1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __NET_IP_TUNNELS_H
3 #define __NET_IP_TUNNELS_H 1
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>
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>
21 #if IS_ENABLED(CONFIG_IPV6)
23 #include <net/ip6_fib.h>
24 #include <net/ip6_route.h>
27 /* Keep error state on tunnel for 30 sec */
28 #define IPTUNNEL_ERR_TIMEO (30*HZ)
30 /* Used to memset ip_tunnel padding. */
31 #define IP_TUNNEL_KEY_SIZE offsetofend(struct ip_tunnel_key, tp_dst)
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))
39 struct ip_tunnel_key {
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 */
61 struct ip_tunnel_encap {
68 /* Flags for ip_tunnel_info mode. */
69 #define IP_TUNNEL_INFO_TX 0x01 /* represents tx tunnel parameters */
70 #define IP_TUNNEL_INFO_IPV6 0x02 /* key contains IPv6 addresses */
71 #define IP_TUNNEL_INFO_BRIDGE 0x04 /* represents a bridged tunnel id */
73 /* Maximum tunnel options length. */
74 #define IP_TUNNEL_OPTS_MAX \
75 GENMASK((sizeof_field(struct ip_tunnel_info, \
76 options_len) * BITS_PER_BYTE) - 1, 0)
78 #define ip_tunnel_info_opts(info) \
80 const struct ip_tunnel_info * : ((const void *)((info) + 1)),\
81 struct ip_tunnel_info * : ((void *)((info) + 1))\
84 struct ip_tunnel_info {
85 struct ip_tunnel_key key;
86 struct ip_tunnel_encap encap;
87 #ifdef CONFIG_DST_CACHE
88 struct dst_cache dst_cache;
94 /* 6rd prefix/relay information */
95 #ifdef CONFIG_IPV6_SIT_6RD
96 struct ip_tunnel_6rd_parm {
97 struct in6_addr prefix;
104 struct ip_tunnel_prl_entry {
105 struct ip_tunnel_prl_entry __rcu *next;
108 struct rcu_head rcu_head;
114 struct ip_tunnel __rcu *next;
115 struct hlist_node hash_node;
117 struct net_device *dev;
118 netdevice_tracker dev_tracker;
120 struct net *net; /* netns for packet i/o */
122 unsigned long err_time; /* Time when the last ICMP error
124 int err_count; /* Number of arrived ICMP errors */
126 /* These four fields used only by GRE */
127 u32 i_seqno; /* The last seen seqno */
128 atomic_t o_seqno; /* The last output seqno */
129 int tun_hlen; /* Precalculated header length */
131 /* These four fields used only by ERSPAN */
132 u32 index; /* ERSPAN type II index */
133 u8 erspan_ver; /* ERSPAN version */
134 u8 dir; /* ERSPAN direction */
135 u16 hwid; /* ERSPAN hardware ID */
137 struct dst_cache dst_cache;
139 struct ip_tunnel_parm parms;
142 int encap_hlen; /* Encap header length (FOU,GUE) */
143 int hlen; /* tun_hlen + encap_hlen */
144 struct ip_tunnel_encap encap;
147 #ifdef CONFIG_IPV6_SIT_6RD
148 struct ip_tunnel_6rd_parm ip6rd;
150 struct ip_tunnel_prl_entry __rcu *prl; /* potential router list */
151 unsigned int prl_count; /* # of entries in PRL */
152 unsigned int ip_tnl_net_id;
153 struct gro_cells gro_cells;
159 struct tnl_ptk_info {
167 #define PACKET_RCVD 0
168 #define PACKET_REJECT 1
169 #define PACKET_NEXT 2
171 #define IP_TNL_HASH_BITS 7
172 #define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS)
174 struct ip_tunnel_net {
175 struct net_device *fb_tunnel_dev;
176 struct rtnl_link_ops *rtnl_link_ops;
177 struct hlist_head tunnels[IP_TNL_HASH_SIZE];
178 struct ip_tunnel __rcu *collect_md_tun;
182 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
183 __be32 saddr, __be32 daddr,
184 u8 tos, u8 ttl, __be32 label,
185 __be16 tp_src, __be16 tp_dst,
186 __be64 tun_id, __be16 tun_flags)
188 key->tun_id = tun_id;
189 key->u.ipv4.src = saddr;
190 key->u.ipv4.dst = daddr;
191 memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
192 0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
196 key->tun_flags = tun_flags;
198 /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
199 * the upper tunnel are used.
200 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
202 key->tp_src = tp_src;
203 key->tp_dst = tp_dst;
205 /* Clear struct padding. */
206 if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
207 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
208 0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
212 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
213 const struct ip_tunnel_info *info)
219 if (info->key.tun_flags & TUNNEL_NOCACHE)
225 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
228 return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
231 static inline __be64 key32_to_tunnel_id(__be32 key)
234 return (__force __be64)key;
236 return (__force __be64)((__force u64)key << 32);
240 /* Returns the least-significant 32 bits of a __be64. */
241 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
244 return (__force __be32)tun_id;
246 return (__force __be32)((__force u64)tun_id >> 32);
252 static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
254 __be32 daddr, __be32 saddr,
255 __be32 key, __u8 tos,
256 struct net *net, int oif,
257 __u32 mark, __u32 tun_inner_hash,
260 memset(fl4, 0, sizeof(*fl4));
263 fl4->flowi4_l3mdev = l3mdev_master_upper_ifindex_by_index_rcu(net, oif);
264 /* Legacy VRF/l3mdev use case */
265 fl4->flowi4_oif = fl4->flowi4_l3mdev ? 0 : oif;
270 fl4->flowi4_tos = tos;
271 fl4->flowi4_proto = proto;
272 fl4->fl4_gre_key = key;
273 fl4->flowi4_mark = mark;
274 fl4->flowi4_multipath_hash = tun_inner_hash;
275 fl4->flowi4_flags = flow_flags;
278 int ip_tunnel_init(struct net_device *dev);
279 void ip_tunnel_uninit(struct net_device *dev);
280 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
281 struct net *ip_tunnel_get_link_net(const struct net_device *dev);
282 int ip_tunnel_get_iflink(const struct net_device *dev);
283 int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
284 struct rtnl_link_ops *ops, char *devname);
286 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
287 struct rtnl_link_ops *ops);
289 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
290 const struct iphdr *tnl_params, const u8 protocol);
291 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
292 const u8 proto, int tunnel_hlen);
293 int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
294 int ip_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
295 void __user *data, int cmd);
296 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
297 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
299 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
300 int link, __be16 flags,
301 __be32 remote, __be32 local,
304 void ip_tunnel_md_udp_encap(struct sk_buff *skb, struct ip_tunnel_info *info);
305 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
306 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
308 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
309 struct ip_tunnel_parm *p, __u32 fwmark);
310 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
311 struct ip_tunnel_parm *p, __u32 fwmark);
312 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
314 bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
315 struct ip_tunnel_encap *encap);
317 void ip_tunnel_netlink_parms(struct nlattr *data[],
318 struct ip_tunnel_parm *parms);
320 extern const struct header_ops ip_tunnel_header_ops;
321 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb);
323 struct ip_tunnel_encap_ops {
324 size_t (*encap_hlen)(struct ip_tunnel_encap *e);
325 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
326 u8 *protocol, struct flowi4 *fl4);
327 int (*err_handler)(struct sk_buff *skb, u32 info);
330 #define MAX_IPTUN_ENCAP_OPS 8
332 extern const struct ip_tunnel_encap_ops __rcu *
333 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
335 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
337 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
340 int ip_tunnel_encap_setup(struct ip_tunnel *t,
341 struct ip_tunnel_encap *ipencap);
343 static inline bool pskb_inet_may_pull(struct sk_buff *skb)
347 switch (skb->protocol) {
348 #if IS_ENABLED(CONFIG_IPV6)
349 case htons(ETH_P_IPV6):
350 nhlen = sizeof(struct ipv6hdr);
353 case htons(ETH_P_IP):
354 nhlen = sizeof(struct iphdr);
360 return pskb_network_may_pull(skb, nhlen);
363 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
365 const struct ip_tunnel_encap_ops *ops;
368 if (e->type == TUNNEL_ENCAP_NONE)
371 if (e->type >= MAX_IPTUN_ENCAP_OPS)
375 ops = rcu_dereference(iptun_encaps[e->type]);
376 if (likely(ops && ops->encap_hlen))
377 hlen = ops->encap_hlen(e);
383 static inline int ip_tunnel_encap(struct sk_buff *skb,
384 struct ip_tunnel_encap *e,
385 u8 *protocol, struct flowi4 *fl4)
387 const struct ip_tunnel_encap_ops *ops;
390 if (e->type == TUNNEL_ENCAP_NONE)
393 if (e->type >= MAX_IPTUN_ENCAP_OPS)
397 ops = rcu_dereference(iptun_encaps[e->type]);
398 if (likely(ops && ops->build_header))
399 ret = ops->build_header(skb, e, protocol, fl4);
405 /* Extract dsfield from inner protocol */
406 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
407 const struct sk_buff *skb)
409 __be16 payload_protocol = skb_protocol(skb, true);
411 if (payload_protocol == htons(ETH_P_IP))
413 else if (payload_protocol == htons(ETH_P_IPV6))
414 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
419 static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
420 const struct sk_buff *skb)
422 __be16 payload_protocol = skb_protocol(skb, true);
424 if (payload_protocol == htons(ETH_P_IP))
426 else if (payload_protocol == htons(ETH_P_IPV6))
427 return ((const struct ipv6hdr *)iph)->hop_limit;
432 /* Propogate ECN bits out */
433 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
434 const struct sk_buff *skb)
436 u8 inner = ip_tunnel_get_dsfield(iph, skb);
438 return INET_ECN_encapsulate(tos, inner);
441 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
442 __be16 inner_proto, bool raw_proto, bool xnet);
444 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
445 __be16 inner_proto, bool xnet)
447 return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
450 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
451 __be32 src, __be32 dst, u8 proto,
452 u8 tos, u8 ttl, __be16 df, bool xnet);
453 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
455 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
456 int headroom, bool reply);
458 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
460 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
462 if (skb_is_gso(skb)) {
465 err = skb_unclone(skb, GFP_ATOMIC);
468 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
472 skb->encapsulation = 0;
476 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
479 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
481 u64_stats_update_begin(&tstats->syncp);
482 u64_stats_add(&tstats->tx_bytes, pkt_len);
483 u64_stats_inc(&tstats->tx_packets);
484 u64_stats_update_end(&tstats->syncp);
490 DEV_STATS_INC(dev, tx_errors);
491 DEV_STATS_INC(dev, tx_aborted_errors);
493 DEV_STATS_INC(dev, tx_dropped);
497 static inline void ip_tunnel_info_opts_get(void *to,
498 const struct ip_tunnel_info *info)
500 memcpy(to, info + 1, info->options_len);
503 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
504 const void *from, int len,
507 info->options_len = len;
509 memcpy(ip_tunnel_info_opts(info), from, len);
510 info->key.tun_flags |= flags;
514 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
516 return (struct ip_tunnel_info *)lwtstate->data;
519 DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
521 /* Returns > 0 if metadata should be collected */
522 static inline int ip_tunnel_collect_metadata(void)
524 return static_branch_unlikely(&ip_tunnel_metadata_cnt);
527 void __init ip_tunnel_core_init(void);
529 void ip_tunnel_need_metadata(void);
530 void ip_tunnel_unneed_metadata(void);
532 #else /* CONFIG_INET */
534 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
539 static inline void ip_tunnel_need_metadata(void)
543 static inline void ip_tunnel_unneed_metadata(void)
547 static inline void ip_tunnel_info_opts_get(void *to,
548 const struct ip_tunnel_info *info)
552 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
553 const void *from, int len,
556 info->options_len = 0;
559 #endif /* CONFIG_INET */
561 #endif /* __NET_IP_TUNNELS_H */