tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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
39struct 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 __u8 flow_flags;
58};
59
60struct ip_tunnel_encap {
61 u16 type;
62 u16 flags;
63 __be16 sport;
64 __be16 dport;
65};
66
67/* Flags for ip_tunnel_info mode. */
68#define IP_TUNNEL_INFO_TX 0x01 /* represents tx tunnel parameters */
69#define IP_TUNNEL_INFO_IPV6 0x02 /* key contains IPv6 addresses */
70#define IP_TUNNEL_INFO_BRIDGE 0x04 /* represents a bridged tunnel id */
71
72/* Maximum tunnel options length. */
73#define IP_TUNNEL_OPTS_MAX \
74 GENMASK((sizeof_field(struct ip_tunnel_info, \
75 options_len) * BITS_PER_BYTE) - 1, 0)
76
77#define ip_tunnel_info_opts(info) \
78 _Generic(info, \
79 const struct ip_tunnel_info * : ((const void *)((info) + 1)),\
80 struct ip_tunnel_info * : ((void *)((info) + 1))\
81 )
82
83struct ip_tunnel_info {
84 struct ip_tunnel_key key;
85 struct ip_tunnel_encap encap;
86#ifdef CONFIG_DST_CACHE
87 struct dst_cache dst_cache;
88#endif
89 u8 options_len;
90 u8 mode;
91};
92
93/* 6rd prefix/relay information */
94#ifdef CONFIG_IPV6_SIT_6RD
95struct ip_tunnel_6rd_parm {
96 struct in6_addr prefix;
97 __be32 relay_prefix;
98 u16 prefixlen;
99 u16 relay_prefixlen;
100};
101#endif
102
103struct ip_tunnel_prl_entry {
104 struct ip_tunnel_prl_entry __rcu *next;
105 __be32 addr;
106 u16 flags;
107 struct rcu_head rcu_head;
108};
109
110struct metadata_dst;
111
112struct ip_tunnel {
113 struct ip_tunnel __rcu *next;
114 struct hlist_node hash_node;
115
116 struct net_device *dev;
117 netdevice_tracker dev_tracker;
118
119 struct net *net; /* netns for packet i/o */
120
121 unsigned long err_time; /* Time when the last ICMP error
122 * arrived */
123 int err_count; /* Number of arrived ICMP errors */
124
125 /* These four fields used only by GRE */
126 u32 i_seqno; /* The last seen seqno */
127 atomic_t o_seqno; /* The last output seqno */
128 int tun_hlen; /* Precalculated header length */
129
130 /* These four fields used only by ERSPAN */
131 u32 index; /* ERSPAN type II index */
132 u8 erspan_ver; /* ERSPAN version */
133 u8 dir; /* ERSPAN direction */
134 u16 hwid; /* ERSPAN hardware ID */
135
136 struct dst_cache dst_cache;
137
138 struct ip_tunnel_parm parms;
139
140 int mlink;
141 int encap_hlen; /* Encap header length (FOU,GUE) */
142 int hlen; /* tun_hlen + encap_hlen */
143 struct ip_tunnel_encap encap;
144
145 /* for SIT */
146#ifdef CONFIG_IPV6_SIT_6RD
147 struct ip_tunnel_6rd_parm ip6rd;
148#endif
149 struct ip_tunnel_prl_entry __rcu *prl; /* potential router list */
150 unsigned int prl_count; /* # of entries in PRL */
151 unsigned int ip_tnl_net_id;
152 struct gro_cells gro_cells;
153 __u32 fwmark;
154 bool collect_md;
155 bool ignore_df;
156};
157
158struct tnl_ptk_info {
159 __be16 flags;
160 __be16 proto;
161 __be32 key;
162 __be32 seq;
163 int hdr_len;
164};
165
166#define PACKET_RCVD 0
167#define PACKET_REJECT 1
168#define PACKET_NEXT 2
169
170#define IP_TNL_HASH_BITS 7
171#define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS)
172
173struct ip_tunnel_net {
174 struct net_device *fb_tunnel_dev;
175 struct rtnl_link_ops *rtnl_link_ops;
176 struct hlist_head tunnels[IP_TNL_HASH_SIZE];
177 struct ip_tunnel __rcu *collect_md_tun;
178 int type;
179};
180
181static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
182 __be32 saddr, __be32 daddr,
183 u8 tos, u8 ttl, __be32 label,
184 __be16 tp_src, __be16 tp_dst,
185 __be64 tun_id, __be16 tun_flags)
186{
187 key->tun_id = tun_id;
188 key->u.ipv4.src = saddr;
189 key->u.ipv4.dst = daddr;
190 memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
191 0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
192 key->tos = tos;
193 key->ttl = ttl;
194 key->label = label;
195 key->tun_flags = tun_flags;
196
197 /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
198 * the upper tunnel are used.
199 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
200 */
201 key->tp_src = tp_src;
202 key->tp_dst = tp_dst;
203
204 /* Clear struct padding. */
205 if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
206 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
207 0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
208}
209
210static inline bool
211ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
212 const struct ip_tunnel_info *info)
213{
214 if (skb->mark)
215 return false;
216 if (!info)
217 return true;
218 if (info->key.tun_flags & TUNNEL_NOCACHE)
219 return false;
220
221 return true;
222}
223
224static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
225 *tun_info)
226{
227 return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
228}
229
230static inline __be64 key32_to_tunnel_id(__be32 key)
231{
232#ifdef __BIG_ENDIAN
233 return (__force __be64)key;
234#else
235 return (__force __be64)((__force u64)key << 32);
236#endif
237}
238
239/* Returns the least-significant 32 bits of a __be64. */
240static inline __be32 tunnel_id_to_key32(__be64 tun_id)
241{
242#ifdef __BIG_ENDIAN
243 return (__force __be32)tun_id;
244#else
245 return (__force __be32)((__force u64)tun_id >> 32);
246#endif
247}
248
249#ifdef CONFIG_INET
250
251static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
252 int proto,
253 __be32 daddr, __be32 saddr,
254 __be32 key, __u8 tos,
255 struct net *net, int oif,
256 __u32 mark, __u32 tun_inner_hash,
257 __u8 flow_flags)
258{
259 memset(fl4, 0, sizeof(*fl4));
260
261 if (oif) {
262 fl4->flowi4_l3mdev = l3mdev_master_upper_ifindex_by_index_rcu(net, oif);
263 /* Legacy VRF/l3mdev use case */
264 fl4->flowi4_oif = fl4->flowi4_l3mdev ? 0 : oif;
265 }
266
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 fl4->flowi4_multipath_hash = tun_inner_hash;
274 fl4->flowi4_flags = flow_flags;
275}
276
277int ip_tunnel_init(struct net_device *dev);
278void ip_tunnel_uninit(struct net_device *dev);
279void ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
280struct net *ip_tunnel_get_link_net(const struct net_device *dev);
281int ip_tunnel_get_iflink(const struct net_device *dev);
282int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
283 struct rtnl_link_ops *ops, char *devname);
284
285void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
286 struct rtnl_link_ops *ops);
287
288void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
289 const struct iphdr *tnl_params, const u8 protocol);
290void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
291 const u8 proto, int tunnel_hlen);
292int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
293int ip_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
294 void __user *data, int cmd);
295int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
296int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
297
298struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
299 int link, __be16 flags,
300 __be32 remote, __be32 local,
301 __be32 key);
302
303void ip_tunnel_md_udp_encap(struct sk_buff *skb, struct ip_tunnel_info *info);
304int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
305 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
306 bool log_ecn_error);
307int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
308 struct ip_tunnel_parm *p, __u32 fwmark);
309int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
310 struct ip_tunnel_parm *p, __u32 fwmark);
311void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
312
313bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
314 struct ip_tunnel_encap *encap);
315
316void ip_tunnel_netlink_parms(struct nlattr *data[],
317 struct ip_tunnel_parm *parms);
318
319extern const struct header_ops ip_tunnel_header_ops;
320__be16 ip_tunnel_parse_protocol(const struct sk_buff *skb);
321
322struct ip_tunnel_encap_ops {
323 size_t (*encap_hlen)(struct ip_tunnel_encap *e);
324 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
325 u8 *protocol, struct flowi4 *fl4);
326 int (*err_handler)(struct sk_buff *skb, u32 info);
327};
328
329#define MAX_IPTUN_ENCAP_OPS 8
330
331extern const struct ip_tunnel_encap_ops __rcu *
332 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
333
334int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
335 unsigned int num);
336int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
337 unsigned int num);
338
339int ip_tunnel_encap_setup(struct ip_tunnel *t,
340 struct ip_tunnel_encap *ipencap);
341
342static inline bool pskb_inet_may_pull(struct sk_buff *skb)
343{
344 int nhlen;
345
346 switch (skb->protocol) {
347#if IS_ENABLED(CONFIG_IPV6)
348 case htons(ETH_P_IPV6):
349 nhlen = sizeof(struct ipv6hdr);
350 break;
351#endif
352 case htons(ETH_P_IP):
353 nhlen = sizeof(struct iphdr);
354 break;
355 default:
356 nhlen = 0;
357 }
358
359 return pskb_network_may_pull(skb, nhlen);
360}
361
362static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
363{
364 const struct ip_tunnel_encap_ops *ops;
365 int hlen = -EINVAL;
366
367 if (e->type == TUNNEL_ENCAP_NONE)
368 return 0;
369
370 if (e->type >= MAX_IPTUN_ENCAP_OPS)
371 return -EINVAL;
372
373 rcu_read_lock();
374 ops = rcu_dereference(iptun_encaps[e->type]);
375 if (likely(ops && ops->encap_hlen))
376 hlen = ops->encap_hlen(e);
377 rcu_read_unlock();
378
379 return hlen;
380}
381
382static inline int ip_tunnel_encap(struct sk_buff *skb,
383 struct ip_tunnel_encap *e,
384 u8 *protocol, struct flowi4 *fl4)
385{
386 const struct ip_tunnel_encap_ops *ops;
387 int ret = -EINVAL;
388
389 if (e->type == TUNNEL_ENCAP_NONE)
390 return 0;
391
392 if (e->type >= MAX_IPTUN_ENCAP_OPS)
393 return -EINVAL;
394
395 rcu_read_lock();
396 ops = rcu_dereference(iptun_encaps[e->type]);
397 if (likely(ops && ops->build_header))
398 ret = ops->build_header(skb, e, protocol, fl4);
399 rcu_read_unlock();
400
401 return ret;
402}
403
404/* Extract dsfield from inner protocol */
405static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
406 const struct sk_buff *skb)
407{
408 __be16 payload_protocol = skb_protocol(skb, true);
409
410 if (payload_protocol == htons(ETH_P_IP))
411 return iph->tos;
412 else if (payload_protocol == htons(ETH_P_IPV6))
413 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
414 else
415 return 0;
416}
417
418static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
419 const struct sk_buff *skb)
420{
421 __be16 payload_protocol = skb_protocol(skb, true);
422
423 if (payload_protocol == htons(ETH_P_IP))
424 return iph->ttl;
425 else if (payload_protocol == htons(ETH_P_IPV6))
426 return ((const struct ipv6hdr *)iph)->hop_limit;
427 else
428 return 0;
429}
430
431/* Propogate ECN bits out */
432static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
433 const struct sk_buff *skb)
434{
435 u8 inner = ip_tunnel_get_dsfield(iph, skb);
436
437 return INET_ECN_encapsulate(tos, inner);
438}
439
440int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
441 __be16 inner_proto, bool raw_proto, bool xnet);
442
443static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
444 __be16 inner_proto, bool xnet)
445{
446 return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
447}
448
449void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
450 __be32 src, __be32 dst, u8 proto,
451 u8 tos, u8 ttl, __be16 df, bool xnet);
452struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
453 gfp_t flags);
454int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
455 int headroom, bool reply);
456
457int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
458
459static inline int iptunnel_pull_offloads(struct sk_buff *skb)
460{
461 if (skb_is_gso(skb)) {
462 int err;
463
464 err = skb_unclone(skb, GFP_ATOMIC);
465 if (unlikely(err))
466 return err;
467 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
468 NETIF_F_GSO_SHIFT);
469 }
470
471 skb->encapsulation = 0;
472 return 0;
473}
474
475static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
476{
477 if (pkt_len > 0) {
478 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
479
480 u64_stats_update_begin(&tstats->syncp);
481 u64_stats_add(&tstats->tx_bytes, pkt_len);
482 u64_stats_inc(&tstats->tx_packets);
483 u64_stats_update_end(&tstats->syncp);
484 put_cpu_ptr(tstats);
485 } else {
486 struct net_device_stats *err_stats = &dev->stats;
487
488 if (pkt_len < 0) {
489 err_stats->tx_errors++;
490 err_stats->tx_aborted_errors++;
491 } else {
492 err_stats->tx_dropped++;
493 }
494 }
495}
496
497static inline void ip_tunnel_info_opts_get(void *to,
498 const struct ip_tunnel_info *info)
499{
500 memcpy(to, info + 1, info->options_len);
501}
502
503static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
504 const void *from, int len,
505 __be16 flags)
506{
507 info->options_len = len;
508 if (len > 0) {
509 memcpy(ip_tunnel_info_opts(info), from, len);
510 info->key.tun_flags |= flags;
511 }
512}
513
514static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
515{
516 return (struct ip_tunnel_info *)lwtstate->data;
517}
518
519DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
520
521/* Returns > 0 if metadata should be collected */
522static inline int ip_tunnel_collect_metadata(void)
523{
524 return static_branch_unlikely(&ip_tunnel_metadata_cnt);
525}
526
527void __init ip_tunnel_core_init(void);
528
529void ip_tunnel_need_metadata(void);
530void ip_tunnel_unneed_metadata(void);
531
532#else /* CONFIG_INET */
533
534static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
535{
536 return NULL;
537}
538
539static inline void ip_tunnel_need_metadata(void)
540{
541}
542
543static inline void ip_tunnel_unneed_metadata(void)
544{
545}
546
547static inline void ip_tunnel_info_opts_get(void *to,
548 const struct ip_tunnel_info *info)
549{
550}
551
552static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
553 const void *from, int len,
554 __be16 flags)
555{
556 info->options_len = 0;
557}
558
559#endif /* CONFIG_INET */
560
561#endif /* __NET_IP_TUNNELS_H */