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 */
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 */
64 /* Maximum tunnel options length. */
65 #define IP_TUNNEL_OPTS_MAX \
66 GENMASK((sizeof_field(struct ip_tunnel_info, \
67 options_len) * BITS_PER_BYTE) - 1, 0)
69 struct ip_tunnel_info {
70 struct ip_tunnel_key key;
71 #ifdef CONFIG_DST_CACHE
72 struct dst_cache dst_cache;
78 /* 6rd prefix/relay information */
79 #ifdef CONFIG_IPV6_SIT_6RD
80 struct ip_tunnel_6rd_parm {
81 struct in6_addr prefix;
88 struct ip_tunnel_encap {
95 struct ip_tunnel_prl_entry {
96 struct ip_tunnel_prl_entry __rcu *next;
99 struct rcu_head rcu_head;
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 */
110 unsigned long err_time; /* Time when the last ICMP error
112 int err_count; /* Number of arrived ICMP errors */
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 */
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 */
125 struct dst_cache dst_cache;
127 struct ip_tunnel_parm parms;
130 int encap_hlen; /* Encap header length (FOU,GUE) */
131 int hlen; /* tun_hlen + encap_hlen */
132 struct ip_tunnel_encap encap;
135 #ifdef CONFIG_IPV6_SIT_6RD
136 struct ip_tunnel_6rd_parm ip6rd;
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;
147 struct tnl_ptk_info {
155 #define PACKET_RCVD 0
156 #define PACKET_REJECT 1
157 #define PACKET_NEXT 2
159 #define IP_TNL_HASH_BITS 7
160 #define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS)
162 struct ip_tunnel_net {
163 struct net_device *fb_tunnel_dev;
164 struct rtnl_link_ops *rtnl_link_ops;
165 struct hlist_head tunnels[IP_TNL_HASH_SIZE];
166 struct ip_tunnel __rcu *collect_md_tun;
170 static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
171 __be32 saddr, __be32 daddr,
172 u8 tos, u8 ttl, __be32 label,
173 __be16 tp_src, __be16 tp_dst,
174 __be64 tun_id, __be16 tun_flags)
176 key->tun_id = tun_id;
177 key->u.ipv4.src = saddr;
178 key->u.ipv4.dst = daddr;
179 memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
180 0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
184 key->tun_flags = tun_flags;
186 /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
187 * the upper tunnel are used.
188 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
190 key->tp_src = tp_src;
191 key->tp_dst = tp_dst;
193 /* Clear struct padding. */
194 if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
195 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
196 0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
200 ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
201 const struct ip_tunnel_info *info)
207 if (info->key.tun_flags & TUNNEL_NOCACHE)
213 static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
216 return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
219 static inline __be64 key32_to_tunnel_id(__be32 key)
222 return (__force __be64)key;
224 return (__force __be64)((__force u64)key << 32);
228 /* Returns the least-significant 32 bits of a __be64. */
229 static inline __be32 tunnel_id_to_key32(__be64 tun_id)
232 return (__force __be32)tun_id;
234 return (__force __be32)((__force u64)tun_id >> 32);
240 static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
242 __be32 daddr, __be32 saddr,
243 __be32 key, __u8 tos, int oif,
244 __u32 mark, __u32 tun_inner_hash)
246 memset(fl4, 0, sizeof(*fl4));
247 fl4->flowi4_oif = oif;
250 fl4->flowi4_tos = tos;
251 fl4->flowi4_proto = proto;
252 fl4->fl4_gre_key = key;
253 fl4->flowi4_mark = mark;
254 fl4->flowi4_multipath_hash = tun_inner_hash;
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);
265 void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
266 struct rtnl_link_ops *ops);
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, int tunnel_hlen);
272 int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
273 int ip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
274 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
275 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
277 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
278 int link, __be16 flags,
279 __be32 remote, __be32 local,
282 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
283 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
285 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
286 struct ip_tunnel_parm *p, __u32 fwmark);
287 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
288 struct ip_tunnel_parm *p, __u32 fwmark);
289 void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
291 extern const struct header_ops ip_tunnel_header_ops;
292 __be16 ip_tunnel_parse_protocol(const struct sk_buff *skb);
294 struct ip_tunnel_encap_ops {
295 size_t (*encap_hlen)(struct ip_tunnel_encap *e);
296 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
297 u8 *protocol, struct flowi4 *fl4);
298 int (*err_handler)(struct sk_buff *skb, u32 info);
301 #define MAX_IPTUN_ENCAP_OPS 8
303 extern const struct ip_tunnel_encap_ops __rcu *
304 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
306 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
308 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
311 int ip_tunnel_encap_setup(struct ip_tunnel *t,
312 struct ip_tunnel_encap *ipencap);
314 static inline bool pskb_inet_may_pull(struct sk_buff *skb)
318 switch (skb->protocol) {
319 #if IS_ENABLED(CONFIG_IPV6)
320 case htons(ETH_P_IPV6):
321 nhlen = sizeof(struct ipv6hdr);
324 case htons(ETH_P_IP):
325 nhlen = sizeof(struct iphdr);
331 return pskb_network_may_pull(skb, nhlen);
334 static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
336 const struct ip_tunnel_encap_ops *ops;
339 if (e->type == TUNNEL_ENCAP_NONE)
342 if (e->type >= MAX_IPTUN_ENCAP_OPS)
346 ops = rcu_dereference(iptun_encaps[e->type]);
347 if (likely(ops && ops->encap_hlen))
348 hlen = ops->encap_hlen(e);
354 static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
355 u8 *protocol, struct flowi4 *fl4)
357 const struct ip_tunnel_encap_ops *ops;
360 if (t->encap.type == TUNNEL_ENCAP_NONE)
363 if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
367 ops = rcu_dereference(iptun_encaps[t->encap.type]);
368 if (likely(ops && ops->build_header))
369 ret = ops->build_header(skb, &t->encap, protocol, fl4);
375 /* Extract dsfield from inner protocol */
376 static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
377 const struct sk_buff *skb)
379 if (skb->protocol == htons(ETH_P_IP))
381 else if (skb->protocol == htons(ETH_P_IPV6))
382 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
387 static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
388 const struct sk_buff *skb)
390 if (skb->protocol == htons(ETH_P_IP))
392 else if (skb->protocol == htons(ETH_P_IPV6))
393 return ((const struct ipv6hdr *)iph)->hop_limit;
398 /* Propogate ECN bits out */
399 static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
400 const struct sk_buff *skb)
402 u8 inner = ip_tunnel_get_dsfield(iph, skb);
404 return INET_ECN_encapsulate(tos, inner);
407 int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
408 __be16 inner_proto, bool raw_proto, bool xnet);
410 static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
411 __be16 inner_proto, bool xnet)
413 return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
416 void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
417 __be32 src, __be32 dst, u8 proto,
418 u8 tos, u8 ttl, __be16 df, bool xnet);
419 struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
421 int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
422 int headroom, bool reply);
424 int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
426 static inline int iptunnel_pull_offloads(struct sk_buff *skb)
428 if (skb_is_gso(skb)) {
431 err = skb_unclone(skb, GFP_ATOMIC);
434 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
438 skb->encapsulation = 0;
442 static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
445 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
447 u64_stats_update_begin(&tstats->syncp);
448 tstats->tx_bytes += pkt_len;
449 tstats->tx_packets++;
450 u64_stats_update_end(&tstats->syncp);
453 struct net_device_stats *err_stats = &dev->stats;
456 err_stats->tx_errors++;
457 err_stats->tx_aborted_errors++;
459 err_stats->tx_dropped++;
464 static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
469 static inline void ip_tunnel_info_opts_get(void *to,
470 const struct ip_tunnel_info *info)
472 memcpy(to, info + 1, info->options_len);
475 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
476 const void *from, int len,
479 info->options_len = len;
481 memcpy(ip_tunnel_info_opts(info), from, len);
482 info->key.tun_flags |= flags;
486 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
488 return (struct ip_tunnel_info *)lwtstate->data;
491 DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
493 /* Returns > 0 if metadata should be collected */
494 static inline int ip_tunnel_collect_metadata(void)
496 return static_branch_unlikely(&ip_tunnel_metadata_cnt);
499 void __init ip_tunnel_core_init(void);
501 void ip_tunnel_need_metadata(void);
502 void ip_tunnel_unneed_metadata(void);
504 #else /* CONFIG_INET */
506 static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
511 static inline void ip_tunnel_need_metadata(void)
515 static inline void ip_tunnel_unneed_metadata(void)
519 static inline void ip_tunnel_info_opts_get(void *to,
520 const struct ip_tunnel_info *info)
524 static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
525 const void *from, int len,
528 info->options_len = 0;
531 #endif /* CONFIG_INET */
533 #endif /* __NET_IP_TUNNELS_H */