1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __NET_IP_TUNNELS_H
3 #define __NET_IP_TUNNELS_H 1
4
5 #include <linux/if_tunnel.h>
6 #include <linux/netdevice.h>
7 #include <linux/skbuff.h>
8 #include <linux/socket.h>
9 #include <linux/types.h>
10 #include <linux/u64_stats_sync.h>
11 #include <linux/bitops.h>
12
13 #include <net/dsfield.h>
14 #include <net/gro_cells.h>
15 #include <net/inet_ecn.h>
16 #include <net/netns/generic.h>
17 #include <net/rtnetlink.h>
18 #include <net/lwtunnel.h>
19 #include <net/dst_cache.h>
20
21 #if IS_ENABLED(CONFIG_IPV6)
22 #include <net/ipv6.h>
23 #include <net/ip6_fib.h>
24 #include <net/ip6_route.h>
25 #endif
26
27 /* Keep error state on tunnel for 30 sec */
28 #define IPTUNNEL_ERR_TIMEO (30*HZ)
29
30 /* Used to memset ip_tunnel padding. */
31 #define IP_TUNNEL_KEY_SIZE offsetofend(struct ip_tunnel_key, tp_dst)
32
33 /* Used to memset ipv4 address padding. */
34 #define IP_TUNNEL_KEY_IPV4_PAD offsetofend(struct ip_tunnel_key, u.ipv4.dst)
35 #define IP_TUNNEL_KEY_IPV4_PAD_LEN \
36 (FIELD_SIZEOF(struct ip_tunnel_key, u) - \
37 FIELD_SIZEOF(struct ip_tunnel_key, u.ipv4))
38
39 struct ip_tunnel_key {
40 __be64 tun_id;
41 union {
42 struct {
43 __be32 src;
44 __be32 dst;
45 } ipv4;
46 struct {
47 struct in6_addr src;
48 struct in6_addr dst;
49 } ipv6;
50 } u;
51 __be16 tun_flags;
52 u8 tos; /* TOS for IPv4, TC for IPv6 */
53 u8 ttl; /* TTL for IPv4, HL for IPv6 */
54 __be32 label; /* Flow Label for IPv6 */
55 __be16 tp_src;
56 __be16 tp_dst;
57 };
58
59 /* Flags for ip_tunnel_info mode. */
60 #define IP_TUNNEL_INFO_TX 0x01 /* represents tx tunnel parameters */
61 #define IP_TUNNEL_INFO_IPV6 0x02 /* key contains IPv6 addresses */
62 #define IP_TUNNEL_INFO_BRIDGE 0x04 /* represents a bridged tunnel id */
63
64 /* Maximum tunnel options length. */
65 #define IP_TUNNEL_OPTS_MAX \
66 GENMASK((FIELD_SIZEOF(struct ip_tunnel_info, \
67 options_len) * BITS_PER_BYTE) - 1, 0)
68
69 struct ip_tunnel_info {
70 struct ip_tunnel_key key;
71 #ifdef CONFIG_DST_CACHE
72 struct dst_cache dst_cache;
73 #endif
74 u8 options_len;
75 u8 mode;
76 };
77
78 /* 6rd prefix/relay information */
79 #ifdef CONFIG_IPV6_SIT_6RD
80 struct ip_tunnel_6rd_parm {
81 struct in6_addr prefix;
82 __be32 relay_prefix;
83 u16 prefixlen;
84 u16 relay_prefixlen;
85 };
86 #endif
87
88 struct ip_tunnel_encap {
89 u16 type;
90 u16 flags;
91 __be16 sport;
92 __be16 dport;
93 };
94
95 struct ip_tunnel_prl_entry {
96 struct ip_tunnel_prl_entry __rcu *next;
97 __be32 addr;
98 u16 flags;
99 struct rcu_head rcu_head;
100 };
101
102 struct metadata_dst;
103
104 struct ip_tunnel {
105 struct ip_tunnel __rcu *next;
106 struct hlist_node hash_node;
107 struct net_device *dev;
108 struct net *net; /* netns for packet i/o */
109
110 unsigned long err_time; /* Time when the last ICMP error
111 * arrived */
112 int err_count; /* Number of arrived ICMP errors */
113
114 /* These four fields used only by GRE */
115 u32 i_seqno; /* The last seen seqno */
116 u32 o_seqno; /* The last output seqno */
117 int tun_hlen; /* Precalculated header length */
118
119 /* This field used only by ERSPAN */
120 u32 index; /* ERSPAN type II index */
121
122 struct dst_cache dst_cache;
123
124 struct ip_tunnel_parm parms;
125
126 int mlink;
127 int encap_hlen; /* Encap header length (FOU,GUE) */
128 int hlen; /* tun_hlen + encap_hlen */
129 struct ip_tunnel_encap encap;
130
131 /* for SIT */
132 #ifdef CONFIG_IPV6_SIT_6RD
133 struct ip_tunnel_6rd_parm ip6rd;
134 #endif
135 struct ip_tunnel_prl_entry __rcu *prl; /* potential router list */
136 unsigned int prl_count; /* # of entries in PRL */
137 unsigned int ip_tnl_net_id;
138 struct gro_cells gro_cells;
139 __u32 fwmark;
140 bool collect_md;
141 bool ignore_df;
142 };
143
144 #define TUNNEL_CSUM __cpu_to_be16(0x01)
145 #define TUNNEL_ROUTING __cpu_to_be16(0x02)
146 #define TUNNEL_KEY __cpu_to_be16(0x04)
147 #define TUNNEL_SEQ __cpu_to_be16(0x08)
148 #define TUNNEL_STRICT __cpu_to_be16(0x10)
149 #define TUNNEL_REC __cpu_to_be16(0x20)
150 #define TUNNEL_VERSION __cpu_to_be16(0x40)
151 #define TUNNEL_NO_KEY __cpu_to_be16(0x80)
152 #define TUNNEL_DONT_FRAGMENT __cpu_to_be16(0x0100)
153 #define TUNNEL_OAM __cpu_to_be16(0x0200)
154 #define TUNNEL_CRIT_OPT __cpu_to_be16(0x0400)
155 #define TUNNEL_GENEVE_OPT __cpu_to_be16(0x0800)
156 #define TUNNEL_VXLAN_OPT __cpu_to_be16(0x1000)
157 #define TUNNEL_NOCACHE __cpu_to_be16(0x2000)
158 #define TUNNEL_ERSPAN_OPT __cpu_to_be16(0x4000)
159
160 #define TUNNEL_OPTIONS_PRESENT \
161 (TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT | TUNNEL_ERSPAN_OPT)
162
163 struct tnl_ptk_info {
164 __be16 flags;
165 __be16 proto;
166 __be32 key;
167 __be32 seq;
168 int hdr_len;
169 };
170
171 #define PACKET_RCVD 0
172 #define PACKET_REJECT 1
173 #define PACKET_NEXT 2
174
175 #define IP_TNL_HASH_BITS 7
176 #define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS)
177
178 struct ip_tunnel_net {
179 struct net_device *fb_tunnel_dev;
180 struct hlist_head tunnels[IP_TNL_HASH_SIZE];
181 struct ip_tunnel __rcu *collect_md_tun;
182 };
183
ip_tunnel_key_init(struct ip_tunnel_key * key,__be32 saddr,__be32 daddr,u8 tos,u8 ttl,__be32 label,__be16 tp_src,__be16 tp_dst,__be64 tun_id,__be16 tun_flags)184 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
185 __be32 saddr, __be32 daddr,
186 u8 tos, u8 ttl, __be32 label,
187 __be16 tp_src, __be16 tp_dst,
188 __be64 tun_id, __be16 tun_flags)
189 {
190 key->tun_id = tun_id;
191 key->u.ipv4.src = saddr;
192 key->u.ipv4.dst = daddr;
193 memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
194 0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
195 key->tos = tos;
196 key->ttl = ttl;
197 key->label = label;
198 key->tun_flags = tun_flags;
199
200 /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
201 * the upper tunnel are used.
202 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
203 */
204 key->tp_src = tp_src;
205 key->tp_dst = tp_dst;
206
207 /* Clear struct padding. */
208 if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
209 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
210 0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
211 }
212
213 static inline bool
ip_tunnel_dst_cache_usable(const struct sk_buff * skb,const struct ip_tunnel_info * info)214 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
215 const struct ip_tunnel_info *info)
216 {
217 if (skb->mark)
218 return false;
219 if (!info)
220 return true;
221 if (info->key.tun_flags & TUNNEL_NOCACHE)
222 return false;
223
224 return true;
225 }
226
ip_tunnel_info_af(const struct ip_tunnel_info * tun_info)227 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
228 *tun_info)
229 {
230 return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
231 }
232
key32_to_tunnel_id(__be32 key)233 static inline __be64 key32_to_tunnel_id(__be32 key)
234 {
235 #ifdef __BIG_ENDIAN
236 return (__force __be64)key;
237 #else
238 return (__force __be64)((__force u64)key << 32);
239 #endif
240 }
241
242 /* Returns the least-significant 32 bits of a __be64. */
tunnel_id_to_key32(__be64 tun_id)243 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
244 {
245 #ifdef __BIG_ENDIAN
246 return (__force __be32)tun_id;
247 #else
248 return (__force __be32)((__force u64)tun_id >> 32);
249 #endif
250 }
251
252 #ifdef CONFIG_INET
253
254 int ip_tunnel_init(struct net_device *dev);
255 void ip_tunnel_uninit(struct net_device *dev);
256 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
257 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
258 int ip_tunnel_get_iflink(const struct net_device *dev);
259 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
260 struct rtnl_link_ops *ops, char *devname);
261
262 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops);
263
264 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
265 const struct iphdr *tnl_params, const u8 protocol);
266 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
267 const u8 proto);
268 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
269 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
270 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
271
272 void ip_tunnel_get_stats64(struct net_device *dev,
273 struct rtnl_link_stats64 *tot);
274 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
275 int link, __be16 flags,
276 __be32 remote, __be32 local,
277 __be32 key);
278
279 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
280 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
281 bool log_ecn_error);
282 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
283 struct ip_tunnel_parm *p, __u32 fwmark);
284 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
285 struct ip_tunnel_parm *p, __u32 fwmark);
286 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
287
288 struct ip_tunnel_encap_ops {
289 size_t (*encap_hlen)(struct ip_tunnel_encap *e);
290 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
291 u8 *protocol, struct flowi4 *fl4);
292 };
293
294 #define MAX_IPTUN_ENCAP_OPS 8
295
296 extern const struct ip_tunnel_encap_ops __rcu *
297 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
298
299 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
300 unsigned int num);
301 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
302 unsigned int num);
303
304 int ip_tunnel_encap_setup(struct ip_tunnel *t,
305 struct ip_tunnel_encap *ipencap);
306
ip_encap_hlen(struct ip_tunnel_encap * e)307 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
308 {
309 const struct ip_tunnel_encap_ops *ops;
310 int hlen = -EINVAL;
311
312 if (e->type == TUNNEL_ENCAP_NONE)
313 return 0;
314
315 if (e->type >= MAX_IPTUN_ENCAP_OPS)
316 return -EINVAL;
317
318 rcu_read_lock();
319 ops = rcu_dereference(iptun_encaps[e->type]);
320 if (likely(ops && ops->encap_hlen))
321 hlen = ops->encap_hlen(e);
322 rcu_read_unlock();
323
324 return hlen;
325 }
326
ip_tunnel_encap(struct sk_buff * skb,struct ip_tunnel * t,u8 * protocol,struct flowi4 * fl4)327 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
328 u8 *protocol, struct flowi4 *fl4)
329 {
330 const struct ip_tunnel_encap_ops *ops;
331 int ret = -EINVAL;
332
333 if (t->encap.type == TUNNEL_ENCAP_NONE)
334 return 0;
335
336 if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
337 return -EINVAL;
338
339 rcu_read_lock();
340 ops = rcu_dereference(iptun_encaps[t->encap.type]);
341 if (likely(ops && ops->build_header))
342 ret = ops->build_header(skb, &t->encap, protocol, fl4);
343 rcu_read_unlock();
344
345 return ret;
346 }
347
348 /* Extract dsfield from inner protocol */
ip_tunnel_get_dsfield(const struct iphdr * iph,const struct sk_buff * skb)349 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
350 const struct sk_buff *skb)
351 {
352 if (skb->protocol == htons(ETH_P_IP))
353 return iph->tos;
354 else if (skb->protocol == htons(ETH_P_IPV6))
355 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
356 else
357 return 0;
358 }
359
360 /* Propogate ECN bits out */
ip_tunnel_ecn_encap(u8 tos,const struct iphdr * iph,const struct sk_buff * skb)361 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
362 const struct sk_buff *skb)
363 {
364 u8 inner = ip_tunnel_get_dsfield(iph, skb);
365
366 return INET_ECN_encapsulate(tos, inner);
367 }
368
369 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
370 __be16 inner_proto, bool raw_proto, bool xnet);
371
iptunnel_pull_header(struct sk_buff * skb,int hdr_len,__be16 inner_proto,bool xnet)372 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
373 __be16 inner_proto, bool xnet)
374 {
375 return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
376 }
377
378 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
379 __be32 src, __be32 dst, u8 proto,
380 u8 tos, u8 ttl, __be16 df, bool xnet);
381 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
382 gfp_t flags);
383
384 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
385
iptunnel_pull_offloads(struct sk_buff * skb)386 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
387 {
388 if (skb_is_gso(skb)) {
389 int err;
390
391 err = skb_unclone(skb, GFP_ATOMIC);
392 if (unlikely(err))
393 return err;
394 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
395 NETIF_F_GSO_SHIFT);
396 }
397
398 skb->encapsulation = 0;
399 return 0;
400 }
401
iptunnel_xmit_stats(struct net_device * dev,int pkt_len)402 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
403 {
404 if (pkt_len > 0) {
405 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
406
407 u64_stats_update_begin(&tstats->syncp);
408 tstats->tx_bytes += pkt_len;
409 tstats->tx_packets++;
410 u64_stats_update_end(&tstats->syncp);
411 put_cpu_ptr(tstats);
412 } else {
413 struct net_device_stats *err_stats = &dev->stats;
414
415 if (pkt_len < 0) {
416 err_stats->tx_errors++;
417 err_stats->tx_aborted_errors++;
418 } else {
419 err_stats->tx_dropped++;
420 }
421 }
422 }
423
ip_tunnel_info_opts(struct ip_tunnel_info * info)424 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
425 {
426 return info + 1;
427 }
428
ip_tunnel_info_opts_get(void * to,const struct ip_tunnel_info * info)429 static inline void ip_tunnel_info_opts_get(void *to,
430 const struct ip_tunnel_info *info)
431 {
432 memcpy(to, info + 1, info->options_len);
433 }
434
ip_tunnel_info_opts_set(struct ip_tunnel_info * info,const void * from,int len)435 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
436 const void *from, int len)
437 {
438 memcpy(ip_tunnel_info_opts(info), from, len);
439 info->options_len = len;
440 }
441
lwt_tun_info(struct lwtunnel_state * lwtstate)442 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
443 {
444 return (struct ip_tunnel_info *)lwtstate->data;
445 }
446
447 extern struct static_key ip_tunnel_metadata_cnt;
448
449 /* Returns > 0 if metadata should be collected */
ip_tunnel_collect_metadata(void)450 static inline int ip_tunnel_collect_metadata(void)
451 {
452 return static_key_false(&ip_tunnel_metadata_cnt);
453 }
454
455 void __init ip_tunnel_core_init(void);
456
457 void ip_tunnel_need_metadata(void);
458 void ip_tunnel_unneed_metadata(void);
459
460 #else /* CONFIG_INET */
461
lwt_tun_info(struct lwtunnel_state * lwtstate)462 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
463 {
464 return NULL;
465 }
466
ip_tunnel_need_metadata(void)467 static inline void ip_tunnel_need_metadata(void)
468 {
469 }
470
ip_tunnel_unneed_metadata(void)471 static inline void ip_tunnel_unneed_metadata(void)
472 {
473 }
474
ip_tunnel_info_opts_get(void * to,const struct ip_tunnel_info * info)475 static inline void ip_tunnel_info_opts_get(void *to,
476 const struct ip_tunnel_info *info)
477 {
478 }
479
ip_tunnel_info_opts_set(struct ip_tunnel_info * info,const void * from,int len)480 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
481 const void *from, int len)
482 {
483 info->options_len = 0;
484 }
485
486 #endif /* CONFIG_INET */
487
488 #endif /* __NET_IP_TUNNELS_H */
489