xref: /openbmc/linux/include/net/ip_tunnels.h (revision e3b9f1e8)
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 	/* 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 #define TUNNEL_CSUM		__cpu_to_be16(0x01)
148 #define TUNNEL_ROUTING		__cpu_to_be16(0x02)
149 #define TUNNEL_KEY		__cpu_to_be16(0x04)
150 #define TUNNEL_SEQ		__cpu_to_be16(0x08)
151 #define TUNNEL_STRICT		__cpu_to_be16(0x10)
152 #define TUNNEL_REC		__cpu_to_be16(0x20)
153 #define TUNNEL_VERSION		__cpu_to_be16(0x40)
154 #define TUNNEL_NO_KEY		__cpu_to_be16(0x80)
155 #define TUNNEL_DONT_FRAGMENT    __cpu_to_be16(0x0100)
156 #define TUNNEL_OAM		__cpu_to_be16(0x0200)
157 #define TUNNEL_CRIT_OPT		__cpu_to_be16(0x0400)
158 #define TUNNEL_GENEVE_OPT	__cpu_to_be16(0x0800)
159 #define TUNNEL_VXLAN_OPT	__cpu_to_be16(0x1000)
160 #define TUNNEL_NOCACHE		__cpu_to_be16(0x2000)
161 #define TUNNEL_ERSPAN_OPT	__cpu_to_be16(0x4000)
162 
163 #define TUNNEL_OPTIONS_PRESENT \
164 		(TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT | TUNNEL_ERSPAN_OPT)
165 
166 struct tnl_ptk_info {
167 	__be16 flags;
168 	__be16 proto;
169 	__be32 key;
170 	__be32 seq;
171 	int hdr_len;
172 };
173 
174 #define PACKET_RCVD	0
175 #define PACKET_REJECT	1
176 #define PACKET_NEXT	2
177 
178 #define IP_TNL_HASH_BITS   7
179 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
180 
181 struct ip_tunnel_net {
182 	struct net_device *fb_tunnel_dev;
183 	struct hlist_head tunnels[IP_TNL_HASH_SIZE];
184 	struct ip_tunnel __rcu *collect_md_tun;
185 };
186 
187 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
188 				      __be32 saddr, __be32 daddr,
189 				      u8 tos, u8 ttl, __be32 label,
190 				      __be16 tp_src, __be16 tp_dst,
191 				      __be64 tun_id, __be16 tun_flags)
192 {
193 	key->tun_id = tun_id;
194 	key->u.ipv4.src = saddr;
195 	key->u.ipv4.dst = daddr;
196 	memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
197 	       0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
198 	key->tos = tos;
199 	key->ttl = ttl;
200 	key->label = label;
201 	key->tun_flags = tun_flags;
202 
203 	/* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
204 	 * the upper tunnel are used.
205 	 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
206 	 */
207 	key->tp_src = tp_src;
208 	key->tp_dst = tp_dst;
209 
210 	/* Clear struct padding. */
211 	if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
212 		memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
213 		       0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
214 }
215 
216 static inline bool
217 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
218 			   const struct ip_tunnel_info *info)
219 {
220 	if (skb->mark)
221 		return false;
222 	if (!info)
223 		return true;
224 	if (info->key.tun_flags & TUNNEL_NOCACHE)
225 		return false;
226 
227 	return true;
228 }
229 
230 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
231 					       *tun_info)
232 {
233 	return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
234 }
235 
236 static inline __be64 key32_to_tunnel_id(__be32 key)
237 {
238 #ifdef __BIG_ENDIAN
239 	return (__force __be64)key;
240 #else
241 	return (__force __be64)((__force u64)key << 32);
242 #endif
243 }
244 
245 /* Returns the least-significant 32 bits of a __be64. */
246 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
247 {
248 #ifdef __BIG_ENDIAN
249 	return (__force __be32)tun_id;
250 #else
251 	return (__force __be32)((__force u64)tun_id >> 32);
252 #endif
253 }
254 
255 #ifdef CONFIG_INET
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);
272 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
273 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
274 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
275 
276 void ip_tunnel_get_stats64(struct net_device *dev,
277 			   struct rtnl_link_stats64 *tot);
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 struct ip_tunnel_encap_ops {
293 	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
294 	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
295 			    u8 *protocol, struct flowi4 *fl4);
296 };
297 
298 #define MAX_IPTUN_ENCAP_OPS 8
299 
300 extern const struct ip_tunnel_encap_ops __rcu *
301 		iptun_encaps[MAX_IPTUN_ENCAP_OPS];
302 
303 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
304 			    unsigned int num);
305 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
306 			    unsigned int num);
307 
308 int ip_tunnel_encap_setup(struct ip_tunnel *t,
309 			  struct ip_tunnel_encap *ipencap);
310 
311 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
312 {
313 	const struct ip_tunnel_encap_ops *ops;
314 	int hlen = -EINVAL;
315 
316 	if (e->type == TUNNEL_ENCAP_NONE)
317 		return 0;
318 
319 	if (e->type >= MAX_IPTUN_ENCAP_OPS)
320 		return -EINVAL;
321 
322 	rcu_read_lock();
323 	ops = rcu_dereference(iptun_encaps[e->type]);
324 	if (likely(ops && ops->encap_hlen))
325 		hlen = ops->encap_hlen(e);
326 	rcu_read_unlock();
327 
328 	return hlen;
329 }
330 
331 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
332 				  u8 *protocol, struct flowi4 *fl4)
333 {
334 	const struct ip_tunnel_encap_ops *ops;
335 	int ret = -EINVAL;
336 
337 	if (t->encap.type == TUNNEL_ENCAP_NONE)
338 		return 0;
339 
340 	if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
341 		return -EINVAL;
342 
343 	rcu_read_lock();
344 	ops = rcu_dereference(iptun_encaps[t->encap.type]);
345 	if (likely(ops && ops->build_header))
346 		ret = ops->build_header(skb, &t->encap, protocol, fl4);
347 	rcu_read_unlock();
348 
349 	return ret;
350 }
351 
352 /* Extract dsfield from inner protocol */
353 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
354 				       const struct sk_buff *skb)
355 {
356 	if (skb->protocol == htons(ETH_P_IP))
357 		return iph->tos;
358 	else if (skb->protocol == htons(ETH_P_IPV6))
359 		return ipv6_get_dsfield((const struct ipv6hdr *)iph);
360 	else
361 		return 0;
362 }
363 
364 /* Propogate ECN bits out */
365 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
366 				     const struct sk_buff *skb)
367 {
368 	u8 inner = ip_tunnel_get_dsfield(iph, skb);
369 
370 	return INET_ECN_encapsulate(tos, inner);
371 }
372 
373 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
374 			   __be16 inner_proto, bool raw_proto, bool xnet);
375 
376 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
377 				       __be16 inner_proto, bool xnet)
378 {
379 	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
380 }
381 
382 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
383 		   __be32 src, __be32 dst, u8 proto,
384 		   u8 tos, u8 ttl, __be16 df, bool xnet);
385 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
386 					     gfp_t flags);
387 
388 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
389 
390 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
391 {
392 	if (skb_is_gso(skb)) {
393 		int err;
394 
395 		err = skb_unclone(skb, GFP_ATOMIC);
396 		if (unlikely(err))
397 			return err;
398 		skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
399 					       NETIF_F_GSO_SHIFT);
400 	}
401 
402 	skb->encapsulation = 0;
403 	return 0;
404 }
405 
406 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
407 {
408 	if (pkt_len > 0) {
409 		struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
410 
411 		u64_stats_update_begin(&tstats->syncp);
412 		tstats->tx_bytes += pkt_len;
413 		tstats->tx_packets++;
414 		u64_stats_update_end(&tstats->syncp);
415 		put_cpu_ptr(tstats);
416 	} else {
417 		struct net_device_stats *err_stats = &dev->stats;
418 
419 		if (pkt_len < 0) {
420 			err_stats->tx_errors++;
421 			err_stats->tx_aborted_errors++;
422 		} else {
423 			err_stats->tx_dropped++;
424 		}
425 	}
426 }
427 
428 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
429 {
430 	return info + 1;
431 }
432 
433 static inline void ip_tunnel_info_opts_get(void *to,
434 					   const struct ip_tunnel_info *info)
435 {
436 	memcpy(to, info + 1, info->options_len);
437 }
438 
439 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
440 					   const void *from, int len)
441 {
442 	memcpy(ip_tunnel_info_opts(info), from, len);
443 	info->options_len = len;
444 }
445 
446 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
447 {
448 	return (struct ip_tunnel_info *)lwtstate->data;
449 }
450 
451 extern struct static_key ip_tunnel_metadata_cnt;
452 
453 /* Returns > 0 if metadata should be collected */
454 static inline int ip_tunnel_collect_metadata(void)
455 {
456 	return static_key_false(&ip_tunnel_metadata_cnt);
457 }
458 
459 void __init ip_tunnel_core_init(void);
460 
461 void ip_tunnel_need_metadata(void);
462 void ip_tunnel_unneed_metadata(void);
463 
464 #else /* CONFIG_INET */
465 
466 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
467 {
468 	return NULL;
469 }
470 
471 static inline void ip_tunnel_need_metadata(void)
472 {
473 }
474 
475 static inline void ip_tunnel_unneed_metadata(void)
476 {
477 }
478 
479 static inline void ip_tunnel_info_opts_get(void *to,
480 					   const struct ip_tunnel_info *info)
481 {
482 }
483 
484 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
485 					   const void *from, int len)
486 {
487 	info->options_len = 0;
488 }
489 
490 #endif /* CONFIG_INET */
491 
492 #endif /* __NET_IP_TUNNELS_H */
493