Loading...
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
57struct 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
79struct 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) + 1)),\
99 struct ip_tunnel_info * : ((void *)((info) + 1))\
100 )
101
102struct 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};
111
112/* 6rd prefix/relay information */
113#ifdef CONFIG_IPV6_SIT_6RD
114struct ip_tunnel_6rd_parm {
115 struct in6_addr prefix;
116 __be32 relay_prefix;
117 u16 prefixlen;
118 u16 relay_prefixlen;
119};
120#endif
121
122struct ip_tunnel_prl_entry {
123 struct ip_tunnel_prl_entry __rcu *next;
124 __be32 addr;
125 u16 flags;
126 struct rcu_head rcu_head;
127};
128
129struct metadata_dst;
130
131/* Kernel-side variant of ip_tunnel_parm */
132struct ip_tunnel_parm_kern {
133 char name[IFNAMSIZ];
134 IP_TUNNEL_DECLARE_FLAGS(i_flags);
135 IP_TUNNEL_DECLARE_FLAGS(o_flags);
136 __be32 i_key;
137 __be32 o_key;
138 int link;
139 struct iphdr iph;
140};
141
142struct ip_tunnel {
143 struct ip_tunnel __rcu *next;
144 struct hlist_node hash_node;
145
146 struct net_device *dev;
147 netdevice_tracker dev_tracker;
148
149 struct net *net; /* netns for packet i/o */
150
151 unsigned long err_time; /* Time when the last ICMP error
152 * arrived */
153 int err_count; /* Number of arrived ICMP errors */
154
155 /* These four fields used only by GRE */
156 u32 i_seqno; /* The last seen seqno */
157 atomic_t o_seqno; /* The last output seqno */
158 int tun_hlen; /* Precalculated header length */
159
160 /* These four fields used only by ERSPAN */
161 u32 index; /* ERSPAN type II index */
162 u8 erspan_ver; /* ERSPAN version */
163 u8 dir; /* ERSPAN direction */
164 u16 hwid; /* ERSPAN hardware ID */
165
166 struct dst_cache dst_cache;
167
168 struct ip_tunnel_parm_kern parms;
169
170 int mlink;
171 int encap_hlen; /* Encap header length (FOU,GUE) */
172 int hlen; /* tun_hlen + encap_hlen */
173 struct ip_tunnel_encap encap;
174
175 /* for SIT */
176#ifdef CONFIG_IPV6_SIT_6RD
177 struct ip_tunnel_6rd_parm ip6rd;
178#endif
179 struct ip_tunnel_prl_entry __rcu *prl; /* potential router list */
180 unsigned int prl_count; /* # of entries in PRL */
181 unsigned int ip_tnl_net_id;
182 struct gro_cells gro_cells;
183 __u32 fwmark;
184 bool collect_md;
185 bool ignore_df;
186};
187
188struct tnl_ptk_info {
189 IP_TUNNEL_DECLARE_FLAGS(flags);
190 __be16 proto;
191 __be32 key;
192 __be32 seq;
193 int hdr_len;
194};
195
196#define PACKET_RCVD 0
197#define PACKET_REJECT 1
198#define PACKET_NEXT 2
199
200#define IP_TNL_HASH_BITS 7
201#define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS)
202
203struct ip_tunnel_net {
204 struct net_device *fb_tunnel_dev;
205 struct rtnl_link_ops *rtnl_link_ops;
206 struct hlist_head tunnels[IP_TNL_HASH_SIZE];
207 struct ip_tunnel __rcu *collect_md_tun;
208 int type;
209};
210
211static inline void ip_tunnel_set_options_present(unsigned long *flags)
212{
213 IP_TUNNEL_DECLARE_FLAGS(present) = { };
214
215 __set_bit(IP_TUNNEL_GENEVE_OPT_BIT, present);
216 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, present);
217 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, present);
218 __set_bit(IP_TUNNEL_GTP_OPT_BIT, present);
219 __set_bit(IP_TUNNEL_PFCP_OPT_BIT, present);
220
221 ip_tunnel_flags_or(flags, flags, present);
222}
223
224static inline void ip_tunnel_clear_options_present(unsigned long *flags)
225{
226 IP_TUNNEL_DECLARE_FLAGS(present) = { };
227
228 __set_bit(IP_TUNNEL_GENEVE_OPT_BIT, present);
229 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, present);
230 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, present);
231 __set_bit(IP_TUNNEL_GTP_OPT_BIT, present);
232 __set_bit(IP_TUNNEL_PFCP_OPT_BIT, present);
233
234 __ipt_flag_op(bitmap_andnot, flags, flags, present);
235}
236
237static inline bool ip_tunnel_is_options_present(const unsigned long *flags)
238{
239 IP_TUNNEL_DECLARE_FLAGS(present) = { };
240
241 __set_bit(IP_TUNNEL_GENEVE_OPT_BIT, present);
242 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, present);
243 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, present);
244 __set_bit(IP_TUNNEL_GTP_OPT_BIT, present);
245 __set_bit(IP_TUNNEL_PFCP_OPT_BIT, present);
246
247 return ip_tunnel_flags_intersect(flags, present);
248}
249
250static inline bool ip_tunnel_flags_is_be16_compat(const unsigned long *flags)
251{
252 IP_TUNNEL_DECLARE_FLAGS(supp) = { };
253
254 bitmap_set(supp, 0, BITS_PER_TYPE(__be16));
255 __set_bit(IP_TUNNEL_VTI_BIT, supp);
256
257 return ip_tunnel_flags_subset(flags, supp);
258}
259
260static inline void ip_tunnel_flags_from_be16(unsigned long *dst, __be16 flags)
261{
262 ip_tunnel_flags_zero(dst);
263
264 bitmap_write(dst, be16_to_cpu(flags), 0, BITS_PER_TYPE(__be16));
265 __assign_bit(IP_TUNNEL_VTI_BIT, dst, flags & VTI_ISVTI);
266}
267
268static inline __be16 ip_tunnel_flags_to_be16(const unsigned long *flags)
269{
270 __be16 ret;
271
272 ret = cpu_to_be16(bitmap_read(flags, 0, BITS_PER_TYPE(__be16)));
273 if (test_bit(IP_TUNNEL_VTI_BIT, flags))
274 ret |= VTI_ISVTI;
275
276 return ret;
277}
278
279static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
280 __be32 saddr, __be32 daddr,
281 u8 tos, u8 ttl, __be32 label,
282 __be16 tp_src, __be16 tp_dst,
283 __be64 tun_id,
284 const unsigned long *tun_flags)
285{
286 key->tun_id = tun_id;
287 key->u.ipv4.src = saddr;
288 key->u.ipv4.dst = daddr;
289 memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
290 0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
291 key->tos = tos;
292 key->ttl = ttl;
293 key->label = label;
294 ip_tunnel_flags_copy(key->tun_flags, tun_flags);
295
296 /* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
297 * the upper tunnel are used.
298 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
299 */
300 key->tp_src = tp_src;
301 key->tp_dst = tp_dst;
302
303 /* Clear struct padding. */
304 if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
305 memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
306 0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
307}
308
309static inline bool
310ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
311 const struct ip_tunnel_info *info)
312{
313 if (skb->mark)
314 return false;
315
316 return !info || !test_bit(IP_TUNNEL_NOCACHE_BIT, info->key.tun_flags);
317}
318
319static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
320 *tun_info)
321{
322 return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
323}
324
325static inline __be64 key32_to_tunnel_id(__be32 key)
326{
327#ifdef __BIG_ENDIAN
328 return (__force __be64)key;
329#else
330 return (__force __be64)((__force u64)key << 32);
331#endif
332}
333
334/* Returns the least-significant 32 bits of a __be64. */
335static inline __be32 tunnel_id_to_key32(__be64 tun_id)
336{
337#ifdef __BIG_ENDIAN
338 return (__force __be32)tun_id;
339#else
340 return (__force __be32)((__force u64)tun_id >> 32);
341#endif
342}
343
344#ifdef CONFIG_INET
345
346static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
347 int proto,
348 __be32 daddr, __be32 saddr,
349 __be32 key, __u8 tos,
350 struct net *net, int oif,
351 __u32 mark, __u32 tun_inner_hash,
352 __u8 flow_flags)
353{
354 memset(fl4, 0, sizeof(*fl4));
355
356 if (oif) {
357 fl4->flowi4_l3mdev = l3mdev_master_upper_ifindex_by_index(net, oif);
358 /* Legacy VRF/l3mdev use case */
359 fl4->flowi4_oif = fl4->flowi4_l3mdev ? 0 : oif;
360 }
361
362 fl4->daddr = daddr;
363 fl4->saddr = saddr;
364 fl4->flowi4_tos = tos;
365 fl4->flowi4_proto = proto;
366 fl4->fl4_gre_key = key;
367 fl4->flowi4_mark = mark;
368 fl4->flowi4_multipath_hash = tun_inner_hash;
369 fl4->flowi4_flags = flow_flags;
370}
371
372int ip_tunnel_init(struct net_device *dev);
373void ip_tunnel_uninit(struct net_device *dev);
374void ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
375struct net *ip_tunnel_get_link_net(const struct net_device *dev);
376int ip_tunnel_get_iflink(const struct net_device *dev);
377int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
378 struct rtnl_link_ops *ops, char *devname);
379
380void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
381 struct rtnl_link_ops *ops,
382 struct list_head *dev_to_kill);
383
384void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
385 const struct iphdr *tnl_params, const u8 protocol);
386void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
387 const u8 proto, int tunnel_hlen);
388int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm_kern *p,
389 int cmd);
390bool ip_tunnel_parm_from_user(struct ip_tunnel_parm_kern *kp,
391 const void __user *data);
392bool ip_tunnel_parm_to_user(void __user *data, struct ip_tunnel_parm_kern *kp);
393int ip_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
394 void __user *data, int cmd);
395int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
396int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
397
398struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
399 int link, const unsigned long *flags,
400 __be32 remote, __be32 local,
401 __be32 key);
402
403void ip_tunnel_md_udp_encap(struct sk_buff *skb, struct ip_tunnel_info *info);
404int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
405 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
406 bool log_ecn_error);
407int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
408 struct ip_tunnel_parm_kern *p, __u32 fwmark);
409int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
410 struct ip_tunnel_parm_kern *p, __u32 fwmark);
411void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
412
413bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
414 struct ip_tunnel_encap *encap);
415
416void ip_tunnel_netlink_parms(struct nlattr *data[],
417 struct ip_tunnel_parm_kern *parms);
418
419extern const struct header_ops ip_tunnel_header_ops;
420__be16 ip_tunnel_parse_protocol(const struct sk_buff *skb);
421
422struct ip_tunnel_encap_ops {
423 size_t (*encap_hlen)(struct ip_tunnel_encap *e);
424 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
425 u8 *protocol, struct flowi4 *fl4);
426 int (*err_handler)(struct sk_buff *skb, u32 info);
427};
428
429#define MAX_IPTUN_ENCAP_OPS 8
430
431extern const struct ip_tunnel_encap_ops __rcu *
432 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
433
434int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
435 unsigned int num);
436int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
437 unsigned int num);
438
439int ip_tunnel_encap_setup(struct ip_tunnel *t,
440 struct ip_tunnel_encap *ipencap);
441
442static inline enum skb_drop_reason
443pskb_inet_may_pull_reason(struct sk_buff *skb)
444{
445 int nhlen;
446
447 switch (skb->protocol) {
448#if IS_ENABLED(CONFIG_IPV6)
449 case htons(ETH_P_IPV6):
450 nhlen = sizeof(struct ipv6hdr);
451 break;
452#endif
453 case htons(ETH_P_IP):
454 nhlen = sizeof(struct iphdr);
455 break;
456 default:
457 nhlen = 0;
458 }
459
460 return pskb_network_may_pull_reason(skb, nhlen);
461}
462
463static inline bool pskb_inet_may_pull(struct sk_buff *skb)
464{
465 return pskb_inet_may_pull_reason(skb) == SKB_NOT_DROPPED_YET;
466}
467
468/* Variant of pskb_inet_may_pull().
469 */
470static inline enum skb_drop_reason
471skb_vlan_inet_prepare(struct sk_buff *skb, bool inner_proto_inherit)
472{
473 int nhlen = 0, maclen = inner_proto_inherit ? 0 : ETH_HLEN;
474 __be16 type = skb->protocol;
475 enum skb_drop_reason reason;
476
477 /* Essentially this is skb_protocol(skb, true)
478 * And we get MAC len.
479 */
480 if (eth_type_vlan(type))
481 type = __vlan_get_protocol(skb, type, &maclen);
482
483 switch (type) {
484#if IS_ENABLED(CONFIG_IPV6)
485 case htons(ETH_P_IPV6):
486 nhlen = sizeof(struct ipv6hdr);
487 break;
488#endif
489 case htons(ETH_P_IP):
490 nhlen = sizeof(struct iphdr);
491 break;
492 }
493 /* For ETH_P_IPV6/ETH_P_IP we make sure to pull
494 * a base network header in skb->head.
495 */
496 reason = pskb_may_pull_reason(skb, maclen + nhlen);
497 if (reason)
498 return reason;
499
500 skb_set_network_header(skb, maclen);
501
502 return SKB_NOT_DROPPED_YET;
503}
504
505static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
506{
507 const struct ip_tunnel_encap_ops *ops;
508 int hlen = -EINVAL;
509
510 if (e->type == TUNNEL_ENCAP_NONE)
511 return 0;
512
513 if (e->type >= MAX_IPTUN_ENCAP_OPS)
514 return -EINVAL;
515
516 rcu_read_lock();
517 ops = rcu_dereference(iptun_encaps[e->type]);
518 if (likely(ops && ops->encap_hlen))
519 hlen = ops->encap_hlen(e);
520 rcu_read_unlock();
521
522 return hlen;
523}
524
525static inline int ip_tunnel_encap(struct sk_buff *skb,
526 struct ip_tunnel_encap *e,
527 u8 *protocol, struct flowi4 *fl4)
528{
529 const struct ip_tunnel_encap_ops *ops;
530 int ret = -EINVAL;
531
532 if (e->type == TUNNEL_ENCAP_NONE)
533 return 0;
534
535 if (e->type >= MAX_IPTUN_ENCAP_OPS)
536 return -EINVAL;
537
538 rcu_read_lock();
539 ops = rcu_dereference(iptun_encaps[e->type]);
540 if (likely(ops && ops->build_header))
541 ret = ops->build_header(skb, e, protocol, fl4);
542 rcu_read_unlock();
543
544 return ret;
545}
546
547/* Extract dsfield from inner protocol */
548static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
549 const struct sk_buff *skb)
550{
551 __be16 payload_protocol = skb_protocol(skb, true);
552
553 if (payload_protocol == htons(ETH_P_IP))
554 return iph->tos;
555 else if (payload_protocol == htons(ETH_P_IPV6))
556 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
557 else
558 return 0;
559}
560
561static inline __be32 ip_tunnel_get_flowlabel(const struct iphdr *iph,
562 const struct sk_buff *skb)
563{
564 __be16 payload_protocol = skb_protocol(skb, true);
565
566 if (payload_protocol == htons(ETH_P_IPV6))
567 return ip6_flowlabel((const struct ipv6hdr *)iph);
568 else
569 return 0;
570}
571
572static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
573 const struct sk_buff *skb)
574{
575 __be16 payload_protocol = skb_protocol(skb, true);
576
577 if (payload_protocol == htons(ETH_P_IP))
578 return iph->ttl;
579 else if (payload_protocol == htons(ETH_P_IPV6))
580 return ((const struct ipv6hdr *)iph)->hop_limit;
581 else
582 return 0;
583}
584
585/* Propagate ECN bits out */
586static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
587 const struct sk_buff *skb)
588{
589 u8 inner = ip_tunnel_get_dsfield(iph, skb);
590
591 return INET_ECN_encapsulate(tos, inner);
592}
593
594int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
595 __be16 inner_proto, bool raw_proto, bool xnet);
596
597static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
598 __be16 inner_proto, bool xnet)
599{
600 return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
601}
602
603void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
604 __be32 src, __be32 dst, u8 proto,
605 u8 tos, u8 ttl, __be16 df, bool xnet);
606struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
607 gfp_t flags);
608int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
609 int headroom, bool reply);
610
611int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
612
613static inline int iptunnel_pull_offloads(struct sk_buff *skb)
614{
615 if (skb_is_gso(skb)) {
616 int err;
617
618 err = skb_unclone(skb, GFP_ATOMIC);
619 if (unlikely(err))
620 return err;
621 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
622 NETIF_F_GSO_SHIFT);
623 }
624
625 skb->encapsulation = 0;
626 return 0;
627}
628
629static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
630{
631 if (pkt_len > 0) {
632 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
633
634 u64_stats_update_begin(&tstats->syncp);
635 u64_stats_add(&tstats->tx_bytes, pkt_len);
636 u64_stats_inc(&tstats->tx_packets);
637 u64_stats_update_end(&tstats->syncp);
638 put_cpu_ptr(tstats);
639 return;
640 }
641
642 if (pkt_len < 0) {
643 DEV_STATS_INC(dev, tx_errors);
644 DEV_STATS_INC(dev, tx_aborted_errors);
645 } else {
646 DEV_STATS_INC(dev, tx_dropped);
647 }
648}
649
650static inline void ip_tunnel_info_opts_get(void *to,
651 const struct ip_tunnel_info *info)
652{
653 memcpy(to, info + 1, info->options_len);
654}
655
656static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
657 const void *from, int len,
658 const unsigned long *flags)
659{
660 info->options_len = len;
661 if (len > 0) {
662 memcpy(ip_tunnel_info_opts(info), from, len);
663 ip_tunnel_flags_or(info->key.tun_flags, info->key.tun_flags,
664 flags);
665 }
666}
667
668static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
669{
670 return (struct ip_tunnel_info *)lwtstate->data;
671}
672
673DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
674
675/* Returns > 0 if metadata should be collected */
676static inline int ip_tunnel_collect_metadata(void)
677{
678 return static_branch_unlikely(&ip_tunnel_metadata_cnt);
679}
680
681void __init ip_tunnel_core_init(void);
682
683void ip_tunnel_need_metadata(void);
684void ip_tunnel_unneed_metadata(void);
685
686#else /* CONFIG_INET */
687
688static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
689{
690 return NULL;
691}
692
693static inline void ip_tunnel_need_metadata(void)
694{
695}
696
697static inline void ip_tunnel_unneed_metadata(void)
698{
699}
700
701static inline void ip_tunnel_info_opts_get(void *to,
702 const struct ip_tunnel_info *info)
703{
704}
705
706static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
707 const void *from, int len,
708 const unsigned long *flags)
709{
710 info->options_len = 0;
711}
712
713#endif /* CONFIG_INET */
714
715#endif /* __NET_IP_TUNNELS_H */
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};
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((sizeof_field(struct ip_tunnel_info, \
67 options_len) * BITS_PER_BYTE) - 1, 0)
68
69struct 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
80struct ip_tunnel_6rd_parm {
81 struct in6_addr prefix;
82 __be32 relay_prefix;
83 u16 prefixlen;
84 u16 relay_prefixlen;
85};
86#endif
87
88struct ip_tunnel_encap {
89 u16 type;
90 u16 flags;
91 __be16 sport;
92 __be16 dport;
93};
94
95struct 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
102struct metadata_dst;
103
104struct 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
147struct tnl_ptk_info {
148 __be16 flags;
149 __be16 proto;
150 __be32 key;
151 __be32 seq;
152 int hdr_len;
153};
154
155#define PACKET_RCVD 0
156#define PACKET_REJECT 1
157#define PACKET_NEXT 2
158
159#define IP_TNL_HASH_BITS 7
160#define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS)
161
162struct 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;
167 int type;
168};
169
170static 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)
175{
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);
181 key->tos = tos;
182 key->ttl = ttl;
183 key->label = label;
184 key->tun_flags = tun_flags;
185
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.
189 */
190 key->tp_src = tp_src;
191 key->tp_dst = tp_dst;
192
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);
197}
198
199static inline bool
200ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
201 const struct ip_tunnel_info *info)
202{
203 if (skb->mark)
204 return false;
205 if (!info)
206 return true;
207 if (info->key.tun_flags & TUNNEL_NOCACHE)
208 return false;
209
210 return true;
211}
212
213static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
214 *tun_info)
215{
216 return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
217}
218
219static inline __be64 key32_to_tunnel_id(__be32 key)
220{
221#ifdef __BIG_ENDIAN
222 return (__force __be64)key;
223#else
224 return (__force __be64)((__force u64)key << 32);
225#endif
226}
227
228/* Returns the least-significant 32 bits of a __be64. */
229static inline __be32 tunnel_id_to_key32(__be64 tun_id)
230{
231#ifdef __BIG_ENDIAN
232 return (__force __be32)tun_id;
233#else
234 return (__force __be32)((__force u64)tun_id >> 32);
235#endif
236}
237
238#ifdef CONFIG_INET
239
240static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
241 int proto,
242 __be32 daddr, __be32 saddr,
243 __be32 key, __u8 tos, int oif,
244 __u32 mark, __u32 tun_inner_hash)
245{
246 memset(fl4, 0, sizeof(*fl4));
247 fl4->flowi4_oif = oif;
248 fl4->daddr = daddr;
249 fl4->saddr = saddr;
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;
255}
256
257int ip_tunnel_init(struct net_device *dev);
258void ip_tunnel_uninit(struct net_device *dev);
259void ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
260struct net *ip_tunnel_get_link_net(const struct net_device *dev);
261int ip_tunnel_get_iflink(const struct net_device *dev);
262int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
263 struct rtnl_link_ops *ops, char *devname);
264
265void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
266 struct rtnl_link_ops *ops);
267
268void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
269 const struct iphdr *tnl_params, const u8 protocol);
270void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
271 const u8 proto, int tunnel_hlen);
272int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
273int ip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
274int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
275int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
276
277void ip_tunnel_get_stats64(struct net_device *dev,
278 struct rtnl_link_stats64 *tot);
279struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
280 int link, __be16 flags,
281 __be32 remote, __be32 local,
282 __be32 key);
283
284int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
285 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
286 bool log_ecn_error);
287int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
288 struct ip_tunnel_parm *p, __u32 fwmark);
289int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
290 struct ip_tunnel_parm *p, __u32 fwmark);
291void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
292
293extern const struct header_ops ip_tunnel_header_ops;
294__be16 ip_tunnel_parse_protocol(const struct sk_buff *skb);
295
296struct ip_tunnel_encap_ops {
297 size_t (*encap_hlen)(struct ip_tunnel_encap *e);
298 int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
299 u8 *protocol, struct flowi4 *fl4);
300 int (*err_handler)(struct sk_buff *skb, u32 info);
301};
302
303#define MAX_IPTUN_ENCAP_OPS 8
304
305extern const struct ip_tunnel_encap_ops __rcu *
306 iptun_encaps[MAX_IPTUN_ENCAP_OPS];
307
308int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
309 unsigned int num);
310int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
311 unsigned int num);
312
313int ip_tunnel_encap_setup(struct ip_tunnel *t,
314 struct ip_tunnel_encap *ipencap);
315
316static inline bool pskb_inet_may_pull(struct sk_buff *skb)
317{
318 int nhlen;
319
320 switch (skb->protocol) {
321#if IS_ENABLED(CONFIG_IPV6)
322 case htons(ETH_P_IPV6):
323 nhlen = sizeof(struct ipv6hdr);
324 break;
325#endif
326 case htons(ETH_P_IP):
327 nhlen = sizeof(struct iphdr);
328 break;
329 default:
330 nhlen = 0;
331 }
332
333 return pskb_network_may_pull(skb, nhlen);
334}
335
336static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
337{
338 const struct ip_tunnel_encap_ops *ops;
339 int hlen = -EINVAL;
340
341 if (e->type == TUNNEL_ENCAP_NONE)
342 return 0;
343
344 if (e->type >= MAX_IPTUN_ENCAP_OPS)
345 return -EINVAL;
346
347 rcu_read_lock();
348 ops = rcu_dereference(iptun_encaps[e->type]);
349 if (likely(ops && ops->encap_hlen))
350 hlen = ops->encap_hlen(e);
351 rcu_read_unlock();
352
353 return hlen;
354}
355
356static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
357 u8 *protocol, struct flowi4 *fl4)
358{
359 const struct ip_tunnel_encap_ops *ops;
360 int ret = -EINVAL;
361
362 if (t->encap.type == TUNNEL_ENCAP_NONE)
363 return 0;
364
365 if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
366 return -EINVAL;
367
368 rcu_read_lock();
369 ops = rcu_dereference(iptun_encaps[t->encap.type]);
370 if (likely(ops && ops->build_header))
371 ret = ops->build_header(skb, &t->encap, protocol, fl4);
372 rcu_read_unlock();
373
374 return ret;
375}
376
377/* Extract dsfield from inner protocol */
378static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
379 const struct sk_buff *skb)
380{
381 if (skb->protocol == htons(ETH_P_IP))
382 return iph->tos;
383 else if (skb->protocol == htons(ETH_P_IPV6))
384 return ipv6_get_dsfield((const struct ipv6hdr *)iph);
385 else
386 return 0;
387}
388
389static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
390 const struct sk_buff *skb)
391{
392 if (skb->protocol == htons(ETH_P_IP))
393 return iph->ttl;
394 else if (skb->protocol == htons(ETH_P_IPV6))
395 return ((const struct ipv6hdr *)iph)->hop_limit;
396 else
397 return 0;
398}
399
400/* Propogate ECN bits out */
401static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
402 const struct sk_buff *skb)
403{
404 u8 inner = ip_tunnel_get_dsfield(iph, skb);
405
406 return INET_ECN_encapsulate(tos, inner);
407}
408
409int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
410 __be16 inner_proto, bool raw_proto, bool xnet);
411
412static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
413 __be16 inner_proto, bool xnet)
414{
415 return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
416}
417
418void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
419 __be32 src, __be32 dst, u8 proto,
420 u8 tos, u8 ttl, __be16 df, bool xnet);
421struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
422 gfp_t flags);
423int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
424 int headroom, bool reply);
425
426int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
427
428static inline int iptunnel_pull_offloads(struct sk_buff *skb)
429{
430 if (skb_is_gso(skb)) {
431 int err;
432
433 err = skb_unclone(skb, GFP_ATOMIC);
434 if (unlikely(err))
435 return err;
436 skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
437 NETIF_F_GSO_SHIFT);
438 }
439
440 skb->encapsulation = 0;
441 return 0;
442}
443
444static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
445{
446 if (pkt_len > 0) {
447 struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
448
449 u64_stats_update_begin(&tstats->syncp);
450 tstats->tx_bytes += pkt_len;
451 tstats->tx_packets++;
452 u64_stats_update_end(&tstats->syncp);
453 put_cpu_ptr(tstats);
454 } else {
455 struct net_device_stats *err_stats = &dev->stats;
456
457 if (pkt_len < 0) {
458 err_stats->tx_errors++;
459 err_stats->tx_aborted_errors++;
460 } else {
461 err_stats->tx_dropped++;
462 }
463 }
464}
465
466static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
467{
468 return info + 1;
469}
470
471static inline void ip_tunnel_info_opts_get(void *to,
472 const struct ip_tunnel_info *info)
473{
474 memcpy(to, info + 1, info->options_len);
475}
476
477static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
478 const void *from, int len,
479 __be16 flags)
480{
481 memcpy(ip_tunnel_info_opts(info), from, len);
482 info->options_len = len;
483 info->key.tun_flags |= flags;
484}
485
486static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
487{
488 return (struct ip_tunnel_info *)lwtstate->data;
489}
490
491DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
492
493/* Returns > 0 if metadata should be collected */
494static inline int ip_tunnel_collect_metadata(void)
495{
496 return static_branch_unlikely(&ip_tunnel_metadata_cnt);
497}
498
499void __init ip_tunnel_core_init(void);
500
501void ip_tunnel_need_metadata(void);
502void ip_tunnel_unneed_metadata(void);
503
504#else /* CONFIG_INET */
505
506static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
507{
508 return NULL;
509}
510
511static inline void ip_tunnel_need_metadata(void)
512{
513}
514
515static inline void ip_tunnel_unneed_metadata(void)
516{
517}
518
519static inline void ip_tunnel_info_opts_get(void *to,
520 const struct ip_tunnel_info *info)
521{
522}
523
524static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
525 const void *from, int len,
526 __be16 flags)
527{
528 info->options_len = 0;
529 info->key.tun_flags |= flags;
530}
531
532#endif /* CONFIG_INET */
533
534#endif /* __NET_IP_TUNNELS_H */