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	u32			nhid;
56	__be16			tp_src;
57	__be16			tp_dst;
58	__u8			flow_flags;
59};
60
61struct ip_tunnel_encap {
62	u16			type;
63	u16			flags;
64	__be16			sport;
65	__be16			dport;
66};
67
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 */
72
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)
77
78#define ip_tunnel_info_opts(info)				\
79	_Generic(info,						\
80		 const struct ip_tunnel_info * : ((const void *)((info) + 1)),\
81		 struct ip_tunnel_info * : ((void *)((info) + 1))\
82	)
83
84struct 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;
89#endif
90	u8			options_len;
91	u8			mode;
92};
93
94/* 6rd prefix/relay information */
95#ifdef CONFIG_IPV6_SIT_6RD
96struct ip_tunnel_6rd_parm {
97	struct in6_addr		prefix;
98	__be32			relay_prefix;
99	u16			prefixlen;
100	u16			relay_prefixlen;
101};
102#endif
103
104struct ip_tunnel_prl_entry {
105	struct ip_tunnel_prl_entry __rcu *next;
106	__be32				addr;
107	u16				flags;
108	struct rcu_head			rcu_head;
109};
110
111struct metadata_dst;
112
113struct ip_tunnel {
114	struct ip_tunnel __rcu	*next;
115	struct hlist_node hash_node;
116
117	struct net_device	*dev;
118	netdevice_tracker	dev_tracker;
119
120	struct net		*net;	/* netns for packet i/o */
121
122	unsigned long	err_time;	/* Time when the last ICMP error
123					 * arrived */
124	int		err_count;	/* Number of arrived ICMP errors */
125
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 */
130
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 */
136
137	struct dst_cache dst_cache;
138
139	struct ip_tunnel_parm parms;
140
141	int		mlink;
142	int		encap_hlen;	/* Encap header length (FOU,GUE) */
143	int		hlen;		/* tun_hlen + encap_hlen */
144	struct ip_tunnel_encap encap;
145
146	/* for SIT */
147#ifdef CONFIG_IPV6_SIT_6RD
148	struct ip_tunnel_6rd_parm ip6rd;
149#endif
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;
154	__u32			fwmark;
155	bool			collect_md;
156	bool			ignore_df;
157};
158
159struct tnl_ptk_info {
160	__be16 flags;
161	__be16 proto;
162	__be32 key;
163	__be32 seq;
164	int hdr_len;
165};
166
167#define PACKET_RCVD	0
168#define PACKET_REJECT	1
169#define PACKET_NEXT	2
170
171#define IP_TNL_HASH_BITS   7
172#define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)
173
174struct 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;
179	int type;
180};
181
182static 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)
187{
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);
193	key->tos = tos;
194	key->ttl = ttl;
195	key->label = label;
196	key->tun_flags = tun_flags;
197
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.
201	 */
202	key->tp_src = tp_src;
203	key->tp_dst = tp_dst;
204
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);
209}
210
211static inline bool
212ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
213			   const struct ip_tunnel_info *info)
214{
215	if (skb->mark)
216		return false;
217	if (!info)
218		return true;
219	if (info->key.tun_flags & TUNNEL_NOCACHE)
220		return false;
221
222	return true;
223}
224
225static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
226					       *tun_info)
227{
228	return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
229}
230
231static inline __be64 key32_to_tunnel_id(__be32 key)
232{
233#ifdef __BIG_ENDIAN
234	return (__force __be64)key;
235#else
236	return (__force __be64)((__force u64)key << 32);
237#endif
238}
239
240/* Returns the least-significant 32 bits of a __be64. */
241static inline __be32 tunnel_id_to_key32(__be64 tun_id)
242{
243#ifdef __BIG_ENDIAN
244	return (__force __be32)tun_id;
245#else
246	return (__force __be32)((__force u64)tun_id >> 32);
247#endif
248}
249
250#ifdef CONFIG_INET
251
252static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
253				       int proto,
254				       __be32 daddr, __be32 saddr,
255				       __be32 key, __u8 tos,
256				       struct net *net, int oif,
257				       __u32 mark, __u32 tun_inner_hash,
258				       __u8 flow_flags)
259{
260	memset(fl4, 0, sizeof(*fl4));
261
262	if (oif) {
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;
266	}
267
268	fl4->daddr = daddr;
269	fl4->saddr = saddr;
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;
276}
277
278int ip_tunnel_init(struct net_device *dev);
279void ip_tunnel_uninit(struct net_device *dev);
280void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
281struct net *ip_tunnel_get_link_net(const struct net_device *dev);
282int ip_tunnel_get_iflink(const struct net_device *dev);
283int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
284		       struct rtnl_link_ops *ops, char *devname);
285
286void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
287			   struct rtnl_link_ops *ops,
288			   struct list_head *dev_to_kill);
289
290void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
291		    const struct iphdr *tnl_params, const u8 protocol);
292void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
293		       const u8 proto, int tunnel_hlen);
294int ip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
295int ip_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
296			     void __user *data, int cmd);
297int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
298int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
299
300struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
301				   int link, __be16 flags,
302				   __be32 remote, __be32 local,
303				   __be32 key);
304
305void ip_tunnel_md_udp_encap(struct sk_buff *skb, struct ip_tunnel_info *info);
306int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
307		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
308		  bool log_ecn_error);
309int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
310			 struct ip_tunnel_parm *p, __u32 fwmark);
311int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
312		      struct ip_tunnel_parm *p, __u32 fwmark);
313void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
314
315bool ip_tunnel_netlink_encap_parms(struct nlattr *data[],
316				   struct ip_tunnel_encap *encap);
317
318void ip_tunnel_netlink_parms(struct nlattr *data[],
319			     struct ip_tunnel_parm *parms);
320
321extern const struct header_ops ip_tunnel_header_ops;
322__be16 ip_tunnel_parse_protocol(const struct sk_buff *skb);
323
324struct ip_tunnel_encap_ops {
325	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
326	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
327			    u8 *protocol, struct flowi4 *fl4);
328	int (*err_handler)(struct sk_buff *skb, u32 info);
329};
330
331#define MAX_IPTUN_ENCAP_OPS 8
332
333extern const struct ip_tunnel_encap_ops __rcu *
334		iptun_encaps[MAX_IPTUN_ENCAP_OPS];
335
336int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
337			    unsigned int num);
338int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
339			    unsigned int num);
340
341int ip_tunnel_encap_setup(struct ip_tunnel *t,
342			  struct ip_tunnel_encap *ipencap);
343
344static inline bool pskb_inet_may_pull(struct sk_buff *skb)
345{
346	int nhlen;
347
348	switch (skb->protocol) {
349#if IS_ENABLED(CONFIG_IPV6)
350	case htons(ETH_P_IPV6):
351		nhlen = sizeof(struct ipv6hdr);
352		break;
353#endif
354	case htons(ETH_P_IP):
355		nhlen = sizeof(struct iphdr);
356		break;
357	default:
358		nhlen = 0;
359	}
360
361	return pskb_network_may_pull(skb, nhlen);
362}
363
364/* Variant of pskb_inet_may_pull().
365 */
366static inline bool skb_vlan_inet_prepare(struct sk_buff *skb)
367{
368	int nhlen = 0, maclen = ETH_HLEN;
369	__be16 type = skb->protocol;
370
371	/* Essentially this is skb_protocol(skb, true)
372	 * And we get MAC len.
373	 */
374	if (eth_type_vlan(type))
375		type = __vlan_get_protocol(skb, type, &maclen);
376
377	switch (type) {
378#if IS_ENABLED(CONFIG_IPV6)
379	case htons(ETH_P_IPV6):
380		nhlen = sizeof(struct ipv6hdr);
381		break;
382#endif
383	case htons(ETH_P_IP):
384		nhlen = sizeof(struct iphdr);
385		break;
386	}
387	/* For ETH_P_IPV6/ETH_P_IP we make sure to pull
388	 * a base network header in skb->head.
389	 */
390	if (!pskb_may_pull(skb, maclen + nhlen))
391		return false;
392
393	skb_set_network_header(skb, maclen);
394	return true;
395}
396
397static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
398{
399	const struct ip_tunnel_encap_ops *ops;
400	int hlen = -EINVAL;
401
402	if (e->type == TUNNEL_ENCAP_NONE)
403		return 0;
404
405	if (e->type >= MAX_IPTUN_ENCAP_OPS)
406		return -EINVAL;
407
408	rcu_read_lock();
409	ops = rcu_dereference(iptun_encaps[e->type]);
410	if (likely(ops && ops->encap_hlen))
411		hlen = ops->encap_hlen(e);
412	rcu_read_unlock();
413
414	return hlen;
415}
416
417static inline int ip_tunnel_encap(struct sk_buff *skb,
418				  struct ip_tunnel_encap *e,
419				  u8 *protocol, struct flowi4 *fl4)
420{
421	const struct ip_tunnel_encap_ops *ops;
422	int ret = -EINVAL;
423
424	if (e->type == TUNNEL_ENCAP_NONE)
425		return 0;
426
427	if (e->type >= MAX_IPTUN_ENCAP_OPS)
428		return -EINVAL;
429
430	rcu_read_lock();
431	ops = rcu_dereference(iptun_encaps[e->type]);
432	if (likely(ops && ops->build_header))
433		ret = ops->build_header(skb, e, protocol, fl4);
434	rcu_read_unlock();
435
436	return ret;
437}
438
439/* Extract dsfield from inner protocol */
440static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
441				       const struct sk_buff *skb)
442{
443	__be16 payload_protocol = skb_protocol(skb, true);
444
445	if (payload_protocol == htons(ETH_P_IP))
446		return iph->tos;
447	else if (payload_protocol == htons(ETH_P_IPV6))
448		return ipv6_get_dsfield((const struct ipv6hdr *)iph);
449	else
450		return 0;
451}
452
453static inline __be32 ip_tunnel_get_flowlabel(const struct iphdr *iph,
454					     const struct sk_buff *skb)
455{
456	__be16 payload_protocol = skb_protocol(skb, true);
457
458	if (payload_protocol == htons(ETH_P_IPV6))
459		return ip6_flowlabel((const struct ipv6hdr *)iph);
460	else
461		return 0;
462}
463
464static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
465				       const struct sk_buff *skb)
466{
467	__be16 payload_protocol = skb_protocol(skb, true);
468
469	if (payload_protocol == htons(ETH_P_IP))
470		return iph->ttl;
471	else if (payload_protocol == htons(ETH_P_IPV6))
472		return ((const struct ipv6hdr *)iph)->hop_limit;
473	else
474		return 0;
475}
476
477/* Propogate ECN bits out */
478static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
479				     const struct sk_buff *skb)
480{
481	u8 inner = ip_tunnel_get_dsfield(iph, skb);
482
483	return INET_ECN_encapsulate(tos, inner);
484}
485
486int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
487			   __be16 inner_proto, bool raw_proto, bool xnet);
488
489static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
490				       __be16 inner_proto, bool xnet)
491{
492	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
493}
494
495void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
496		   __be32 src, __be32 dst, u8 proto,
497		   u8 tos, u8 ttl, __be16 df, bool xnet);
498struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
499					     gfp_t flags);
500int skb_tunnel_check_pmtu(struct sk_buff *skb, struct dst_entry *encap_dst,
501			  int headroom, bool reply);
502
503int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
504
505static inline int iptunnel_pull_offloads(struct sk_buff *skb)
506{
507	if (skb_is_gso(skb)) {
508		int err;
509
510		err = skb_unclone(skb, GFP_ATOMIC);
511		if (unlikely(err))
512			return err;
513		skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
514					       NETIF_F_GSO_SHIFT);
515	}
516
517	skb->encapsulation = 0;
518	return 0;
519}
520
521static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
522{
523	if (pkt_len > 0) {
524		struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
525
526		u64_stats_update_begin(&tstats->syncp);
527		u64_stats_add(&tstats->tx_bytes, pkt_len);
528		u64_stats_inc(&tstats->tx_packets);
529		u64_stats_update_end(&tstats->syncp);
530		put_cpu_ptr(tstats);
531		return;
532	}
533
534	if (pkt_len < 0) {
535		DEV_STATS_INC(dev, tx_errors);
536		DEV_STATS_INC(dev, tx_aborted_errors);
537	} else {
538		DEV_STATS_INC(dev, tx_dropped);
539	}
540}
541
542static inline void ip_tunnel_info_opts_get(void *to,
543					   const struct ip_tunnel_info *info)
544{
545	memcpy(to, info + 1, info->options_len);
546}
547
548static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
549					   const void *from, int len,
550					   __be16 flags)
551{
552	info->options_len = len;
553	if (len > 0) {
554		memcpy(ip_tunnel_info_opts(info), from, len);
555		info->key.tun_flags |= flags;
556	}
557}
558
559static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
560{
561	return (struct ip_tunnel_info *)lwtstate->data;
562}
563
564DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
565
566/* Returns > 0 if metadata should be collected */
567static inline int ip_tunnel_collect_metadata(void)
568{
569	return static_branch_unlikely(&ip_tunnel_metadata_cnt);
570}
571
572void __init ip_tunnel_core_init(void);
573
574void ip_tunnel_need_metadata(void);
575void ip_tunnel_unneed_metadata(void);
576
577#else /* CONFIG_INET */
578
579static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
580{
581	return NULL;
582}
583
584static inline void ip_tunnel_need_metadata(void)
585{
586}
587
588static inline void ip_tunnel_unneed_metadata(void)
589{
590}
591
592static inline void ip_tunnel_info_opts_get(void *to,
593					   const struct ip_tunnel_info *info)
594{
595}
596
597static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
598					   const void *from, int len,
599					   __be16 flags)
600{
601	info->options_len = 0;
602}
603
604#endif /* CONFIG_INET */
605
606#endif /* __NET_IP_TUNNELS_H */
607