xref: /openbmc/linux/include/net/ip_tunnels.h (revision 8730046c)
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 
62 /* Maximum tunnel options length. */
63 #define IP_TUNNEL_OPTS_MAX					\
64 	GENMASK((FIELD_SIZEOF(struct ip_tunnel_info,		\
65 			      options_len) * BITS_PER_BYTE) - 1, 0)
66 
67 struct ip_tunnel_info {
68 	struct ip_tunnel_key	key;
69 #ifdef CONFIG_DST_CACHE
70 	struct dst_cache	dst_cache;
71 #endif
72 	u8			options_len;
73 	u8			mode;
74 };
75 
76 /* 6rd prefix/relay information */
77 #ifdef CONFIG_IPV6_SIT_6RD
78 struct ip_tunnel_6rd_parm {
79 	struct in6_addr		prefix;
80 	__be32			relay_prefix;
81 	u16			prefixlen;
82 	u16			relay_prefixlen;
83 };
84 #endif
85 
86 struct ip_tunnel_encap {
87 	u16			type;
88 	u16			flags;
89 	__be16			sport;
90 	__be16			dport;
91 };
92 
93 struct ip_tunnel_prl_entry {
94 	struct ip_tunnel_prl_entry __rcu *next;
95 	__be32				addr;
96 	u16				flags;
97 	struct rcu_head			rcu_head;
98 };
99 
100 struct metadata_dst;
101 
102 struct ip_tunnel {
103 	struct ip_tunnel __rcu	*next;
104 	struct hlist_node hash_node;
105 	struct net_device	*dev;
106 	struct net		*net;	/* netns for packet i/o */
107 
108 	unsigned long	err_time;	/* Time when the last ICMP error
109 					 * arrived */
110 	int		err_count;	/* Number of arrived ICMP errors */
111 
112 	/* These four fields used only by GRE */
113 	u32		i_seqno;	/* The last seen seqno	*/
114 	u32		o_seqno;	/* The last output seqno */
115 	int		tun_hlen;	/* Precalculated header length */
116 
117 	struct dst_cache dst_cache;
118 
119 	struct ip_tunnel_parm parms;
120 
121 	int		mlink;
122 	int		encap_hlen;	/* Encap header length (FOU,GUE) */
123 	int		hlen;		/* tun_hlen + encap_hlen */
124 	struct ip_tunnel_encap encap;
125 
126 	/* for SIT */
127 #ifdef CONFIG_IPV6_SIT_6RD
128 	struct ip_tunnel_6rd_parm ip6rd;
129 #endif
130 	struct ip_tunnel_prl_entry __rcu *prl;	/* potential router list */
131 	unsigned int		prl_count;	/* # of entries in PRL */
132 	unsigned int		ip_tnl_net_id;
133 	struct gro_cells	gro_cells;
134 	bool			collect_md;
135 	bool			ignore_df;
136 };
137 
138 #define TUNNEL_CSUM		__cpu_to_be16(0x01)
139 #define TUNNEL_ROUTING		__cpu_to_be16(0x02)
140 #define TUNNEL_KEY		__cpu_to_be16(0x04)
141 #define TUNNEL_SEQ		__cpu_to_be16(0x08)
142 #define TUNNEL_STRICT		__cpu_to_be16(0x10)
143 #define TUNNEL_REC		__cpu_to_be16(0x20)
144 #define TUNNEL_VERSION		__cpu_to_be16(0x40)
145 #define TUNNEL_NO_KEY		__cpu_to_be16(0x80)
146 #define TUNNEL_DONT_FRAGMENT    __cpu_to_be16(0x0100)
147 #define TUNNEL_OAM		__cpu_to_be16(0x0200)
148 #define TUNNEL_CRIT_OPT		__cpu_to_be16(0x0400)
149 #define TUNNEL_GENEVE_OPT	__cpu_to_be16(0x0800)
150 #define TUNNEL_VXLAN_OPT	__cpu_to_be16(0x1000)
151 #define TUNNEL_NOCACHE		__cpu_to_be16(0x2000)
152 
153 #define TUNNEL_OPTIONS_PRESENT	(TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT)
154 
155 struct tnl_ptk_info {
156 	__be16 flags;
157 	__be16 proto;
158 	__be32 key;
159 	__be32 seq;
160 	int hdr_len;
161 };
162 
163 #define PACKET_RCVD	0
164 #define PACKET_REJECT	1
165 #define PACKET_NEXT	2
166 
167 #define IP_TNL_HASH_BITS   7
168 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
169 
170 struct ip_tunnel_net {
171 	struct net_device *fb_tunnel_dev;
172 	struct hlist_head tunnels[IP_TNL_HASH_SIZE];
173 	struct ip_tunnel __rcu *collect_md_tun;
174 };
175 
176 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
177 				      __be32 saddr, __be32 daddr,
178 				      u8 tos, u8 ttl, __be32 label,
179 				      __be16 tp_src, __be16 tp_dst,
180 				      __be64 tun_id, __be16 tun_flags)
181 {
182 	key->tun_id = tun_id;
183 	key->u.ipv4.src = saddr;
184 	key->u.ipv4.dst = daddr;
185 	memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
186 	       0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
187 	key->tos = tos;
188 	key->ttl = ttl;
189 	key->label = label;
190 	key->tun_flags = tun_flags;
191 
192 	/* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
193 	 * the upper tunnel are used.
194 	 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
195 	 */
196 	key->tp_src = tp_src;
197 	key->tp_dst = tp_dst;
198 
199 	/* Clear struct padding. */
200 	if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
201 		memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
202 		       0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
203 }
204 
205 static inline bool
206 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
207 			   const struct ip_tunnel_info *info)
208 {
209 	if (skb->mark)
210 		return false;
211 	if (!info)
212 		return true;
213 	if (info->key.tun_flags & TUNNEL_NOCACHE)
214 		return false;
215 
216 	return true;
217 }
218 
219 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
220 					       *tun_info)
221 {
222 	return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
223 }
224 
225 static inline __be64 key32_to_tunnel_id(__be32 key)
226 {
227 #ifdef __BIG_ENDIAN
228 	return (__force __be64)key;
229 #else
230 	return (__force __be64)((__force u64)key << 32);
231 #endif
232 }
233 
234 /* Returns the least-significant 32 bits of a __be64. */
235 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
236 {
237 #ifdef __BIG_ENDIAN
238 	return (__force __be32)tun_id;
239 #else
240 	return (__force __be32)((__force u64)tun_id >> 32);
241 #endif
242 }
243 
244 #ifdef CONFIG_INET
245 
246 int ip_tunnel_init(struct net_device *dev);
247 void ip_tunnel_uninit(struct net_device *dev);
248 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
249 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
250 int ip_tunnel_get_iflink(const struct net_device *dev);
251 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
252 		       struct rtnl_link_ops *ops, char *devname);
253 
254 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops);
255 
256 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
257 		    const struct iphdr *tnl_params, const u8 protocol);
258 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
259 		       const u8 proto);
260 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
261 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
262 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
263 
264 struct rtnl_link_stats64 *ip_tunnel_get_stats64(struct net_device *dev,
265 						struct rtnl_link_stats64 *tot);
266 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
267 				   int link, __be16 flags,
268 				   __be32 remote, __be32 local,
269 				   __be32 key);
270 
271 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
272 		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
273 		  bool log_ecn_error);
274 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
275 			 struct ip_tunnel_parm *p);
276 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
277 		      struct ip_tunnel_parm *p);
278 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
279 
280 struct ip_tunnel_encap_ops {
281 	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
282 	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
283 			    u8 *protocol, struct flowi4 *fl4);
284 };
285 
286 #define MAX_IPTUN_ENCAP_OPS 8
287 
288 extern const struct ip_tunnel_encap_ops __rcu *
289 		iptun_encaps[MAX_IPTUN_ENCAP_OPS];
290 
291 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
292 			    unsigned int num);
293 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
294 			    unsigned int num);
295 
296 int ip_tunnel_encap_setup(struct ip_tunnel *t,
297 			  struct ip_tunnel_encap *ipencap);
298 
299 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
300 {
301 	const struct ip_tunnel_encap_ops *ops;
302 	int hlen = -EINVAL;
303 
304 	if (e->type == TUNNEL_ENCAP_NONE)
305 		return 0;
306 
307 	if (e->type >= MAX_IPTUN_ENCAP_OPS)
308 		return -EINVAL;
309 
310 	rcu_read_lock();
311 	ops = rcu_dereference(iptun_encaps[e->type]);
312 	if (likely(ops && ops->encap_hlen))
313 		hlen = ops->encap_hlen(e);
314 	rcu_read_unlock();
315 
316 	return hlen;
317 }
318 
319 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
320 				  u8 *protocol, struct flowi4 *fl4)
321 {
322 	const struct ip_tunnel_encap_ops *ops;
323 	int ret = -EINVAL;
324 
325 	if (t->encap.type == TUNNEL_ENCAP_NONE)
326 		return 0;
327 
328 	if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
329 		return -EINVAL;
330 
331 	rcu_read_lock();
332 	ops = rcu_dereference(iptun_encaps[t->encap.type]);
333 	if (likely(ops && ops->build_header))
334 		ret = ops->build_header(skb, &t->encap, protocol, fl4);
335 	rcu_read_unlock();
336 
337 	return ret;
338 }
339 
340 /* Extract dsfield from inner protocol */
341 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
342 				       const struct sk_buff *skb)
343 {
344 	if (skb->protocol == htons(ETH_P_IP))
345 		return iph->tos;
346 	else if (skb->protocol == htons(ETH_P_IPV6))
347 		return ipv6_get_dsfield((const struct ipv6hdr *)iph);
348 	else
349 		return 0;
350 }
351 
352 /* Propogate ECN bits out */
353 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
354 				     const struct sk_buff *skb)
355 {
356 	u8 inner = ip_tunnel_get_dsfield(iph, skb);
357 
358 	return INET_ECN_encapsulate(tos, inner);
359 }
360 
361 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
362 			   __be16 inner_proto, bool raw_proto, bool xnet);
363 
364 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
365 				       __be16 inner_proto, bool xnet)
366 {
367 	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
368 }
369 
370 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
371 		   __be32 src, __be32 dst, u8 proto,
372 		   u8 tos, u8 ttl, __be16 df, bool xnet);
373 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
374 					     gfp_t flags);
375 
376 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
377 
378 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
379 {
380 	if (skb_is_gso(skb)) {
381 		int err;
382 
383 		err = skb_unclone(skb, GFP_ATOMIC);
384 		if (unlikely(err))
385 			return err;
386 		skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
387 					       NETIF_F_GSO_SHIFT);
388 	}
389 
390 	skb->encapsulation = 0;
391 	return 0;
392 }
393 
394 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
395 {
396 	if (pkt_len > 0) {
397 		struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
398 
399 		u64_stats_update_begin(&tstats->syncp);
400 		tstats->tx_bytes += pkt_len;
401 		tstats->tx_packets++;
402 		u64_stats_update_end(&tstats->syncp);
403 		put_cpu_ptr(tstats);
404 	} else {
405 		struct net_device_stats *err_stats = &dev->stats;
406 
407 		if (pkt_len < 0) {
408 			err_stats->tx_errors++;
409 			err_stats->tx_aborted_errors++;
410 		} else {
411 			err_stats->tx_dropped++;
412 		}
413 	}
414 }
415 
416 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
417 {
418 	return info + 1;
419 }
420 
421 static inline void ip_tunnel_info_opts_get(void *to,
422 					   const struct ip_tunnel_info *info)
423 {
424 	memcpy(to, info + 1, info->options_len);
425 }
426 
427 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
428 					   const void *from, int len)
429 {
430 	memcpy(ip_tunnel_info_opts(info), from, len);
431 	info->options_len = len;
432 }
433 
434 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
435 {
436 	return (struct ip_tunnel_info *)lwtstate->data;
437 }
438 
439 extern struct static_key ip_tunnel_metadata_cnt;
440 
441 /* Returns > 0 if metadata should be collected */
442 static inline int ip_tunnel_collect_metadata(void)
443 {
444 	return static_key_false(&ip_tunnel_metadata_cnt);
445 }
446 
447 void __init ip_tunnel_core_init(void);
448 
449 void ip_tunnel_need_metadata(void);
450 void ip_tunnel_unneed_metadata(void);
451 
452 #else /* CONFIG_INET */
453 
454 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
455 {
456 	return NULL;
457 }
458 
459 static inline void ip_tunnel_need_metadata(void)
460 {
461 }
462 
463 static inline void ip_tunnel_unneed_metadata(void)
464 {
465 }
466 
467 static inline void ip_tunnel_info_opts_get(void *to,
468 					   const struct ip_tunnel_info *info)
469 {
470 }
471 
472 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
473 					   const void *from, int len)
474 {
475 	info->options_len = 0;
476 }
477 
478 #endif /* CONFIG_INET */
479 
480 #endif /* __NET_IP_TUNNELS_H */
481