xref: /linux/include/net/ip_tunnels.h (revision 1a9239bb4253f9076b5b4b2a1a4e8d7defd77a95)
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 #define __ipt_flag_op(op, ...)					\
40 	op(__VA_ARGS__, __IP_TUNNEL_FLAG_NUM)
41 
42 #define IP_TUNNEL_DECLARE_FLAGS(...)				\
43 	__ipt_flag_op(DECLARE_BITMAP, __VA_ARGS__)
44 
45 #define ip_tunnel_flags_zero(...)	__ipt_flag_op(bitmap_zero, __VA_ARGS__)
46 #define ip_tunnel_flags_copy(...)	__ipt_flag_op(bitmap_copy, __VA_ARGS__)
47 #define ip_tunnel_flags_and(...)	__ipt_flag_op(bitmap_and, __VA_ARGS__)
48 #define ip_tunnel_flags_or(...)		__ipt_flag_op(bitmap_or, __VA_ARGS__)
49 
50 #define ip_tunnel_flags_empty(...)				\
51 	__ipt_flag_op(bitmap_empty, __VA_ARGS__)
52 #define ip_tunnel_flags_intersect(...)				\
53 	__ipt_flag_op(bitmap_intersects, __VA_ARGS__)
54 #define ip_tunnel_flags_subset(...)				\
55 	__ipt_flag_op(bitmap_subset, __VA_ARGS__)
56 
57 struct ip_tunnel_key {
58 	__be64			tun_id;
59 	union {
60 		struct {
61 			__be32	src;
62 			__be32	dst;
63 		} ipv4;
64 		struct {
65 			struct in6_addr src;
66 			struct in6_addr dst;
67 		} ipv6;
68 	} u;
69 	IP_TUNNEL_DECLARE_FLAGS(tun_flags);
70 	__be32			label;		/* Flow Label for IPv6 */
71 	u32			nhid;
72 	u8			tos;		/* TOS for IPv4, TC for IPv6 */
73 	u8			ttl;		/* TTL for IPv4, HL for IPv6 */
74 	__be16			tp_src;
75 	__be16			tp_dst;
76 	__u8			flow_flags;
77 };
78 
79 struct ip_tunnel_encap {
80 	u16			type;
81 	u16			flags;
82 	__be16			sport;
83 	__be16			dport;
84 };
85 
86 /* Flags for ip_tunnel_info mode. */
87 #define IP_TUNNEL_INFO_TX	0x01	/* represents tx tunnel parameters */
88 #define IP_TUNNEL_INFO_IPV6	0x02	/* key contains IPv6 addresses */
89 #define IP_TUNNEL_INFO_BRIDGE	0x04	/* represents a bridged tunnel id */
90 
91 /* Maximum tunnel options length. */
92 #define IP_TUNNEL_OPTS_MAX					\
93 	GENMASK((sizeof_field(struct ip_tunnel_info,		\
94 			      options_len) * BITS_PER_BYTE) - 1, 0)
95 
96 #define ip_tunnel_info_opts(info)				\
97 	_Generic(info,						\
98 		 const struct ip_tunnel_info * : ((const void *)(info)->options),\
99 		 struct ip_tunnel_info * : ((void *)(info)->options)\
100 	)
101 
102 struct ip_tunnel_info {
103 	struct ip_tunnel_key	key;
104 	struct ip_tunnel_encap	encap;
105 #ifdef CONFIG_DST_CACHE
106 	struct dst_cache	dst_cache;
107 #endif
108 	u8			options_len;
109 	u8			mode;
110 	u8			options[] __aligned_largest __counted_by(options_len);
111 };
112 
113 /* 6rd prefix/relay information */
114 #ifdef CONFIG_IPV6_SIT_6RD
115 struct ip_tunnel_6rd_parm {
116 	struct in6_addr		prefix;
117 	__be32			relay_prefix;
118 	u16			prefixlen;
119 	u16			relay_prefixlen;
120 };
121 #endif
122 
123 struct ip_tunnel_prl_entry {
124 	struct ip_tunnel_prl_entry __rcu *next;
125 	__be32				addr;
126 	u16				flags;
127 	struct rcu_head			rcu_head;
128 };
129 
130 struct metadata_dst;
131 
132 /* Kernel-side variant of ip_tunnel_parm */
133 struct ip_tunnel_parm_kern {
134 	char			name[IFNAMSIZ];
135 	IP_TUNNEL_DECLARE_FLAGS(i_flags);
136 	IP_TUNNEL_DECLARE_FLAGS(o_flags);
137 	__be32			i_key;
138 	__be32			o_key;
139 	int			link;
140 	struct iphdr		iph;
141 };
142 
143 struct ip_tunnel {
144 	struct ip_tunnel __rcu	*next;
145 	struct hlist_node hash_node;
146 
147 	struct net_device	*dev;
148 	netdevice_tracker	dev_tracker;
149 
150 	struct net		*net;	/* netns for packet i/o */
151 
152 	unsigned long	err_time;	/* Time when the last ICMP error
153 					 * arrived */
154 	int		err_count;	/* Number of arrived ICMP errors */
155 
156 	/* These four fields used only by GRE */
157 	u32		i_seqno;	/* The last seen seqno	*/
158 	atomic_t	o_seqno;	/* The last output seqno */
159 	int		tun_hlen;	/* Precalculated header length */
160 
161 	/* These four fields used only by ERSPAN */
162 	u32		index;		/* ERSPAN type II index */
163 	u8		erspan_ver;	/* ERSPAN version */
164 	u8		dir;		/* ERSPAN direction */
165 	u16		hwid;		/* ERSPAN hardware ID */
166 
167 	struct dst_cache dst_cache;
168 
169 	struct ip_tunnel_parm_kern parms;
170 
171 	int		mlink;
172 	int		encap_hlen;	/* Encap header length (FOU,GUE) */
173 	int		hlen;		/* tun_hlen + encap_hlen */
174 	struct ip_tunnel_encap encap;
175 
176 	/* for SIT */
177 #ifdef CONFIG_IPV6_SIT_6RD
178 	struct ip_tunnel_6rd_parm ip6rd;
179 #endif
180 	struct ip_tunnel_prl_entry __rcu *prl;	/* potential router list */
181 	unsigned int		prl_count;	/* # of entries in PRL */
182 	unsigned int		ip_tnl_net_id;
183 	struct gro_cells	gro_cells;
184 	__u32			fwmark;
185 	bool			collect_md;
186 	bool			ignore_df;
187 };
188 
189 struct tnl_ptk_info {
190 	IP_TUNNEL_DECLARE_FLAGS(flags);
191 	__be16 proto;
192 	__be32 key;
193 	__be32 seq;
194 	int hdr_len;
195 };
196 
197 #define PACKET_RCVD	0
198 #define PACKET_REJECT	1
199 #define PACKET_NEXT	2
200 
201 #define IP_TNL_HASH_BITS   7
202 #define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
203 
204 struct ip_tunnel_net {
205 	struct net_device *fb_tunnel_dev;
206 	struct rtnl_link_ops *rtnl_link_ops;
207 	struct hlist_head tunnels[IP_TNL_HASH_SIZE];
208 	struct ip_tunnel __rcu *collect_md_tun;
209 	int type;
210 };
211 
ip_tunnel_set_options_present(unsigned long * flags)212 static inline void ip_tunnel_set_options_present(unsigned long *flags)
213 {
214 	IP_TUNNEL_DECLARE_FLAGS(present) = { };
215 
216 	__set_bit(IP_TUNNEL_GENEVE_OPT_BIT, present);
217 	__set_bit(IP_TUNNEL_VXLAN_OPT_BIT, present);
218 	__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, present);
219 	__set_bit(IP_TUNNEL_GTP_OPT_BIT, present);
220 	__set_bit(IP_TUNNEL_PFCP_OPT_BIT, present);
221 
222 	ip_tunnel_flags_or(flags, flags, present);
223 }
224 
ip_tunnel_clear_options_present(unsigned long * flags)225 static inline void ip_tunnel_clear_options_present(unsigned long *flags)
226 {
227 	IP_TUNNEL_DECLARE_FLAGS(present) = { };
228 
229 	__set_bit(IP_TUNNEL_GENEVE_OPT_BIT, present);
230 	__set_bit(IP_TUNNEL_VXLAN_OPT_BIT, present);
231 	__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, present);
232 	__set_bit(IP_TUNNEL_GTP_OPT_BIT, present);
233 	__set_bit(IP_TUNNEL_PFCP_OPT_BIT, present);
234 
235 	__ipt_flag_op(bitmap_andnot, flags, flags, present);
236 }
237 
ip_tunnel_is_options_present(const unsigned long * flags)238 static inline bool ip_tunnel_is_options_present(const unsigned long *flags)
239 {
240 	IP_TUNNEL_DECLARE_FLAGS(present) = { };
241 
242 	__set_bit(IP_TUNNEL_GENEVE_OPT_BIT, present);
243 	__set_bit(IP_TUNNEL_VXLAN_OPT_BIT, present);
244 	__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, present);
245 	__set_bit(IP_TUNNEL_GTP_OPT_BIT, present);
246 	__set_bit(IP_TUNNEL_PFCP_OPT_BIT, present);
247 
248 	return ip_tunnel_flags_intersect(flags, present);
249 }
250 
ip_tunnel_flags_is_be16_compat(const unsigned long * flags)251 static inline bool ip_tunnel_flags_is_be16_compat(const unsigned long *flags)
252 {
253 	IP_TUNNEL_DECLARE_FLAGS(supp) = { };
254 
255 	bitmap_set(supp, 0, BITS_PER_TYPE(__be16));
256 	__set_bit(IP_TUNNEL_VTI_BIT, supp);
257 
258 	return ip_tunnel_flags_subset(flags, supp);
259 }
260 
ip_tunnel_flags_from_be16(unsigned long * dst,__be16 flags)261 static inline void ip_tunnel_flags_from_be16(unsigned long *dst, __be16 flags)
262 {
263 	ip_tunnel_flags_zero(dst);
264 
265 	bitmap_write(dst, be16_to_cpu(flags), 0, BITS_PER_TYPE(__be16));
266 	__assign_bit(IP_TUNNEL_VTI_BIT, dst, flags & VTI_ISVTI);
267 }
268 
ip_tunnel_flags_to_be16(const unsigned long * flags)269 static inline __be16 ip_tunnel_flags_to_be16(const unsigned long *flags)
270 {
271 	__be16 ret;
272 
273 	ret = cpu_to_be16(bitmap_read(flags, 0, BITS_PER_TYPE(__be16)));
274 	if (test_bit(IP_TUNNEL_VTI_BIT, flags))
275 		ret |= VTI_ISVTI;
276 
277 	return ret;
278 }
279 
ip_tunnel_key_init(struct ip_tunnel_key * key,__be32 saddr,__be32 daddr,u8 tos,u8 ttl,__be32 label,__be16 tp_src,__be16 tp_dst,__be64 tun_id,const unsigned long * tun_flags)280 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
281 				      __be32 saddr, __be32 daddr,
282 				      u8 tos, u8 ttl, __be32 label,
283 				      __be16 tp_src, __be16 tp_dst,
284 				      __be64 tun_id,
285 				      const unsigned long *tun_flags)
286 {
287 	key->tun_id = tun_id;
288 	key->u.ipv4.src = saddr;
289 	key->u.ipv4.dst = daddr;
290 	memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
291 	       0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
292 	key->tos = tos;
293 	key->ttl = ttl;
294 	key->label = label;
295 	ip_tunnel_flags_copy(key->tun_flags, tun_flags);
296 
297 	/* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
298 	 * the upper tunnel are used.
299 	 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
300 	 */
301 	key->tp_src = tp_src;
302 	key->tp_dst = tp_dst;
303 
304 	/* Clear struct padding. */
305 	if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
306 		memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
307 		       0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
308 }
309 
310 static inline bool
ip_tunnel_dst_cache_usable(const struct sk_buff * skb,const struct ip_tunnel_info * info)311 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
312 			   const struct ip_tunnel_info *info)
313 {
314 	if (skb->mark)
315 		return false;
316 
317 	return !info || !test_bit(IP_TUNNEL_NOCACHE_BIT, info->key.tun_flags);
318 }
319 
ip_tunnel_info_af(const struct ip_tunnel_info * tun_info)320 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
321 					       *tun_info)
322 {
323 	return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
324 }
325 
key32_to_tunnel_id(__be32 key)326 static inline __be64 key32_to_tunnel_id(__be32 key)
327 {
328 #ifdef __BIG_ENDIAN
329 	return (__force __be64)key;
330 #else
331 	return (__force __be64)((__force u64)key << 32);
332 #endif
333 }
334 
335 /* Returns the least-significant 32 bits of a __be64. */
tunnel_id_to_key32(__be64 tun_id)336 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
337 {
338 #ifdef __BIG_ENDIAN
339 	return (__force __be32)tun_id;
340 #else
341 	return (__force __be32)((__force u64)tun_id >> 32);
342 #endif
343 }
344 
345 #ifdef CONFIG_INET
346 
ip_tunnel_init_flow(struct flowi4 * fl4,int proto,__be32 daddr,__be32 saddr,__be32 key,__u8 tos,struct net * net,int oif,__u32 mark,__u32 tun_inner_hash,__u8 flow_flags)347 static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
348 				       int proto,
349 				       __be32 daddr, __be32 saddr,
350 				       __be32 key, __u8 tos,
351 				       struct net *net, int oif,
352 				       __u32 mark, __u32 tun_inner_hash,
353 				       __u8 flow_flags)
354 {
355 	memset(fl4, 0, sizeof(*fl4));
356 
357 	if (oif) {
358 		fl4->flowi4_l3mdev = l3mdev_master_upper_ifindex_by_index(net, oif);
359 		/* Legacy VRF/l3mdev use case */
360 		fl4->flowi4_oif = fl4->flowi4_l3mdev ? 0 : oif;
361 	}
362 
363 	fl4->daddr = daddr;
364 	fl4->saddr = saddr;
365 	fl4->flowi4_tos = tos;
366 	fl4->flowi4_proto = proto;
367 	fl4->fl4_gre_key = key;
368 	fl4->flowi4_mark = mark;
369 	fl4->flowi4_multipath_hash = tun_inner_hash;
370 	fl4->flowi4_flags = flow_flags;
371 }
372 
373 int ip_tunnel_init(struct net_device *dev);
374 void ip_tunnel_uninit(struct net_device *dev);
375 void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
376 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
377 int ip_tunnel_get_iflink(const struct net_device *dev);
378 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
379 		       struct rtnl_link_ops *ops, char *devname);
380 
381 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
382 			   struct rtnl_link_ops *ops,
383 			   struct list_head *dev_to_kill);
384 
385 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
386 		    const struct iphdr *tnl_params, const u8 protocol);
387 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
388 		       const u8 proto, int tunnel_hlen);
389 int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm_kern *p,
390 		  int cmd);
391 bool ip_tunnel_parm_from_user(struct ip_tunnel_parm_kern *kp,
392 			      const void __user *data);
393 bool ip_tunnel_parm_to_user(void __user *data, struct ip_tunnel_parm_kern *kp);
394 int ip_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
395 			     void __user *data, int cmd);
396 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
397 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
398 
399 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
400 				   int link, const unsigned long *flags,
401 				   __be32 remote, __be32 local,
402 				   __be32 key);
403 
404 void ip_tunnel_md_udp_encap(struct sk_buff *skb, struct ip_tunnel_info *info);
405 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
406 		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
407 		  bool log_ecn_error);
408 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
409 			 struct ip_tunnel_parm_kern *p, __u32 fwmark);
410 int ip_tunnel_newlink(struct net *net, struct net_device *dev,
411 		      struct nlattr *tb[], struct ip_tunnel_parm_kern *p,
412 		      __u32 fwmark);
413 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
414 
415 bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
416 				   struct ip_tunnel_encap *encap);
417 
418 void ip_tunnel_netlink_parms(struct nlattr *data[],
419 			     struct ip_tunnel_parm_kern *parms);
420 
421 extern const struct header_ops ip_tunnel_header_ops;
422 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb);
423 
424 struct ip_tunnel_encap_ops {
425 	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
426 	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
427 			    u8 *protocol, struct flowi4 *fl4);
428 	int (*err_handler)(struct sk_buff *skb, u32 info);
429 };
430 
431 #define MAX_IPTUN_ENCAP_OPS 8
432 
433 extern const struct ip_tunnel_encap_ops __rcu *
434 		iptun_encaps[MAX_IPTUN_ENCAP_OPS];
435 
436 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
437 			    unsigned int num);
438 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
439 			    unsigned int num);
440 
441 int ip_tunnel_encap_setup(struct ip_tunnel *t,
442 			  struct ip_tunnel_encap *ipencap);
443 
444 static inline enum skb_drop_reason
pskb_inet_may_pull_reason(struct sk_buff * skb)445 pskb_inet_may_pull_reason(struct sk_buff *skb)
446 {
447 	int nhlen;
448 
449 	switch (skb->protocol) {
450 #if IS_ENABLED(CONFIG_IPV6)
451 	case htons(ETH_P_IPV6):
452 		nhlen = sizeof(struct ipv6hdr);
453 		break;
454 #endif
455 	case htons(ETH_P_IP):
456 		nhlen = sizeof(struct iphdr);
457 		break;
458 	default:
459 		nhlen = 0;
460 	}
461 
462 	return pskb_network_may_pull_reason(skb, nhlen);
463 }
464 
pskb_inet_may_pull(struct sk_buff * skb)465 static inline bool pskb_inet_may_pull(struct sk_buff *skb)
466 {
467 	return pskb_inet_may_pull_reason(skb) == SKB_NOT_DROPPED_YET;
468 }
469 
470 /* Variant of pskb_inet_may_pull().
471  */
472 static inline enum skb_drop_reason
skb_vlan_inet_prepare(struct sk_buff * skb,bool inner_proto_inherit)473 skb_vlan_inet_prepare(struct sk_buff *skb, bool inner_proto_inherit)
474 {
475 	int nhlen = 0, maclen = inner_proto_inherit ? 0 : ETH_HLEN;
476 	__be16 type = skb->protocol;
477 	enum skb_drop_reason reason;
478 
479 	/* Essentially this is skb_protocol(skb, true)
480 	 * And we get MAC len.
481 	 */
482 	if (eth_type_vlan(type))
483 		type = __vlan_get_protocol(skb, type, &maclen);
484 
485 	switch (type) {
486 #if IS_ENABLED(CONFIG_IPV6)
487 	case htons(ETH_P_IPV6):
488 		nhlen = sizeof(struct ipv6hdr);
489 		break;
490 #endif
491 	case htons(ETH_P_IP):
492 		nhlen = sizeof(struct iphdr);
493 		break;
494 	}
495 	/* For ETH_P_IPV6/ETH_P_IP we make sure to pull
496 	 * a base network header in skb->head.
497 	 */
498 	reason = pskb_may_pull_reason(skb, maclen + nhlen);
499 	if (reason)
500 		return reason;
501 
502 	skb_set_network_header(skb, maclen);
503 
504 	return SKB_NOT_DROPPED_YET;
505 }
506 
ip_encap_hlen(struct ip_tunnel_encap * e)507 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
508 {
509 	const struct ip_tunnel_encap_ops *ops;
510 	int hlen = -EINVAL;
511 
512 	if (e->type == TUNNEL_ENCAP_NONE)
513 		return 0;
514 
515 	if (e->type >= MAX_IPTUN_ENCAP_OPS)
516 		return -EINVAL;
517 
518 	rcu_read_lock();
519 	ops = rcu_dereference(iptun_encaps[e->type]);
520 	if (likely(ops && ops->encap_hlen))
521 		hlen = ops->encap_hlen(e);
522 	rcu_read_unlock();
523 
524 	return hlen;
525 }
526 
ip_tunnel_encap(struct sk_buff * skb,struct ip_tunnel_encap * e,u8 * protocol,struct flowi4 * fl4)527 static inline int ip_tunnel_encap(struct sk_buff *skb,
528 				  struct ip_tunnel_encap *e,
529 				  u8 *protocol, struct flowi4 *fl4)
530 {
531 	const struct ip_tunnel_encap_ops *ops;
532 	int ret = -EINVAL;
533 
534 	if (e->type == TUNNEL_ENCAP_NONE)
535 		return 0;
536 
537 	if (e->type >= MAX_IPTUN_ENCAP_OPS)
538 		return -EINVAL;
539 
540 	rcu_read_lock();
541 	ops = rcu_dereference(iptun_encaps[e->type]);
542 	if (likely(ops && ops->build_header))
543 		ret = ops->build_header(skb, e, protocol, fl4);
544 	rcu_read_unlock();
545 
546 	return ret;
547 }
548 
549 /* Extract dsfield from inner protocol */
ip_tunnel_get_dsfield(const struct iphdr * iph,const struct sk_buff * skb)550 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
551 				       const struct sk_buff *skb)
552 {
553 	__be16 payload_protocol = skb_protocol(skb, true);
554 
555 	if (payload_protocol == htons(ETH_P_IP))
556 		return iph->tos;
557 	else if (payload_protocol == htons(ETH_P_IPV6))
558 		return ipv6_get_dsfield((const struct ipv6hdr *)iph);
559 	else
560 		return 0;
561 }
562 
ip_tunnel_get_flowlabel(const struct iphdr * iph,const struct sk_buff * skb)563 static inline __be32 ip_tunnel_get_flowlabel(const struct iphdr *iph,
564 					     const struct sk_buff *skb)
565 {
566 	__be16 payload_protocol = skb_protocol(skb, true);
567 
568 	if (payload_protocol == htons(ETH_P_IPV6))
569 		return ip6_flowlabel((const struct ipv6hdr *)iph);
570 	else
571 		return 0;
572 }
573 
ip_tunnel_get_ttl(const struct iphdr * iph,const struct sk_buff * skb)574 static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
575 				       const struct sk_buff *skb)
576 {
577 	__be16 payload_protocol = skb_protocol(skb, true);
578 
579 	if (payload_protocol == htons(ETH_P_IP))
580 		return iph->ttl;
581 	else if (payload_protocol == htons(ETH_P_IPV6))
582 		return ((const struct ipv6hdr *)iph)->hop_limit;
583 	else
584 		return 0;
585 }
586 
587 /* Propagate ECN bits out */
ip_tunnel_ecn_encap(u8 tos,const struct iphdr * iph,const struct sk_buff * skb)588 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
589 				     const struct sk_buff *skb)
590 {
591 	u8 inner = ip_tunnel_get_dsfield(iph, skb);
592 
593 	return INET_ECN_encapsulate(tos, inner);
594 }
595 
596 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
597 			   __be16 inner_proto, bool raw_proto, bool xnet);
598 
iptunnel_pull_header(struct sk_buff * skb,int hdr_len,__be16 inner_proto,bool xnet)599 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
600 				       __be16 inner_proto, bool xnet)
601 {
602 	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
603 }
604 
605 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
606 		   __be32 src, __be32 dst, u8 proto,
607 		   u8 tos, u8 ttl, __be16 df, bool xnet);
608 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
609 					     gfp_t flags);
610 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
611 			  int headroom, bool reply);
612 
613 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
614 
iptunnel_pull_offloads(struct sk_buff * skb)615 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
616 {
617 	if (skb_is_gso(skb)) {
618 		int err;
619 
620 		err = skb_unclone(skb, GFP_ATOMIC);
621 		if (unlikely(err))
622 			return err;
623 		skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
624 					       NETIF_F_GSO_SHIFT);
625 	}
626 
627 	skb->encapsulation = 0;
628 	return 0;
629 }
630 
iptunnel_xmit_stats(struct net_device * dev,int pkt_len)631 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
632 {
633 	if (pkt_len > 0) {
634 		struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
635 
636 		u64_stats_update_begin(&tstats->syncp);
637 		u64_stats_add(&tstats->tx_bytes, pkt_len);
638 		u64_stats_inc(&tstats->tx_packets);
639 		u64_stats_update_end(&tstats->syncp);
640 		put_cpu_ptr(tstats);
641 		return;
642 	}
643 
644 	if (pkt_len < 0) {
645 		DEV_STATS_INC(dev, tx_errors);
646 		DEV_STATS_INC(dev, tx_aborted_errors);
647 	} else {
648 		DEV_STATS_INC(dev, tx_dropped);
649 	}
650 }
651 
ip_tunnel_info_opts_get(void * to,const struct ip_tunnel_info * info)652 static inline void ip_tunnel_info_opts_get(void *to,
653 					   const struct ip_tunnel_info *info)
654 {
655 	memcpy(to, ip_tunnel_info_opts(info), info->options_len);
656 }
657 
ip_tunnel_info_opts_set(struct ip_tunnel_info * info,const void * from,int len,const unsigned long * flags)658 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
659 					   const void *from, int len,
660 					   const unsigned long *flags)
661 {
662 	info->options_len = len;
663 	if (len > 0) {
664 		memcpy(ip_tunnel_info_opts(info), from, len);
665 		ip_tunnel_flags_or(info->key.tun_flags, info->key.tun_flags,
666 				   flags);
667 	}
668 }
669 
lwt_tun_info(struct lwtunnel_state * lwtstate)670 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
671 {
672 	return (struct ip_tunnel_info *)lwtstate->data;
673 }
674 
675 DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
676 
677 /* Returns > 0 if metadata should be collected */
ip_tunnel_collect_metadata(void)678 static inline int ip_tunnel_collect_metadata(void)
679 {
680 	return static_branch_unlikely(&ip_tunnel_metadata_cnt);
681 }
682 
683 void __init ip_tunnel_core_init(void);
684 
685 void ip_tunnel_need_metadata(void);
686 void ip_tunnel_unneed_metadata(void);
687 
688 #else /* CONFIG_INET */
689 
lwt_tun_info(struct lwtunnel_state * lwtstate)690 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
691 {
692 	return NULL;
693 }
694 
ip_tunnel_need_metadata(void)695 static inline void ip_tunnel_need_metadata(void)
696 {
697 }
698 
ip_tunnel_unneed_metadata(void)699 static inline void ip_tunnel_unneed_metadata(void)
700 {
701 }
702 
ip_tunnel_info_opts_get(void * to,const struct ip_tunnel_info * info)703 static inline void ip_tunnel_info_opts_get(void *to,
704 					   const struct ip_tunnel_info *info)
705 {
706 }
707 
ip_tunnel_info_opts_set(struct ip_tunnel_info * info,const void * from,int len,const unsigned long * flags)708 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
709 					   const void *from, int len,
710 					   const unsigned long *flags)
711 {
712 	info->options_len = 0;
713 }
714 
715 #endif /* CONFIG_INET */
716 
717 #endif /* __NET_IP_TUNNELS_H */
718