xref: /openbmc/linux/include/net/ip_tunnels.h (revision 42bc47b3)
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 rtnl_link_ops *rtnl_link_ops;
184 	struct hlist_head tunnels[IP_TNL_HASH_SIZE];
185 	struct ip_tunnel __rcu *collect_md_tun;
186 	int type;
187 };
188 
189 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
190 				      __be32 saddr, __be32 daddr,
191 				      u8 tos, u8 ttl, __be32 label,
192 				      __be16 tp_src, __be16 tp_dst,
193 				      __be64 tun_id, __be16 tun_flags)
194 {
195 	key->tun_id = tun_id;
196 	key->u.ipv4.src = saddr;
197 	key->u.ipv4.dst = daddr;
198 	memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
199 	       0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
200 	key->tos = tos;
201 	key->ttl = ttl;
202 	key->label = label;
203 	key->tun_flags = tun_flags;
204 
205 	/* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
206 	 * the upper tunnel are used.
207 	 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
208 	 */
209 	key->tp_src = tp_src;
210 	key->tp_dst = tp_dst;
211 
212 	/* Clear struct padding. */
213 	if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
214 		memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
215 		       0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
216 }
217 
218 static inline bool
219 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
220 			   const struct ip_tunnel_info *info)
221 {
222 	if (skb->mark)
223 		return false;
224 	if (!info)
225 		return true;
226 	if (info->key.tun_flags & TUNNEL_NOCACHE)
227 		return false;
228 
229 	return true;
230 }
231 
232 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
233 					       *tun_info)
234 {
235 	return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
236 }
237 
238 static inline __be64 key32_to_tunnel_id(__be32 key)
239 {
240 #ifdef __BIG_ENDIAN
241 	return (__force __be64)key;
242 #else
243 	return (__force __be64)((__force u64)key << 32);
244 #endif
245 }
246 
247 /* Returns the least-significant 32 bits of a __be64. */
248 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
249 {
250 #ifdef __BIG_ENDIAN
251 	return (__force __be32)tun_id;
252 #else
253 	return (__force __be32)((__force u64)tun_id >> 32);
254 #endif
255 }
256 
257 #ifdef CONFIG_INET
258 
259 static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
260 				       int proto,
261 				       __be32 daddr, __be32 saddr,
262 				       __be32 key, __u8 tos, int oif,
263 				       __u32 mark)
264 {
265 	memset(fl4, 0, sizeof(*fl4));
266 	fl4->flowi4_oif = oif;
267 	fl4->daddr = daddr;
268 	fl4->saddr = saddr;
269 	fl4->flowi4_tos = tos;
270 	fl4->flowi4_proto = proto;
271 	fl4->fl4_gre_key = key;
272 	fl4->flowi4_mark = mark;
273 }
274 
275 int ip_tunnel_init(struct net_device *dev);
276 void ip_tunnel_uninit(struct net_device *dev);
277 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
278 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
279 int ip_tunnel_get_iflink(const struct net_device *dev);
280 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
281 		       struct rtnl_link_ops *ops, char *devname);
282 
283 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
284 			   struct rtnl_link_ops *ops);
285 
286 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
287 		    const struct iphdr *tnl_params, const u8 protocol);
288 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
289 		       const u8 proto);
290 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
291 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
292 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
293 
294 void ip_tunnel_get_stats64(struct net_device *dev,
295 			   struct rtnl_link_stats64 *tot);
296 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
297 				   int link, __be16 flags,
298 				   __be32 remote, __be32 local,
299 				   __be32 key);
300 
301 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
302 		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
303 		  bool log_ecn_error);
304 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
305 			 struct ip_tunnel_parm *p, __u32 fwmark);
306 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
307 		      struct ip_tunnel_parm *p, __u32 fwmark);
308 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
309 
310 struct ip_tunnel_encap_ops {
311 	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
312 	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
313 			    u8 *protocol, struct flowi4 *fl4);
314 };
315 
316 #define MAX_IPTUN_ENCAP_OPS 8
317 
318 extern const struct ip_tunnel_encap_ops __rcu *
319 		iptun_encaps[MAX_IPTUN_ENCAP_OPS];
320 
321 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
322 			    unsigned int num);
323 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
324 			    unsigned int num);
325 
326 int ip_tunnel_encap_setup(struct ip_tunnel *t,
327 			  struct ip_tunnel_encap *ipencap);
328 
329 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
330 {
331 	const struct ip_tunnel_encap_ops *ops;
332 	int hlen = -EINVAL;
333 
334 	if (e->type == TUNNEL_ENCAP_NONE)
335 		return 0;
336 
337 	if (e->type >= MAX_IPTUN_ENCAP_OPS)
338 		return -EINVAL;
339 
340 	rcu_read_lock();
341 	ops = rcu_dereference(iptun_encaps[e->type]);
342 	if (likely(ops && ops->encap_hlen))
343 		hlen = ops->encap_hlen(e);
344 	rcu_read_unlock();
345 
346 	return hlen;
347 }
348 
349 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
350 				  u8 *protocol, struct flowi4 *fl4)
351 {
352 	const struct ip_tunnel_encap_ops *ops;
353 	int ret = -EINVAL;
354 
355 	if (t->encap.type == TUNNEL_ENCAP_NONE)
356 		return 0;
357 
358 	if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
359 		return -EINVAL;
360 
361 	rcu_read_lock();
362 	ops = rcu_dereference(iptun_encaps[t->encap.type]);
363 	if (likely(ops && ops->build_header))
364 		ret = ops->build_header(skb, &t->encap, protocol, fl4);
365 	rcu_read_unlock();
366 
367 	return ret;
368 }
369 
370 /* Extract dsfield from inner protocol */
371 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
372 				       const struct sk_buff *skb)
373 {
374 	if (skb->protocol == htons(ETH_P_IP))
375 		return iph->tos;
376 	else if (skb->protocol == htons(ETH_P_IPV6))
377 		return ipv6_get_dsfield((const struct ipv6hdr *)iph);
378 	else
379 		return 0;
380 }
381 
382 static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
383 				       const struct sk_buff *skb)
384 {
385 	if (skb->protocol == htons(ETH_P_IP))
386 		return iph->ttl;
387 	else if (skb->protocol == htons(ETH_P_IPV6))
388 		return ((const struct ipv6hdr *)iph)->hop_limit;
389 	else
390 		return 0;
391 }
392 
393 /* Propogate ECN bits out */
394 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
395 				     const struct sk_buff *skb)
396 {
397 	u8 inner = ip_tunnel_get_dsfield(iph, skb);
398 
399 	return INET_ECN_encapsulate(tos, inner);
400 }
401 
402 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
403 			   __be16 inner_proto, bool raw_proto, bool xnet);
404 
405 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
406 				       __be16 inner_proto, bool xnet)
407 {
408 	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
409 }
410 
411 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
412 		   __be32 src, __be32 dst, u8 proto,
413 		   u8 tos, u8 ttl, __be16 df, bool xnet);
414 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
415 					     gfp_t flags);
416 
417 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
418 
419 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
420 {
421 	if (skb_is_gso(skb)) {
422 		int err;
423 
424 		err = skb_unclone(skb, GFP_ATOMIC);
425 		if (unlikely(err))
426 			return err;
427 		skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
428 					       NETIF_F_GSO_SHIFT);
429 	}
430 
431 	skb->encapsulation = 0;
432 	return 0;
433 }
434 
435 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
436 {
437 	if (pkt_len > 0) {
438 		struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
439 
440 		u64_stats_update_begin(&tstats->syncp);
441 		tstats->tx_bytes += pkt_len;
442 		tstats->tx_packets++;
443 		u64_stats_update_end(&tstats->syncp);
444 		put_cpu_ptr(tstats);
445 	} else {
446 		struct net_device_stats *err_stats = &dev->stats;
447 
448 		if (pkt_len < 0) {
449 			err_stats->tx_errors++;
450 			err_stats->tx_aborted_errors++;
451 		} else {
452 			err_stats->tx_dropped++;
453 		}
454 	}
455 }
456 
457 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
458 {
459 	return info + 1;
460 }
461 
462 static inline void ip_tunnel_info_opts_get(void *to,
463 					   const struct ip_tunnel_info *info)
464 {
465 	memcpy(to, info + 1, info->options_len);
466 }
467 
468 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
469 					   const void *from, int len)
470 {
471 	memcpy(ip_tunnel_info_opts(info), from, len);
472 	info->options_len = len;
473 }
474 
475 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
476 {
477 	return (struct ip_tunnel_info *)lwtstate->data;
478 }
479 
480 DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
481 
482 /* Returns > 0 if metadata should be collected */
483 static inline int ip_tunnel_collect_metadata(void)
484 {
485 	return static_branch_unlikely(&ip_tunnel_metadata_cnt);
486 }
487 
488 void __init ip_tunnel_core_init(void);
489 
490 void ip_tunnel_need_metadata(void);
491 void ip_tunnel_unneed_metadata(void);
492 
493 #else /* CONFIG_INET */
494 
495 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
496 {
497 	return NULL;
498 }
499 
500 static inline void ip_tunnel_need_metadata(void)
501 {
502 }
503 
504 static inline void ip_tunnel_unneed_metadata(void)
505 {
506 }
507 
508 static inline void ip_tunnel_info_opts_get(void *to,
509 					   const struct ip_tunnel_info *info)
510 {
511 }
512 
513 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
514 					   const void *from, int len)
515 {
516 	info->options_len = 0;
517 }
518 
519 #endif /* CONFIG_INET */
520 
521 #endif /* __NET_IP_TUNNELS_H */
522