xref: /linux/include/net/ip_tunnels.h (revision be58f7103700a68d5c7ca60a2bc0b309907599ab)
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 	(sizeof_field(struct ip_tunnel_key, u) -		\
37 	 sizeof_field(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((sizeof_field(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 	/* These four fields used only by ERSPAN */
120 	u32		index;		/* ERSPAN type II index */
121 	u8		erspan_ver;	/* ERSPAN version */
122 	u8		dir;		/* ERSPAN direction */
123 	u16		hwid;		/* ERSPAN hardware ID */
124 
125 	struct dst_cache dst_cache;
126 
127 	struct ip_tunnel_parm parms;
128 
129 	int		mlink;
130 	int		encap_hlen;	/* Encap header length (FOU,GUE) */
131 	int		hlen;		/* tun_hlen + encap_hlen */
132 	struct ip_tunnel_encap encap;
133 
134 	/* for SIT */
135 #ifdef CONFIG_IPV6_SIT_6RD
136 	struct ip_tunnel_6rd_parm ip6rd;
137 #endif
138 	struct ip_tunnel_prl_entry __rcu *prl;	/* potential router list */
139 	unsigned int		prl_count;	/* # of entries in PRL */
140 	unsigned int		ip_tnl_net_id;
141 	struct gro_cells	gro_cells;
142 	__u32			fwmark;
143 	bool			collect_md;
144 	bool			ignore_df;
145 };
146 
147 struct tnl_ptk_info {
148 	__be16 flags;
149 	__be16 proto;
150 	__be32 key;
151 	__be32 seq;
152 	int hdr_len;
153 };
154 
155 #define PACKET_RCVD	0
156 #define PACKET_REJECT	1
157 #define PACKET_NEXT	2
158 
159 #define IP_TNL_HASH_BITS   7
160 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
161 
162 struct ip_tunnel_net {
163 	struct net_device *fb_tunnel_dev;
164 	struct rtnl_link_ops *rtnl_link_ops;
165 	struct hlist_head tunnels[IP_TNL_HASH_SIZE];
166 	struct ip_tunnel __rcu *collect_md_tun;
167 	int type;
168 };
169 
170 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
171 				      __be32 saddr, __be32 daddr,
172 				      u8 tos, u8 ttl, __be32 label,
173 				      __be16 tp_src, __be16 tp_dst,
174 				      __be64 tun_id, __be16 tun_flags)
175 {
176 	key->tun_id = tun_id;
177 	key->u.ipv4.src = saddr;
178 	key->u.ipv4.dst = daddr;
179 	memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
180 	       0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
181 	key->tos = tos;
182 	key->ttl = ttl;
183 	key->label = label;
184 	key->tun_flags = tun_flags;
185 
186 	/* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
187 	 * the upper tunnel are used.
188 	 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
189 	 */
190 	key->tp_src = tp_src;
191 	key->tp_dst = tp_dst;
192 
193 	/* Clear struct padding. */
194 	if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
195 		memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
196 		       0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
197 }
198 
199 static inline bool
200 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
201 			   const struct ip_tunnel_info *info)
202 {
203 	if (skb->mark)
204 		return false;
205 	if (!info)
206 		return true;
207 	if (info->key.tun_flags & TUNNEL_NOCACHE)
208 		return false;
209 
210 	return true;
211 }
212 
213 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
214 					       *tun_info)
215 {
216 	return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
217 }
218 
219 static inline __be64 key32_to_tunnel_id(__be32 key)
220 {
221 #ifdef __BIG_ENDIAN
222 	return (__force __be64)key;
223 #else
224 	return (__force __be64)((__force u64)key << 32);
225 #endif
226 }
227 
228 /* Returns the least-significant 32 bits of a __be64. */
229 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
230 {
231 #ifdef __BIG_ENDIAN
232 	return (__force __be32)tun_id;
233 #else
234 	return (__force __be32)((__force u64)tun_id >> 32);
235 #endif
236 }
237 
238 #ifdef CONFIG_INET
239 
240 static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
241 				       int proto,
242 				       __be32 daddr, __be32 saddr,
243 				       __be32 key, __u8 tos, int oif,
244 				       __u32 mark, __u32 tun_inner_hash)
245 {
246 	memset(fl4, 0, sizeof(*fl4));
247 	fl4->flowi4_oif = oif;
248 	fl4->daddr = daddr;
249 	fl4->saddr = saddr;
250 	fl4->flowi4_tos = tos;
251 	fl4->flowi4_proto = proto;
252 	fl4->fl4_gre_key = key;
253 	fl4->flowi4_mark = mark;
254 	fl4->flowi4_multipath_hash = tun_inner_hash;
255 }
256 
257 int ip_tunnel_init(struct net_device *dev);
258 void ip_tunnel_uninit(struct net_device *dev);
259 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
260 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
261 int ip_tunnel_get_iflink(const struct net_device *dev);
262 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
263 		       struct rtnl_link_ops *ops, char *devname);
264 
265 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
266 			   struct rtnl_link_ops *ops);
267 
268 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
269 		    const struct iphdr *tnl_params, const u8 protocol);
270 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
271 		       const u8 proto, int tunnel_hlen);
272 int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
273 int ip_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
274 			     void __user *data, int cmd);
275 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
276 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
277 
278 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
279 				   int link, __be16 flags,
280 				   __be32 remote, __be32 local,
281 				   __be32 key);
282 
283 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
284 		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
285 		  bool log_ecn_error);
286 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
287 			 struct ip_tunnel_parm *p, __u32 fwmark);
288 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
289 		      struct ip_tunnel_parm *p, __u32 fwmark);
290 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
291 
292 extern const struct header_ops ip_tunnel_header_ops;
293 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb);
294 
295 struct ip_tunnel_encap_ops {
296 	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
297 	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
298 			    u8 *protocol, struct flowi4 *fl4);
299 	int (*err_handler)(struct sk_buff *skb, u32 info);
300 };
301 
302 #define MAX_IPTUN_ENCAP_OPS 8
303 
304 extern const struct ip_tunnel_encap_ops __rcu *
305 		iptun_encaps[MAX_IPTUN_ENCAP_OPS];
306 
307 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
308 			    unsigned int num);
309 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
310 			    unsigned int num);
311 
312 int ip_tunnel_encap_setup(struct ip_tunnel *t,
313 			  struct ip_tunnel_encap *ipencap);
314 
315 static inline bool pskb_inet_may_pull(struct sk_buff *skb)
316 {
317 	int nhlen;
318 
319 	switch (skb->protocol) {
320 #if IS_ENABLED(CONFIG_IPV6)
321 	case htons(ETH_P_IPV6):
322 		nhlen = sizeof(struct ipv6hdr);
323 		break;
324 #endif
325 	case htons(ETH_P_IP):
326 		nhlen = sizeof(struct iphdr);
327 		break;
328 	default:
329 		nhlen = 0;
330 	}
331 
332 	return pskb_network_may_pull(skb, nhlen);
333 }
334 
335 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
336 {
337 	const struct ip_tunnel_encap_ops *ops;
338 	int hlen = -EINVAL;
339 
340 	if (e->type == TUNNEL_ENCAP_NONE)
341 		return 0;
342 
343 	if (e->type >= MAX_IPTUN_ENCAP_OPS)
344 		return -EINVAL;
345 
346 	rcu_read_lock();
347 	ops = rcu_dereference(iptun_encaps[e->type]);
348 	if (likely(ops && ops->encap_hlen))
349 		hlen = ops->encap_hlen(e);
350 	rcu_read_unlock();
351 
352 	return hlen;
353 }
354 
355 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
356 				  u8 *protocol, struct flowi4 *fl4)
357 {
358 	const struct ip_tunnel_encap_ops *ops;
359 	int ret = -EINVAL;
360 
361 	if (t->encap.type == TUNNEL_ENCAP_NONE)
362 		return 0;
363 
364 	if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
365 		return -EINVAL;
366 
367 	rcu_read_lock();
368 	ops = rcu_dereference(iptun_encaps[t->encap.type]);
369 	if (likely(ops && ops->build_header))
370 		ret = ops->build_header(skb, &t->encap, protocol, fl4);
371 	rcu_read_unlock();
372 
373 	return ret;
374 }
375 
376 /* Extract dsfield from inner protocol */
377 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
378 				       const struct sk_buff *skb)
379 {
380 	if (skb->protocol == htons(ETH_P_IP))
381 		return iph->tos;
382 	else if (skb->protocol == htons(ETH_P_IPV6))
383 		return ipv6_get_dsfield((const struct ipv6hdr *)iph);
384 	else
385 		return 0;
386 }
387 
388 static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
389 				       const struct sk_buff *skb)
390 {
391 	if (skb->protocol == htons(ETH_P_IP))
392 		return iph->ttl;
393 	else if (skb->protocol == htons(ETH_P_IPV6))
394 		return ((const struct ipv6hdr *)iph)->hop_limit;
395 	else
396 		return 0;
397 }
398 
399 /* Propogate ECN bits out */
400 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
401 				     const struct sk_buff *skb)
402 {
403 	u8 inner = ip_tunnel_get_dsfield(iph, skb);
404 
405 	return INET_ECN_encapsulate(tos, inner);
406 }
407 
408 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
409 			   __be16 inner_proto, bool raw_proto, bool xnet);
410 
411 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
412 				       __be16 inner_proto, bool xnet)
413 {
414 	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
415 }
416 
417 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
418 		   __be32 src, __be32 dst, u8 proto,
419 		   u8 tos, u8 ttl, __be16 df, bool xnet);
420 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
421 					     gfp_t flags);
422 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
423 			  int headroom, bool reply);
424 
425 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
426 
427 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
428 {
429 	if (skb_is_gso(skb)) {
430 		int err;
431 
432 		err = skb_unclone(skb, GFP_ATOMIC);
433 		if (unlikely(err))
434 			return err;
435 		skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
436 					       NETIF_F_GSO_SHIFT);
437 	}
438 
439 	skb->encapsulation = 0;
440 	return 0;
441 }
442 
443 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
444 {
445 	if (pkt_len > 0) {
446 		struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
447 
448 		u64_stats_update_begin(&tstats->syncp);
449 		tstats->tx_bytes += pkt_len;
450 		tstats->tx_packets++;
451 		u64_stats_update_end(&tstats->syncp);
452 		put_cpu_ptr(tstats);
453 	} else {
454 		struct net_device_stats *err_stats = &dev->stats;
455 
456 		if (pkt_len < 0) {
457 			err_stats->tx_errors++;
458 			err_stats->tx_aborted_errors++;
459 		} else {
460 			err_stats->tx_dropped++;
461 		}
462 	}
463 }
464 
465 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
466 {
467 	return info + 1;
468 }
469 
470 static inline void ip_tunnel_info_opts_get(void *to,
471 					   const struct ip_tunnel_info *info)
472 {
473 	memcpy(to, info + 1, info->options_len);
474 }
475 
476 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
477 					   const void *from, int len,
478 					   __be16 flags)
479 {
480 	info->options_len = len;
481 	if (len > 0) {
482 		memcpy(ip_tunnel_info_opts(info), from, len);
483 		info->key.tun_flags |= flags;
484 	}
485 }
486 
487 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
488 {
489 	return (struct ip_tunnel_info *)lwtstate->data;
490 }
491 
492 DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
493 
494 /* Returns > 0 if metadata should be collected */
495 static inline int ip_tunnel_collect_metadata(void)
496 {
497 	return static_branch_unlikely(&ip_tunnel_metadata_cnt);
498 }
499 
500 void __init ip_tunnel_core_init(void);
501 
502 void ip_tunnel_need_metadata(void);
503 void ip_tunnel_unneed_metadata(void);
504 
505 #else /* CONFIG_INET */
506 
507 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
508 {
509 	return NULL;
510 }
511 
512 static inline void ip_tunnel_need_metadata(void)
513 {
514 }
515 
516 static inline void ip_tunnel_unneed_metadata(void)
517 {
518 }
519 
520 static inline void ip_tunnel_info_opts_get(void *to,
521 					   const struct ip_tunnel_info *info)
522 {
523 }
524 
525 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
526 					   const void *from, int len,
527 					   __be16 flags)
528 {
529 	info->options_len = 0;
530 }
531 
532 #endif /* CONFIG_INET */
533 
534 #endif /* __NET_IP_TUNNELS_H */
535