLA_BINARY, .len = 128 },
238 };
239
240 static const struct nla_policy
241 vxlan_opt_policy[LWTUNNEL_IP_OPT_VXLAN_MAX + 1] = {
242         [LWTUNNEL_IP_OPT_VXLAN_GBP]     = { .type = NLA_U32 },
243 };
244
245 static const struct nla_policy
246 erspan_opt_policy[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1] = {
247         [LWTUNNEL_IP_OPT_ERSPAN_VER]    = { .type = NLA_U8 },
248         [LWTUNNEL_IP_OPT_ERSPAN_INDEX]  = { .type = NLA_U32 },
249         [LWTUNNEL_IP_OPT_ERSPAN_DIR]    = { .type = NLA_U8 },
250         [LWTUNNEL_IP_OPT_ERSPAN_HWID]   = { .type = NLA_U8 },
251 };
252
253 static int ip_tun_parse_opts_geneve(struct nlattr *attr,
254                                     struct ip_tunnel_info *info,
255                                     struct netlink_ext_ack *extack)
256 {
257         struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1];
258         int data_len, err;
259
260         err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_GENEVE_MAX, attr,
261                                geneve_opt_policy, extack);
262         if (err)
263                 return err;
264
265         if (!tb[LWTUNNEL_IP_OPT_GENEVE_CLASS] ||
266             !tb[LWTUNNEL_IP_OPT_GENEVE_TYPE] ||
267             !tb[LWTUNNEL_IP_OPT_GENEVE_DATA])
268                 return -EINVAL;
269
270         attr = tb[LWTUNNEL_IP_OPT_GENEVE_DATA];
271         data_len = nla_len(attr);
272         if (data_len % 4)
273                 return -EINVAL;
274
275         if (info) {
276                 struct geneve_opt *opt = ip_tunnel_info_opts(info);
277
278                 memcpy(opt->opt_data, nla_data(attr), data_len);
279                 opt->length = data_len / 4;
280                 attr = tb[LWTUNNEL_IP_OPT_GENEVE_CLASS];
281                 opt->opt_class = nla_get_be16(attr);
282                 attr = tb[LWTUNNEL_IP_OPT_GENEVE_TYPE];
283                 opt->type = nla_get_u8(attr);
284                 info->key.tun_flags |= TUNNEL_GENEVE_OPT;
285         }
286
287         return sizeof(struct geneve_opt) + data_len;
288 }
289
290 static int ip_tun_parse_opts_vxlan(struct nlattr *attr,
291                                    struct ip_tunnel_info *info,
292                                    struct netlink_ext_ack *extack)
293 {
294         struct nlattr *tb[LWTUNNEL_IP_OPT_VXLAN_MAX + 1];
295         int err;
296
297         err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_VXLAN_MAX, attr,
298                                vxlan_opt_policy, extack);
299         if (err)
300                 return err;
301
302         if (!tb[LWTUNNEL_IP_OPT_VXLAN_GBP])
303                 return -EINVAL;
304
305         if (info) {
306                 struct vxlan_metadata *md = ip_tunnel_info_opts(info);
307
308                 attr = tb[LWTUNNEL_IP_OPT_VXLAN_GBP];
309                 md->gbp = nla_get_u32(attr);
310                 info->key.tun_flags |= TUNNEL_VXLAN_OPT;
311         }
312
313         return sizeof(struct vxlan_metadata);
314 }
315
316 static int ip_tun_parse_opts_erspan(struct nlattr *attr,
317                                     struct ip_tunnel_info *info,
318                                     struct netlink_ext_ack *extack)
319 {
320         struct nlattr *tb[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1];
321         int err;
322
323         err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_ERSPAN_MAX, attr,
324                                erspan_opt_policy, extack);
325         if (err)
326                 return err;
327
328         if (!tb[LWTUNNEL_IP_OPT_ERSPAN_VER])
329                 return -EINVAL;
330
331         if (info) {
332                 struct erspan_metadata *md = ip_tunnel_info_opts(info);
333
334                 attr = tb[LWTUNNEL_IP_OPT_ERSPAN_VER];
335                 md->version = nla_get_u8(attr);
336
337                 if (md->version == 1 && tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX]) {
338                         attr = tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX];
339                         md->u.index = nla_get_be32(attr);
340                 } else if (md->version == 2 && tb[LWTUNNEL_IP_OPT_ERSPAN_DIR] &&
341                            tb[LWTUNNEL_IP_OPT_ERSPAN_HWID]) {
342                         attr = tb[LWTUNNEL_IP_OPT_ERSPAN_DIR];
343                         md->u.md2.dir = nla_get_u8(attr);
344                         attr = tb[LWTUNNEL_IP_OPT_ERSPAN_HWID];
345                         set_hwid(&md->u.md2, nla_get_u8(attr));
346                 } else {
347                         return -EINVAL;
348                 }
349
350                 info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
351         }
352
353         return sizeof(struct erspan_metadata);
354 }
355
356 static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info,
357                              struct netlink_ext_ack *extack)
358 {
359         struct nlattr *tb[LWTUNNEL_IP_OPTS_MAX + 1];
360         int err;
361
362         if (!attr)
363                 return 0;
364
365         err = nla_parse_nested(tb, LWTUNNEL_IP_OPTS_MAX, attr,
366                                ip_opts_policy, extack);
367         if (err)
368                 return err;
369
370         if (tb[LWTUNNEL_IP_OPTS_GENEVE])
371                 err = ip_tun_parse_opts_geneve(tb[LWTUNNEL_IP_OPTS_GENEVE],
372                                                info, extack);
373         else if (tb[LWTUNNEL_IP_OPTS_VXLAN])
374                 err = ip_tun_parse_opts_vxlan(tb[LWTUNNEL_IP_OPTS_VXLAN],
375                                               info, extack);
376         else if (tb[LWTUNNEL_IP_OPTS_ERSPAN])
377                 err = ip_tun_parse_opts_erspan(tb[LWTUNNEL_IP_OPTS_ERSPAN],
378                                                info, extack);
379         else
380                 err = -EINVAL;
381
382         return err;
383 }
384
385 static int ip_tun_get_optlen(struct nlattr *attr,
386                              struct netlink_ext_ack *extack)
387 {
388         return ip_tun_parse_opts(attr, NULL, extack);
389 }
390
391 static int ip_tun_set_opts(struct nlattr *attr, struct ip_tunnel_info *info,
392                            struct netlink_ext_ack *extack)
393 {
394         return ip_tun_parse_opts(attr, info, extack);
395 }
396
397 static int ip_tun_build_state(struct nlattr *attr,
398                               unsigned int family, const void *cfg,
399                               struct lwtunnel_state **ts,
400                               struct netlink_ext_ack *extack)
401 {
402         struct nlattr *tb[LWTUNNEL_IP_MAX + 1];
403         struct lwtunnel_state *new_state;
404         struct ip_tunnel_info *tun_info;
405         int err, opt_len;
406
407         err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_MAX, attr,
408                                           ip_tun_policy, extack);
409         if (err < 0)
410                 return err;
411
412         opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP_OPTS], extack);
413         if (opt_len < 0)
414                 return opt_len;
415
416         new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
417         if (!new_state)
418                 return -ENOMEM;
419
420         new_state->type = LWTUNNEL_ENCAP_IP;
421
422         tun_info = lwt_tun_info(new_state);
423
424         err = ip_tun_set_opts(tb[LWTUNNEL_IP_OPTS], tun_info, extack);
425         if (err < 0) {
426                 lwtstate_free(new_state);
427                 return err;
428         }
429
430 #ifdef CONFIG_DST_CACHE
431         err = dst_cache_init(&tun_info->dst_cache, GFP_KERNEL);
432         if (err) {
433                 lwtstate_free(new_state);
434                 return err;
435         }
436 #endif
437
438         if (tb[LWTUNNEL_IP_ID])
439                 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP_ID]);
440
441         if (tb[LWTUNNEL_IP_DST])
442                 tun_info->key.u.ipv4.dst = nla_get_in_addr(tb[LWTUNNEL_IP_DST]);
443
444         if (tb[LWTUNNEL_IP_SRC])
445                 tun_info->key.u.ipv4.src = nla_get_in_addr(tb[LWTUNNEL_IP_SRC]);
446
447         if (tb[LWTUNNEL_IP_TTL])
448                 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP_TTL]);
449
450         if (tb[LWTUNNEL_IP_TOS])
451                 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
452
453         if (tb[LWTUNNEL_IP_FLAGS])
454                 tun_info->key.tun_flags |=
455                                 (nla_get_be16(tb[LWTUNNEL_IP_FLAGS]) &
456                                  ~TUNNEL_OPTIONS_PRESENT);
457
458         tun_info->mode = IP_TUNNEL_INFO_TX;
459         tun_info->options_len = opt_len;
460
461         *ts = new_state;
462
463         return 0;
464 }
465
466 static void ip_tun_destroy_state(struct lwtunnel_state *lwtstate)
467 {
468 #ifdef CONFIG_DST_CACHE
469         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
470
471         dst_cache_destroy(&tun_info->dst_cache);
472 #endif
473 }
474
475 static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
476                                          struct ip_tunnel_info *tun_info)
477 {
478         struct geneve_opt *opt;
479         struct nlattr *nest;
480
481         nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE);
482         if (!nest)
483                 return -ENOMEM;
484
485         opt = ip_tunnel_info_opts(tun_info);
486         if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS, opt->opt_class) ||
487             nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
488             nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
489                     opt->opt_data)) {
490                 nla_nest_cancel(skb, nest);
491                 return -ENOMEM;
492         }
493
494         nla_nest_end(skb, nest);
495         return 0;
496 }
497
498 static int ip_tun_fill_encap_opts_vxlan(struct sk_buff *skb,
499                                         struct ip_tunnel_info *tun_info)
500 {
501         struct vxlan_metadata *md;
502         struct nlattr *nest;
503
504         nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_VXLAN);
505         if (!nest)
506                 return -ENOMEM;
507
508         md = ip_tunnel_info_opts(tun_info);
509         if (nla_put_u32(skb, LWTUNNEL_IP_OPT_VXLAN_GBP, md->gbp)) {
510                 nla_nest_cancel(skb, nest);
511                 return -ENOMEM;
512         }
513
514         nla_nest_end(skb, nest);
515         return 0;
516 }
517
518 static int ip_tun_fill_encap_opts_erspan(struct sk_buff *skb,
519                                          struct ip_tunnel_info *tun_info)
520 {
521         struct erspan_metadata *md;
522         struct nlattr *nest;
523
524         nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_ERSPAN);
525         if (!nest)
526                 return -ENOMEM;
527
528         md = ip_tunnel_info_opts(tun_info);
529         if (nla_put_u32(skb, LWTUNNEL_IP_OPT_ERSPAN_VER, md->version))
530                 goto err;
531
532         if (md->version == 1 &&
533             nla_put_be32(skb, LWTUNNEL_IP_OPT_ERSPAN_INDEX, md->u.index))
534                 goto err;
535
536         if (md->version == 2 &&
537             (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_DIR, md->u.md2.dir) ||
538              nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_HWID,
539                         get_hwid(&md->u.md2))))
540                 goto err;
541
542         nla_nest_end(skb, nest);
543         return 0;
544 err:
545         nla_nest_cancel(skb, nest);
546         return -ENOMEM;
547 }
548
549 static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
550                                   struct ip_tunnel_info *tun_info)
551 {
552         struct nlattr *nest;
553         int err = 0;
554
555         if (!(tun_info->key.tun_flags & TUNNEL_OPTIONS_PRESENT))
556                 return 0;
557
558         nest = nla_nest_start_noflag(skb, type);
559         if (!nest)
560                 return -ENOMEM;
561
562         if (tun_info->key.tun_flags & TUNNEL_GENEVE_OPT)
563                 err = ip_tun_fill_encap_opts_geneve(skb, tun_info);
564         else if (tun_info->key.tun_flags & TUNNEL_VXLAN_OPT)
565                 err = ip_tun_fill_encap_opts_vxlan(skb, tun_info);
566         else if (tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)
567                 err = ip_tun_fill_encap_opts_erspan(skb, tun_info);
568
569         if (err) {
570                 nla_nest_cancel(skb, nest);
571                 return err;
572         }
573
574         nla_nest_end(skb, nest);
575         return 0;
576 }
577
578 static int ip_tun_fill_encap_info(struct sk_buff *skb,
579                                   struct lwtunnel_state *lwtstate)
580 {
581         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
582
583         if (nla_put_be64(skb, LWTUNNEL_IP_ID, tun_info->key.tun_id,
584                          LWTUNNEL_IP_PAD) ||
585             nla_put_in_addr(skb, LWTUNNEL_IP_DST, tun_info->key.u.ipv4.dst) ||
586             nla_put_in_addr(skb, LWTUNNEL_IP_SRC, tun_info->key.u.ipv4.src) ||
587             nla_put_u8(skb, LWTUNNEL_IP_TOS, tun_info->key.tos) ||
588             nla_put_u8(skb, LWTUNNEL_IP_TTL, tun_info->key.ttl) ||
589             nla_put_be16(skb, LWTUNNEL_IP_FLAGS, tun_info->key.tun_flags) ||
590             ip_tun_fill_encap_opts(skb, LWTUNNEL_IP_OPTS, tun_info))
591                 return -ENOMEM;
592
593         return 0;
594 }
595
596 static int ip_tun_opts_nlsize(struct ip_tunnel_info *info)
597 {
598         int opt_len;
599
600         if (!(info->key.tun_flags & TUNNEL_OPTIONS_PRESENT))
601                 return 0;
602
603         opt_len = nla_total_size(0);            /* LWTUNNEL_IP_OPTS */
604         if (info->key.tun_flags & TUNNEL_GENEVE_OPT) {
605                 struct geneve_opt *opt = ip_tunnel_info_opts(info);
606
607                 opt_len += nla_total_size(0)    /* LWTUNNEL_IP_OPTS_GENEVE */
608                            + nla_total_size(2)  /* OPT_GENEVE_CLASS */
609                            + nla_total_size(1)  /* OPT_GENEVE_TYPE */
610                            + nla_total_size(opt->length * 4);
611                                                 /* OPT_GENEVE_DATA */
612         } else if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
613                 opt_len += nla_total_size(0)    /* LWTUNNEL_IP_OPTS_VXLAN */
614                            + nla_total_size(4); /* OPT_VXLAN_GBP */
615         } else if (info->key.tun_flags & TUNNEL_ERSPAN_OPT) {
616                 struct erspan_metadata *md = ip_tunnel_info_opts(info);
617
618                 opt_len += nla_total_size(0)    /* LWTUNNEL_IP_OPTS_ERSPAN */
619                            + nla_total_size(1)  /* OPT_ERSPAN_VER */
620                            + (md->version == 1 ? nla_total_size(4)
621                                                 /* OPT_ERSPAN_INDEX (v1) */
622                                                : nla_total_size(1) +
623                                                  nla_total_size(1));
624                                                 /* OPT_ERSPAN_DIR + HWID (v2) */
625         }
626
627         return opt_len;
628 }
629
630 static int ip_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
631 {
632         return nla_total_size_64bit(8)  /* LWTUNNEL_IP_ID */
633                 + nla_total_size(4)     /* LWTUNNEL_IP_DST */
634                 + nla_total_size(4)     /* LWTUNNEL_IP_SRC */
635                 + nla_total_size(1)     /* LWTUNNEL_IP_TOS */
636                 + nla_total_size(1)     /* LWTUNNEL_IP_TTL */
637                 + nla_total_size(2)     /* LWTUNNEL_IP_FLAGS */
638                 + ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
639                                         /* LWTUNNEL_IP_OPTS */
640 }
641
642 static int ip_tun_cmp_encap(struct lwtunnel_state *a, struct lwtunnel_state *b)
643 {
644         struct ip_tunnel_info *info_a = lwt_tun_info(a);
645         struct ip_tunnel_info *info_b = lwt_tun_info(b);
646
647         return memcmp(info_a, info_b, sizeof(info_a->key)) ||
648                info_a->mode != info_b->mode ||
649                info_a->options_len != info_b->options_len ||
650                memcmp(ip_tunnel_info_opts(info_a),
651                       ip_tunnel_info_opts(info_b), info_a->options_len);
652 }
653
654 static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
655         .build_state = ip_tun_build_state,
656         .destroy_state = ip_tun_destroy_state,
657         .fill_encap = ip_tun_fill_encap_info,
658         .get_encap_size = ip_tun_encap_nlsize,
659         .cmp_encap = ip_tun_cmp_encap,
660         .owner = THIS_MODULE,
661 };
662
663 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
664         [LWTUNNEL_IP6_ID]               = { .type = NLA_U64 },
665         [LWTUNNEL_IP6_DST]              = { .len = sizeof(struct in6_addr) },
666         [LWTUNNEL_IP6_SRC]              = { .len = sizeof(struct in6_addr) },
667         [LWTUNNEL_IP6_HOPLIMIT]         = { .type = NLA_U8 },
668         [LWTUNNEL_IP6_TC]               = { .type = NLA_U8 },
669         [LWTUNNEL_IP6_FLAGS]            = { .type = NLA_U16 },
670 };
671
672 static int ip6_tun_build_state(struct nlattr *attr,
673                                unsigned int family, const void *cfg,
674                                struct lwtunnel_state **ts,
675                                struct netlink_ext_ack *extack)
676 {
677         struct nlattr *tb[LWTUNNEL_IP6_MAX + 1];
678         struct lwtunnel_state *new_state;
679         struct ip_tunnel_info *tun_info;
680         int err, opt_len;
681
682         err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP6_MAX, attr,
683                                           ip6_tun_policy, extack);
684         if (err < 0)
685                 return err;
686
687         opt_len = ip_tun_get_optlen(tb[LWTUNNEL_IP6_OPTS], extack);
688         if (opt_len < 0)
689                 return opt_len;
690
691         new_state = lwtunnel_state_alloc(sizeof(*tun_info) + opt_len);
692         if (!new_state)
693                 return -ENOMEM;
694
695         new_state->type = LWTUNNEL_ENCAP_IP6;
696
697         tun_info = lwt_tun_info(new_state);
698
699         err = ip_tun_set_opts(tb[LWTUNNEL_IP6_OPTS], tun_info, extack);
700         if (err < 0) {
701                 lwtstate_free(new_state);
702                 return err;
703         }
704
705         if (tb[LWTUNNEL_IP6_ID])
706                 tun_info->key.tun_id = nla_get_be64(tb[LWTUNNEL_IP6_ID]);
707
708         if (tb[LWTUNNEL_IP6_DST])
709                 tun_info->key.u.ipv6.dst = nla_get_in6_addr(tb[LWTUNNEL_IP6_DST]);
710
711         if (tb[LWTUNNEL_IP6_SRC])
712                 tun_info->key.u.ipv6.src = nla_get_in6_addr(tb[LWTUNNEL_IP6_SRC]);
713
714         if (tb[LWTUNNEL_IP6_HOPLIMIT])
715                 tun_info->key.ttl = nla_get_u8(tb[LWTUNNEL_IP6_HOPLIMIT]);
716
717         if (tb[LWTUNNEL_IP6_TC])
718                 tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
719
720         if (tb[LWTUNNEL_IP6_FLAGS])
721                 tun_info->key.tun_flags |=
722                                 (nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]) &
723                                  ~TUNNEL_OPTIONS_PRESENT);
724
725         tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
726         tun_info->options_len = opt_len;
727
728         *ts = new_state;
729
730         return 0;
731 }
732
733 static int ip6_tun_fill_encap_info(struct sk_buff *skb,
734                                    struct lwtunnel_state *lwtstate)
735 {
736         struct ip_tunnel_info *tun_info = lwt_tun_info(lwtstate);
737
738         if (nla_put_be64(skb, LWTUNNEL_IP6_ID, tun_info->key.tun_id,
739                          LWTUNNEL_IP6_PAD) ||
740             nla_put_in6_addr(skb, LWTUNNEL_IP6_DST, &tun_info->key.u.ipv6.dst) ||
741             nla_put_in6_addr(skb, LWTUNNEL_IP6_SRC, &tun_info->key.u.ipv6.src) ||
742             nla_put_u8(skb, LWTUNNEL_IP6_TC, tun_info->key.tos) ||
743             nla_put_u8(skb, LWTUNNEL_IP6_HOPLIMIT, tun_info->key.ttl) ||
744             nla_put_be16(skb, LWTUNNEL_IP6_FLAGS, tun_info->key.tun_flags) ||
745             ip_tun_fill_encap_opts(skb, LWTUNNEL_IP6_OPTS, tun_info))
746                 return -ENOMEM;
747
748         return 0;
749 }
750
751 static int ip6_tun_encap_nlsize(struct lwtunnel_state *lwtstate)
752 {
753         return nla_total_size_64bit(8)  /* LWTUNNEL_IP6_ID */
754                 + nla_total_size(16)    /* LWTUNNEL_IP6_DST */
755                 + nla_total_size(16)    /* LWTUNNEL_IP6_SRC */
756                 + nla_total_size(1)     /* LWTUNNEL_IP6_HOPLIMIT */
757                 + nla_total_size(1)     /* LWTUNNEL_IP6_TC */
758                 + nla_total_size(2)     /* LWTUNNEL_IP6_FLAGS */
759                 + ip_tun_opts_nlsize(lwt_tun_info(lwtstate));
760                                         /* LWTUNNEL_IP6_OPTS */
761 }
762
763 static const struct lwtunnel_encap_ops ip6_tun_lwt_ops = {
764         .build_state = ip6_tun_build_state,
765         .fill_encap = ip6_tun_fill_encap_info,
766         .get_encap_size = ip6_tun_encap_nlsize,
767         .cmp_encap = ip_tun_cmp_encap,
768         .owner = THIS_MODULE,
769 };
770
771 void __init ip_tunnel_core_init(void)
772 {
773         /* If you land here, make sure whether increasing ip_tunnel_info's
774          * options_len is a reasonable choice with its usage in front ends
775          * (f.e., it's part of flow keys, etc).
776          */
777         BUILD_BUG_ON(IP_TUNNEL_OPTS_MAX != 255);
778
779         lwtunnel_encap_add_ops(&ip_tun_lwt_ops, LWTUNNEL_ENCAP_IP);
780         lwtunnel_encap_add_ops(&ip6_tun_lwt_ops, LWTUNNEL_ENCAP_IP6);
781 }
782
783 DEFINE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
784 EXPORT_SYMBOL(ip_tunnel_metadata_cnt);
785
786 void ip_tunnel_need_metadata(void)
787 {
788         static_branch_inc(&ip_tunnel_metadata_cnt);
789 }
790 EXPORT_SYMBOL_GPL(ip_tunnel_need_metadata);
791
792 void ip_tunnel_unneed_metadata(void)
793 {
794         static_branch_dec(&ip_tunnel_metadata_cnt);
795 }
796 EXPORT_SYMBOL_GPL(ip_tunnel_unneed_metadata);