xref: /openbmc/linux/include/net/ip_tunnels.h (revision 8bdc2a19)
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 
108 	struct net_device	*dev;
109 	netdevice_tracker	dev_tracker;
110 
111 	struct net		*net;	/* netns for packet i/o */
112 
113 	unsigned long	err_time;	/* Time when the last ICMP error
114 					 * arrived */
115 	int		err_count;	/* Number of arrived ICMP errors */
116 
117 	/* These four fields used only by GRE */
118 	u32		i_seqno;	/* The last seen seqno	*/
119 	atomic_t	o_seqno;	/* The last output seqno */
120 	int		tun_hlen;	/* Precalculated header length */
121 
122 	/* These four fields used only by ERSPAN */
123 	u32		index;		/* ERSPAN type II index */
124 	u8		erspan_ver;	/* ERSPAN version */
125 	u8		dir;		/* ERSPAN direction */
126 	u16		hwid;		/* ERSPAN hardware ID */
127 
128 	struct dst_cache dst_cache;
129 
130 	struct ip_tunnel_parm parms;
131 
132 	int		mlink;
133 	int		encap_hlen;	/* Encap header length (FOU,GUE) */
134 	int		hlen;		/* tun_hlen + encap_hlen */
135 	struct ip_tunnel_encap encap;
136 
137 	/* for SIT */
138 #ifdef CONFIG_IPV6_SIT_6RD
139 	struct ip_tunnel_6rd_parm ip6rd;
140 #endif
141 	struct ip_tunnel_prl_entry __rcu *prl;	/* potential router list */
142 	unsigned int		prl_count;	/* # of entries in PRL */
143 	unsigned int		ip_tnl_net_id;
144 	struct gro_cells	gro_cells;
145 	__u32			fwmark;
146 	bool			collect_md;
147 	bool			ignore_df;
148 };
149 
150 struct tnl_ptk_info {
151 	__be16 flags;
152 	__be16 proto;
153 	__be32 key;
154 	__be32 seq;
155 	int hdr_len;
156 };
157 
158 #define PACKET_RCVD	0
159 #define PACKET_REJECT	1
160 #define PACKET_NEXT	2
161 
162 #define IP_TNL_HASH_BITS   7
163 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
164 
165 struct ip_tunnel_net {
166 	struct net_device *fb_tunnel_dev;
167 	struct rtnl_link_ops *rtnl_link_ops;
168 	struct hlist_head tunnels[IP_TNL_HASH_SIZE];
169 	struct ip_tunnel __rcu *collect_md_tun;
170 	int type;
171 };
172 
173 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
174 				      __be32 saddr, __be32 daddr,
175 				      u8 tos, u8 ttl, __be32 label,
176 				      __be16 tp_src, __be16 tp_dst,
177 				      __be64 tun_id, __be16 tun_flags)
178 {
179 	key->tun_id = tun_id;
180 	key->u.ipv4.src = saddr;
181 	key->u.ipv4.dst = daddr;
182 	memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
183 	       0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
184 	key->tos = tos;
185 	key->ttl = ttl;
186 	key->label = label;
187 	key->tun_flags = tun_flags;
188 
189 	/* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
190 	 * the upper tunnel are used.
191 	 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
192 	 */
193 	key->tp_src = tp_src;
194 	key->tp_dst = tp_dst;
195 
196 	/* Clear struct padding. */
197 	if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
198 		memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
199 		       0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
200 }
201 
202 static inline bool
203 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
204 			   const struct ip_tunnel_info *info)
205 {
206 	if (skb->mark)
207 		return false;
208 	if (!info)
209 		return true;
210 	if (info->key.tun_flags & TUNNEL_NOCACHE)
211 		return false;
212 
213 	return true;
214 }
215 
216 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
217 					       *tun_info)
218 {
219 	return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
220 }
221 
222 static inline __be64 key32_to_tunnel_id(__be32 key)
223 {
224 #ifdef __BIG_ENDIAN
225 	return (__force __be64)key;
226 #else
227 	return (__force __be64)((__force u64)key << 32);
228 #endif
229 }
230 
231 /* Returns the least-significant 32 bits of a __be64. */
232 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
233 {
234 #ifdef __BIG_ENDIAN
235 	return (__force __be32)tun_id;
236 #else
237 	return (__force __be32)((__force u64)tun_id >> 32);
238 #endif
239 }
240 
241 #ifdef CONFIG_INET
242 
243 static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
244 				       int proto,
245 				       __be32 daddr, __be32 saddr,
246 				       __be32 key, __u8 tos,
247 				       struct net *net, int oif,
248 				       __u32 mark, __u32 tun_inner_hash)
249 {
250 	memset(fl4, 0, sizeof(*fl4));
251 
252 	if (oif) {
253 		fl4->flowi4_l3mdev = l3mdev_master_upper_ifindex_by_index_rcu(net, oif);
254 		/* Legacy VRF/l3mdev use case */
255 		fl4->flowi4_oif = fl4->flowi4_l3mdev ? 0 : oif;
256 	}
257 
258 	fl4->daddr = daddr;
259 	fl4->saddr = saddr;
260 	fl4->flowi4_tos = tos;
261 	fl4->flowi4_proto = proto;
262 	fl4->fl4_gre_key = key;
263 	fl4->flowi4_mark = mark;
264 	fl4->flowi4_multipath_hash = tun_inner_hash;
265 }
266 
267 int ip_tunnel_init(struct net_device *dev);
268 void ip_tunnel_uninit(struct net_device *dev);
269 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
270 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
271 int ip_tunnel_get_iflink(const struct net_device *dev);
272 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
273 		       struct rtnl_link_ops *ops, char *devname);
274 
275 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
276 			   struct rtnl_link_ops *ops);
277 
278 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
279 		    const struct iphdr *tnl_params, const u8 protocol);
280 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
281 		       const u8 proto, int tunnel_hlen);
282 int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
283 int ip_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
284 			     void __user *data, int cmd);
285 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
286 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
287 
288 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
289 				   int link, __be16 flags,
290 				   __be32 remote, __be32 local,
291 				   __be32 key);
292 
293 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
294 		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
295 		  bool log_ecn_error);
296 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
297 			 struct ip_tunnel_parm *p, __u32 fwmark);
298 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
299 		      struct ip_tunnel_parm *p, __u32 fwmark);
300 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
301 
302 extern const struct header_ops ip_tunnel_header_ops;
303 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb);
304 
305 struct ip_tunnel_encap_ops {
306 	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
307 	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
308 			    u8 *protocol, struct flowi4 *fl4);
309 	int (*err_handler)(struct sk_buff *skb, u32 info);
310 };
311 
312 #define MAX_IPTUN_ENCAP_OPS 8
313 
314 extern const struct ip_tunnel_encap_ops __rcu *
315 		iptun_encaps[MAX_IPTUN_ENCAP_OPS];
316 
317 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
318 			    unsigned int num);
319 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
320 			    unsigned int num);
321 
322 int ip_tunnel_encap_setup(struct ip_tunnel *t,
323 			  struct ip_tunnel_encap *ipencap);
324 
325 static inline bool pskb_inet_may_pull(struct sk_buff *skb)
326 {
327 	int nhlen;
328 
329 	switch (skb->protocol) {
330 #if IS_ENABLED(CONFIG_IPV6)
331 	case htons(ETH_P_IPV6):
332 		nhlen = sizeof(struct ipv6hdr);
333 		break;
334 #endif
335 	case htons(ETH_P_IP):
336 		nhlen = sizeof(struct iphdr);
337 		break;
338 	default:
339 		nhlen = 0;
340 	}
341 
342 	return pskb_network_may_pull(skb, nhlen);
343 }
344 
345 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
346 {
347 	const struct ip_tunnel_encap_ops *ops;
348 	int hlen = -EINVAL;
349 
350 	if (e->type == TUNNEL_ENCAP_NONE)
351 		return 0;
352 
353 	if (e->type >= MAX_IPTUN_ENCAP_OPS)
354 		return -EINVAL;
355 
356 	rcu_read_lock();
357 	ops = rcu_dereference(iptun_encaps[e->type]);
358 	if (likely(ops && ops->encap_hlen))
359 		hlen = ops->encap_hlen(e);
360 	rcu_read_unlock();
361 
362 	return hlen;
363 }
364 
365 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
366 				  u8 *protocol, struct flowi4 *fl4)
367 {
368 	const struct ip_tunnel_encap_ops *ops;
369 	int ret = -EINVAL;
370 
371 	if (t->encap.type == TUNNEL_ENCAP_NONE)
372 		return 0;
373 
374 	if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
375 		return -EINVAL;
376 
377 	rcu_read_lock();
378 	ops = rcu_dereference(iptun_encaps[t->encap.type]);
379 	if (likely(ops && ops->build_header))
380 		ret = ops->build_header(skb, &t->encap, protocol, fl4);
381 	rcu_read_unlock();
382 
383 	return ret;
384 }
385 
386 /* Extract dsfield from inner protocol */
387 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
388 				       const struct sk_buff *skb)
389 {
390 	if (skb->protocol == htons(ETH_P_IP))
391 		return iph->tos;
392 	else if (skb->protocol == htons(ETH_P_IPV6))
393 		return ipv6_get_dsfield((const struct ipv6hdr *)iph);
394 	else
395 		return 0;
396 }
397 
398 static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
399 				       const struct sk_buff *skb)
400 {
401 	if (skb->protocol == htons(ETH_P_IP))
402 		return iph->ttl;
403 	else if (skb->protocol == htons(ETH_P_IPV6))
404 		return ((const struct ipv6hdr *)iph)->hop_limit;
405 	else
406 		return 0;
407 }
408 
409 /* Propogate ECN bits out */
410 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
411 				     const struct sk_buff *skb)
412 {
413 	u8 inner = ip_tunnel_get_dsfield(iph, skb);
414 
415 	return INET_ECN_encapsulate(tos, inner);
416 }
417 
418 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
419 			   __be16 inner_proto, bool raw_proto, bool xnet);
420 
421 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
422 				       __be16 inner_proto, bool xnet)
423 {
424 	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
425 }
426 
427 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
428 		   __be32 src, __be32 dst, u8 proto,
429 		   u8 tos, u8 ttl, __be16 df, bool xnet);
430 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
431 					     gfp_t flags);
432 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
433 			  int headroom, bool reply);
434 
435 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
436 
437 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
438 {
439 	if (skb_is_gso(skb)) {
440 		int err;
441 
442 		err = skb_unclone(skb, GFP_ATOMIC);
443 		if (unlikely(err))
444 			return err;
445 		skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
446 					       NETIF_F_GSO_SHIFT);
447 	}
448 
449 	skb->encapsulation = 0;
450 	return 0;
451 }
452 
453 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
454 {
455 	if (pkt_len > 0) {
456 		struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
457 
458 		u64_stats_update_begin(&tstats->syncp);
459 		tstats->tx_bytes += pkt_len;
460 		tstats->tx_packets++;
461 		u64_stats_update_end(&tstats->syncp);
462 		put_cpu_ptr(tstats);
463 	} else {
464 		struct net_device_stats *err_stats = &dev->stats;
465 
466 		if (pkt_len < 0) {
467 			err_stats->tx_errors++;
468 			err_stats->tx_aborted_errors++;
469 		} else {
470 			err_stats->tx_dropped++;
471 		}
472 	}
473 }
474 
475 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
476 {
477 	return info + 1;
478 }
479 
480 static inline void ip_tunnel_info_opts_get(void *to,
481 					   const struct ip_tunnel_info *info)
482 {
483 	memcpy(to, info + 1, info->options_len);
484 }
485 
486 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
487 					   const void *from, int len,
488 					   __be16 flags)
489 {
490 	info->options_len = len;
491 	if (len > 0) {
492 		memcpy(ip_tunnel_info_opts(info), from, len);
493 		info->key.tun_flags |= flags;
494 	}
495 }
496 
497 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
498 {
499 	return (struct ip_tunnel_info *)lwtstate->data;
500 }
501 
502 DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
503 
504 /* Returns > 0 if metadata should be collected */
505 static inline int ip_tunnel_collect_metadata(void)
506 {
507 	return static_branch_unlikely(&ip_tunnel_metadata_cnt);
508 }
509 
510 void __init ip_tunnel_core_init(void);
511 
512 void ip_tunnel_need_metadata(void);
513 void ip_tunnel_unneed_metadata(void);
514 
515 #else /* CONFIG_INET */
516 
517 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
518 {
519 	return NULL;
520 }
521 
522 static inline void ip_tunnel_need_metadata(void)
523 {
524 }
525 
526 static inline void ip_tunnel_unneed_metadata(void)
527 {
528 }
529 
530 static inline void ip_tunnel_info_opts_get(void *to,
531 					   const struct ip_tunnel_info *info)
532 {
533 }
534 
535 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
536 					   const void *from, int len,
537 					   __be16 flags)
538 {
539 	info->options_len = 0;
540 }
541 
542 #endif /* CONFIG_INET */
543 
544 #endif /* __NET_IP_TUNNELS_H */
545