xref: /linux/include/net/ip_tunnels.h (revision a3a4a816b4b194c45d0217e8b9e08b2639802cda)
1 #ifndef __NET_IP_TUNNELS_H
2 #define __NET_IP_TUNNELS_H 1
3 
4 #include <linux/if_tunnel.h>
5 #include <linux/netdevice.h>
6 #include <linux/skbuff.h>
7 #include <linux/socket.h>
8 #include <linux/types.h>
9 #include <linux/u64_stats_sync.h>
10 #include <linux/bitops.h>
11 
12 #include <net/dsfield.h>
13 #include <net/gro_cells.h>
14 #include <net/inet_ecn.h>
15 #include <net/netns/generic.h>
16 #include <net/rtnetlink.h>
17 #include <net/lwtunnel.h>
18 #include <net/dst_cache.h>
19 
20 #if IS_ENABLED(CONFIG_IPV6)
21 #include <net/ipv6.h>
22 #include <net/ip6_fib.h>
23 #include <net/ip6_route.h>
24 #endif
25 
26 /* Keep error state on tunnel for 30 sec */
27 #define IPTUNNEL_ERR_TIMEO	(30*HZ)
28 
29 /* Used to memset ip_tunnel padding. */
30 #define IP_TUNNEL_KEY_SIZE	offsetofend(struct ip_tunnel_key, tp_dst)
31 
32 /* Used to memset ipv4 address padding. */
33 #define IP_TUNNEL_KEY_IPV4_PAD	offsetofend(struct ip_tunnel_key, u.ipv4.dst)
34 #define IP_TUNNEL_KEY_IPV4_PAD_LEN				\
35 	(FIELD_SIZEOF(struct ip_tunnel_key, u) -		\
36 	 FIELD_SIZEOF(struct ip_tunnel_key, u.ipv4))
37 
38 struct ip_tunnel_key {
39 	__be64			tun_id;
40 	union {
41 		struct {
42 			__be32	src;
43 			__be32	dst;
44 		} ipv4;
45 		struct {
46 			struct in6_addr src;
47 			struct in6_addr dst;
48 		} ipv6;
49 	} u;
50 	__be16			tun_flags;
51 	u8			tos;		/* TOS for IPv4, TC for IPv6 */
52 	u8			ttl;		/* TTL for IPv4, HL for IPv6 */
53 	__be32			label;		/* Flow Label for IPv6 */
54 	__be16			tp_src;
55 	__be16			tp_dst;
56 };
57 
58 /* Flags for ip_tunnel_info mode. */
59 #define IP_TUNNEL_INFO_TX	0x01	/* represents tx tunnel parameters */
60 #define IP_TUNNEL_INFO_IPV6	0x02	/* key contains IPv6 addresses */
61 #define IP_TUNNEL_INFO_BRIDGE	0x04	/* represents a bridged tunnel id */
62 
63 /* Maximum tunnel options length. */
64 #define IP_TUNNEL_OPTS_MAX					\
65 	GENMASK((FIELD_SIZEOF(struct ip_tunnel_info,		\
66 			      options_len) * BITS_PER_BYTE) - 1, 0)
67 
68 struct ip_tunnel_info {
69 	struct ip_tunnel_key	key;
70 #ifdef CONFIG_DST_CACHE
71 	struct dst_cache	dst_cache;
72 #endif
73 	u8			options_len;
74 	u8			mode;
75 };
76 
77 /* 6rd prefix/relay information */
78 #ifdef CONFIG_IPV6_SIT_6RD
79 struct ip_tunnel_6rd_parm {
80 	struct in6_addr		prefix;
81 	__be32			relay_prefix;
82 	u16			prefixlen;
83 	u16			relay_prefixlen;
84 };
85 #endif
86 
87 struct ip_tunnel_encap {
88 	u16			type;
89 	u16			flags;
90 	__be16			sport;
91 	__be16			dport;
92 };
93 
94 struct ip_tunnel_prl_entry {
95 	struct ip_tunnel_prl_entry __rcu *next;
96 	__be32				addr;
97 	u16				flags;
98 	struct rcu_head			rcu_head;
99 };
100 
101 struct metadata_dst;
102 
103 struct ip_tunnel {
104 	struct ip_tunnel __rcu	*next;
105 	struct hlist_node hash_node;
106 	struct net_device	*dev;
107 	struct net		*net;	/* netns for packet i/o */
108 
109 	unsigned long	err_time;	/* Time when the last ICMP error
110 					 * arrived */
111 	int		err_count;	/* Number of arrived ICMP errors */
112 
113 	/* These four fields used only by GRE */
114 	u32		i_seqno;	/* The last seen seqno	*/
115 	u32		o_seqno;	/* The last output seqno */
116 	int		tun_hlen;	/* Precalculated header length */
117 
118 	struct dst_cache dst_cache;
119 
120 	struct ip_tunnel_parm parms;
121 
122 	int		mlink;
123 	int		encap_hlen;	/* Encap header length (FOU,GUE) */
124 	int		hlen;		/* tun_hlen + encap_hlen */
125 	struct ip_tunnel_encap encap;
126 
127 	/* for SIT */
128 #ifdef CONFIG_IPV6_SIT_6RD
129 	struct ip_tunnel_6rd_parm ip6rd;
130 #endif
131 	struct ip_tunnel_prl_entry __rcu *prl;	/* potential router list */
132 	unsigned int		prl_count;	/* # of entries in PRL */
133 	unsigned int		ip_tnl_net_id;
134 	struct gro_cells	gro_cells;
135 	bool			collect_md;
136 	bool			ignore_df;
137 };
138 
139 #define TUNNEL_CSUM		__cpu_to_be16(0x01)
140 #define TUNNEL_ROUTING		__cpu_to_be16(0x02)
141 #define TUNNEL_KEY		__cpu_to_be16(0x04)
142 #define TUNNEL_SEQ		__cpu_to_be16(0x08)
143 #define TUNNEL_STRICT		__cpu_to_be16(0x10)
144 #define TUNNEL_REC		__cpu_to_be16(0x20)
145 #define TUNNEL_VERSION		__cpu_to_be16(0x40)
146 #define TUNNEL_NO_KEY		__cpu_to_be16(0x80)
147 #define TUNNEL_DONT_FRAGMENT    __cpu_to_be16(0x0100)
148 #define TUNNEL_OAM		__cpu_to_be16(0x0200)
149 #define TUNNEL_CRIT_OPT		__cpu_to_be16(0x0400)
150 #define TUNNEL_GENEVE_OPT	__cpu_to_be16(0x0800)
151 #define TUNNEL_VXLAN_OPT	__cpu_to_be16(0x1000)
152 #define TUNNEL_NOCACHE		__cpu_to_be16(0x2000)
153 
154 #define TUNNEL_OPTIONS_PRESENT	(TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT)
155 
156 struct tnl_ptk_info {
157 	__be16 flags;
158 	__be16 proto;
159 	__be32 key;
160 	__be32 seq;
161 	int hdr_len;
162 };
163 
164 #define PACKET_RCVD	0
165 #define PACKET_REJECT	1
166 #define PACKET_NEXT	2
167 
168 #define IP_TNL_HASH_BITS   7
169 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
170 
171 struct ip_tunnel_net {
172 	struct net_device *fb_tunnel_dev;
173 	struct hlist_head tunnels[IP_TNL_HASH_SIZE];
174 	struct ip_tunnel __rcu *collect_md_tun;
175 };
176 
177 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
178 				      __be32 saddr, __be32 daddr,
179 				      u8 tos, u8 ttl, __be32 label,
180 				      __be16 tp_src, __be16 tp_dst,
181 				      __be64 tun_id, __be16 tun_flags)
182 {
183 	key->tun_id = tun_id;
184 	key->u.ipv4.src = saddr;
185 	key->u.ipv4.dst = daddr;
186 	memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
187 	       0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
188 	key->tos = tos;
189 	key->ttl = ttl;
190 	key->label = label;
191 	key->tun_flags = tun_flags;
192 
193 	/* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
194 	 * the upper tunnel are used.
195 	 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
196 	 */
197 	key->tp_src = tp_src;
198 	key->tp_dst = tp_dst;
199 
200 	/* Clear struct padding. */
201 	if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
202 		memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
203 		       0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
204 }
205 
206 static inline bool
207 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
208 			   const struct ip_tunnel_info *info)
209 {
210 	if (skb->mark)
211 		return false;
212 	if (!info)
213 		return true;
214 	if (info->key.tun_flags & TUNNEL_NOCACHE)
215 		return false;
216 
217 	return true;
218 }
219 
220 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
221 					       *tun_info)
222 {
223 	return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
224 }
225 
226 static inline __be64 key32_to_tunnel_id(__be32 key)
227 {
228 #ifdef __BIG_ENDIAN
229 	return (__force __be64)key;
230 #else
231 	return (__force __be64)((__force u64)key << 32);
232 #endif
233 }
234 
235 /* Returns the least-significant 32 bits of a __be64. */
236 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
237 {
238 #ifdef __BIG_ENDIAN
239 	return (__force __be32)tun_id;
240 #else
241 	return (__force __be32)((__force u64)tun_id >> 32);
242 #endif
243 }
244 
245 #ifdef CONFIG_INET
246 
247 int ip_tunnel_init(struct net_device *dev);
248 void ip_tunnel_uninit(struct net_device *dev);
249 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
250 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
251 int ip_tunnel_get_iflink(const struct net_device *dev);
252 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
253 		       struct rtnl_link_ops *ops, char *devname);
254 
255 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops);
256 
257 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
258 		    const struct iphdr *tnl_params, const u8 protocol);
259 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
260 		       const u8 proto);
261 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
262 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
263 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
264 
265 void ip_tunnel_get_stats64(struct net_device *dev,
266 			   struct rtnl_link_stats64 *tot);
267 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
268 				   int link, __be16 flags,
269 				   __be32 remote, __be32 local,
270 				   __be32 key);
271 
272 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
273 		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
274 		  bool log_ecn_error);
275 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
276 			 struct ip_tunnel_parm *p);
277 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
278 		      struct ip_tunnel_parm *p);
279 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
280 
281 struct ip_tunnel_encap_ops {
282 	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
283 	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
284 			    u8 *protocol, struct flowi4 *fl4);
285 };
286 
287 #define MAX_IPTUN_ENCAP_OPS 8
288 
289 extern const struct ip_tunnel_encap_ops __rcu *
290 		iptun_encaps[MAX_IPTUN_ENCAP_OPS];
291 
292 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
293 			    unsigned int num);
294 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
295 			    unsigned int num);
296 
297 int ip_tunnel_encap_setup(struct ip_tunnel *t,
298 			  struct ip_tunnel_encap *ipencap);
299 
300 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
301 {
302 	const struct ip_tunnel_encap_ops *ops;
303 	int hlen = -EINVAL;
304 
305 	if (e->type == TUNNEL_ENCAP_NONE)
306 		return 0;
307 
308 	if (e->type >= MAX_IPTUN_ENCAP_OPS)
309 		return -EINVAL;
310 
311 	rcu_read_lock();
312 	ops = rcu_dereference(iptun_encaps[e->type]);
313 	if (likely(ops && ops->encap_hlen))
314 		hlen = ops->encap_hlen(e);
315 	rcu_read_unlock();
316 
317 	return hlen;
318 }
319 
320 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
321 				  u8 *protocol, struct flowi4 *fl4)
322 {
323 	const struct ip_tunnel_encap_ops *ops;
324 	int ret = -EINVAL;
325 
326 	if (t->encap.type == TUNNEL_ENCAP_NONE)
327 		return 0;
328 
329 	if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
330 		return -EINVAL;
331 
332 	rcu_read_lock();
333 	ops = rcu_dereference(iptun_encaps[t->encap.type]);
334 	if (likely(ops && ops->build_header))
335 		ret = ops->build_header(skb, &t->encap, protocol, fl4);
336 	rcu_read_unlock();
337 
338 	return ret;
339 }
340 
341 /* Extract dsfield from inner protocol */
342 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
343 				       const struct sk_buff *skb)
344 {
345 	if (skb->protocol == htons(ETH_P_IP))
346 		return iph->tos;
347 	else if (skb->protocol == htons(ETH_P_IPV6))
348 		return ipv6_get_dsfield((const struct ipv6hdr *)iph);
349 	else
350 		return 0;
351 }
352 
353 /* Propogate ECN bits out */
354 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
355 				     const struct sk_buff *skb)
356 {
357 	u8 inner = ip_tunnel_get_dsfield(iph, skb);
358 
359 	return INET_ECN_encapsulate(tos, inner);
360 }
361 
362 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
363 			   __be16 inner_proto, bool raw_proto, bool xnet);
364 
365 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
366 				       __be16 inner_proto, bool xnet)
367 {
368 	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
369 }
370 
371 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
372 		   __be32 src, __be32 dst, u8 proto,
373 		   u8 tos, u8 ttl, __be16 df, bool xnet);
374 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
375 					     gfp_t flags);
376 
377 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
378 
379 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
380 {
381 	if (skb_is_gso(skb)) {
382 		int err;
383 
384 		err = skb_unclone(skb, GFP_ATOMIC);
385 		if (unlikely(err))
386 			return err;
387 		skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
388 					       NETIF_F_GSO_SHIFT);
389 	}
390 
391 	skb->encapsulation = 0;
392 	return 0;
393 }
394 
395 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
396 {
397 	if (pkt_len > 0) {
398 		struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
399 
400 		u64_stats_update_begin(&tstats->syncp);
401 		tstats->tx_bytes += pkt_len;
402 		tstats->tx_packets++;
403 		u64_stats_update_end(&tstats->syncp);
404 		put_cpu_ptr(tstats);
405 	} else {
406 		struct net_device_stats *err_stats = &dev->stats;
407 
408 		if (pkt_len < 0) {
409 			err_stats->tx_errors++;
410 			err_stats->tx_aborted_errors++;
411 		} else {
412 			err_stats->tx_dropped++;
413 		}
414 	}
415 }
416 
417 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
418 {
419 	return info + 1;
420 }
421 
422 static inline void ip_tunnel_info_opts_get(void *to,
423 					   const struct ip_tunnel_info *info)
424 {
425 	memcpy(to, info + 1, info->options_len);
426 }
427 
428 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
429 					   const void *from, int len)
430 {
431 	memcpy(ip_tunnel_info_opts(info), from, len);
432 	info->options_len = len;
433 }
434 
435 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
436 {
437 	return (struct ip_tunnel_info *)lwtstate->data;
438 }
439 
440 extern struct static_key ip_tunnel_metadata_cnt;
441 
442 /* Returns > 0 if metadata should be collected */
443 static inline int ip_tunnel_collect_metadata(void)
444 {
445 	return static_key_false(&ip_tunnel_metadata_cnt);
446 }
447 
448 void __init ip_tunnel_core_init(void);
449 
450 void ip_tunnel_need_metadata(void);
451 void ip_tunnel_unneed_metadata(void);
452 
453 #else /* CONFIG_INET */
454 
455 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
456 {
457 	return NULL;
458 }
459 
460 static inline void ip_tunnel_need_metadata(void)
461 {
462 }
463 
464 static inline void ip_tunnel_unneed_metadata(void)
465 {
466 }
467 
468 static inline void ip_tunnel_info_opts_get(void *to,
469 					   const struct ip_tunnel_info *info)
470 {
471 }
472 
473 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
474 					   const void *from, int len)
475 {
476 	info->options_len = 0;
477 }
478 
479 #endif /* CONFIG_INET */
480 
481 #endif /* __NET_IP_TUNNELS_H */
482