[thirdparty/kernel/stable.git] / include / net / ip_tunnels.h

/* SPDX-License-Identifier: GPL-2.0 */
#ifndef __NET_IP_TUNNELS_H
#define __NET_IP_TUNNELS_H 1

#include <linux/if_tunnel.h>
#include <linux/netdevice.h>
#include <linux/skbuff.h>
#include <linux/socket.h>
#include <linux/types.h>
#include <linux/u64_stats_sync.h>
#include <linux/bitops.h>

#include <net/dsfield.h>
#include <net/gro_cells.h>
#include <net/inet_ecn.h>
#include <net/netns/generic.h>
#include <net/rtnetlink.h>
#include <net/lwtunnel.h>
#include <net/dst_cache.h>

#if IS_ENABLED(CONFIG_IPV6)
#include <net/ipv6.h>
#include <net/ip6_fib.h>
#include <net/ip6_route.h>
#endif

/* Keep error state on tunnel for 30 sec */
#define IPTUNNEL_ERR_TIMEO	(30*HZ)

/* Used to memset ip_tunnel padding. */
#define IP_TUNNEL_KEY_SIZE	offsetofend(struct ip_tunnel_key, tp_dst)

/* Used to memset ipv4 address padding. */
#define IP_TUNNEL_KEY_IPV4_PAD	offsetofend(struct ip_tunnel_key, u.ipv4.dst)
#define IP_TUNNEL_KEY_IPV4_PAD_LEN				\
	(FIELD_SIZEOF(struct ip_tunnel_key, u) -		\
	 FIELD_SIZEOF(struct ip_tunnel_key, u.ipv4))

struct ip_tunnel_key {
	__be64			tun_id;
	union {
		struct {
			__be32	src;
			__be32	dst;
		} ipv4;
		struct {
			struct in6_addr src;
			struct in6_addr dst;
		} ipv6;
	} u;
	__be16			tun_flags;
	u8			tos;		/* TOS for IPv4, TC for IPv6 */
	u8			ttl;		/* TTL for IPv4, HL for IPv6 */
	__be32			label;		/* Flow Label for IPv6 */
	__be16			tp_src;
	__be16			tp_dst;
};

/* Flags for ip_tunnel_info mode. */
#define IP_TUNNEL_INFO_TX	0x01	/* represents tx tunnel parameters */
#define IP_TUNNEL_INFO_IPV6	0x02	/* key contains IPv6 addresses */
#define IP_TUNNEL_INFO_BRIDGE	0x04	/* represents a bridged tunnel id */

/* Maximum tunnel options length. */
#define IP_TUNNEL_OPTS_MAX					\
	GENMASK((FIELD_SIZEOF(struct ip_tunnel_info,		\
			      options_len) * BITS_PER_BYTE) - 1, 0)

struct ip_tunnel_info {
	struct ip_tunnel_key	key;
#ifdef CONFIG_DST_CACHE
	struct dst_cache	dst_cache;
#endif
	u8			options_len;
	u8			mode;
};

/* 6rd prefix/relay information */
#ifdef CONFIG_IPV6_SIT_6RD
struct ip_tunnel_6rd_parm {
	struct in6_addr		prefix;
	__be32			relay_prefix;
	u16			prefixlen;
	u16			relay_prefixlen;
};
#endif

struct ip_tunnel_encap {
	u16			type;
	u16			flags;
	__be16			sport;
	__be16			dport;
};

struct ip_tunnel_prl_entry {
	struct ip_tunnel_prl_entry __rcu *next;
	__be32				addr;
	u16				flags;
	struct rcu_head			rcu_head;
};

struct metadata_dst;

struct ip_tunnel {
	struct ip_tunnel __rcu	*next;
	struct hlist_node hash_node;
	struct net_device	*dev;
	struct net		*net;	/* netns for packet i/o */

	unsigned long	err_time;	/* Time when the last ICMP error
					 * arrived */
	int		err_count;	/* Number of arrived ICMP errors */

	/* These four fields used only by GRE */
	u32		i_seqno;	/* The last seen seqno	*/
	u32		o_seqno;	/* The last output seqno */
	int		tun_hlen;	/* Precalculated header length */

	/* These four fields used only by ERSPAN */
	u32		index;		/* ERSPAN type II index */
	u8		erspan_ver;	/* ERSPAN version */
	u8		dir;		/* ERSPAN direction */
	u16		hwid;		/* ERSPAN hardware ID */

	struct dst_cache dst_cache;

	struct ip_tunnel_parm parms;

	int		mlink;
	int		encap_hlen;	/* Encap header length (FOU,GUE) */
	int		hlen;		/* tun_hlen + encap_hlen */
	struct ip_tunnel_encap encap;

	/* for SIT */
#ifdef CONFIG_IPV6_SIT_6RD
	struct ip_tunnel_6rd_parm ip6rd;
#endif
	struct ip_tunnel_prl_entry __rcu *prl;	/* potential router list */
	unsigned int		prl_count;	/* # of entries in PRL */
	unsigned int		ip_tnl_net_id;
	struct gro_cells	gro_cells;
	__u32			fwmark;
	bool			collect_md;
	bool			ignore_df;
};

struct tnl_ptk_info {
	__be16 flags;
	__be16 proto;
	__be32 key;
	__be32 seq;
	int hdr_len;
};

#define PACKET_RCVD	0
#define PACKET_REJECT	1
#define PACKET_NEXT	2

#define IP_TNL_HASH_BITS   7
#define IP_TNL_HASH_SIZE   (1 << IP_TNL_HASH_BITS)

struct ip_tunnel_net {
	struct net_device *fb_tunnel_dev;
	struct rtnl_link_ops *rtnl_link_ops;
	struct hlist_head tunnels[IP_TNL_HASH_SIZE];
	struct ip_tunnel __rcu *collect_md_tun;
	int type;
};

static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
				      __be32 saddr, __be32 daddr,
				      u8 tos, u8 ttl, __be32 label,
				      __be16 tp_src, __be16 tp_dst,
				      __be64 tun_id, __be16 tun_flags)
{
	key->tun_id = tun_id;
	key->u.ipv4.src = saddr;
	key->u.ipv4.dst = daddr;
	memset((unsigned char *)key + IP_TUNNEL_KEY_IPV4_PAD,
	       0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
	key->tos = tos;
	key->ttl = ttl;
	key->label = label;
	key->tun_flags = tun_flags;

	/* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
	 * the upper tunnel are used.
	 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
	 */
	key->tp_src = tp_src;
	key->tp_dst = tp_dst;

	/* Clear struct padding. */
	if (sizeof(*key) != IP_TUNNEL_KEY_SIZE)
		memset((unsigned char *)key + IP_TUNNEL_KEY_SIZE,
		       0, sizeof(*key) - IP_TUNNEL_KEY_SIZE);
}

static inline bool
ip_tunnel_dst_cache_usable(const struct sk_buff *skb,
			   const struct ip_tunnel_info *info)
{
	if (skb->mark)
		return false;
	if (!info)
		return true;
	if (info->key.tun_flags & TUNNEL_NOCACHE)
		return false;

	return true;
}

static inline unsigned short ip_tunnel_info_af(const struct ip_tunnel_info
					       *tun_info)
{
	return tun_info->mode & IP_TUNNEL_INFO_IPV6 ? AF_INET6 : AF_INET;
}

static inline __be64 key32_to_tunnel_id(__be32 key)
{
#ifdef __BIG_ENDIAN
	return (__force __be64)key;
#else
	return (__force __be64)((__force u64)key << 32);
#endif
}

/* Returns the least-significant 32 bits of a __be64. */
static inline __be32 tunnel_id_to_key32(__be64 tun_id)
{
#ifdef __BIG_ENDIAN
	return (__force __be32)tun_id;
#else
	return (__force __be32)((__force u64)tun_id >> 32);
#endif
}

#ifdef CONFIG_INET

static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
				       int proto,
				       __be32 daddr, __be32 saddr,
				       __be32 key, __u8 tos, int oif,
				       __u32 mark, __u32 tun_inner_hash)
{
	memset(fl4, 0, sizeof(*fl4));
	fl4->flowi4_oif = oif;
	fl4->daddr = daddr;
	fl4->saddr = saddr;
	fl4->flowi4_tos = tos;
	fl4->flowi4_proto = proto;
	fl4->fl4_gre_key = key;
	fl4->flowi4_mark = mark;
	fl4->flowi4_multipath_hash = tun_inner_hash;
}

int ip_tunnel_init(struct net_device *dev);
void ip_tunnel_uninit(struct net_device *dev);
void  ip_tunnel_dellink(struct net_device *dev, struct list_head *head);
struct net *ip_tunnel_get_link_net(const struct net_device *dev);
int ip_tunnel_get_iflink(const struct net_device *dev);
int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
		       struct rtnl_link_ops *ops, char *devname);

void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
			   struct rtnl_link_ops *ops);

void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
		    const struct iphdr *tnl_params, const u8 protocol);
void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
		       const u8 proto, int tunnel_hlen);
int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd);
int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);

void ip_tunnel_get_stats64(struct net_device *dev,
			   struct rtnl_link_stats64 *tot);
struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
				   int link, __be16 flags,
				   __be32 remote, __be32 local,
				   __be32 key);

int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
		  bool log_ecn_error);
int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
			 struct ip_tunnel_parm *p, __u32 fwmark);
int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
		      struct ip_tunnel_parm *p, __u32 fwmark);
void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);

struct ip_tunnel_encap_ops {
	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
			    u8 *protocol, struct flowi4 *fl4);
	int (*err_handler)(struct sk_buff *skb, u32 info);
};

#define MAX_IPTUN_ENCAP_OPS 8

extern const struct ip_tunnel_encap_ops __rcu *
		iptun_encaps[MAX_IPTUN_ENCAP_OPS];

int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
			    unsigned int num);
int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
			    unsigned int num);

int ip_tunnel_encap_setup(struct ip_tunnel *t,
			  struct ip_tunnel_encap *ipencap);

static inline bool pskb_inet_may_pull(struct sk_buff *skb)
{
	int nhlen;

	switch (skb->protocol) {
#if IS_ENABLED(CONFIG_IPV6)
	case htons(ETH_P_IPV6):
		nhlen = sizeof(struct ipv6hdr);
		break;
#endif
	case htons(ETH_P_IP):
		nhlen = sizeof(struct iphdr);
		break;
	default:
		nhlen = 0;
	}

	return pskb_network_may_pull(skb, nhlen);
}

static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
{
	const struct ip_tunnel_encap_ops *ops;
	int hlen = -EINVAL;

	if (e->type == TUNNEL_ENCAP_NONE)
		return 0;

	if (e->type >= MAX_IPTUN_ENCAP_OPS)
		return -EINVAL;

	rcu_read_lock();
	ops = rcu_dereference(iptun_encaps[e->type]);
	if (likely(ops && ops->encap_hlen))
		hlen = ops->encap_hlen(e);
	rcu_read_unlock();

	return hlen;
}

static inline int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
				  u8 *protocol, struct flowi4 *fl4)
{
	const struct ip_tunnel_encap_ops *ops;
	int ret = -EINVAL;

	if (t->encap.type == TUNNEL_ENCAP_NONE)
		return 0;

	if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
		return -EINVAL;

	rcu_read_lock();
	ops = rcu_dereference(iptun_encaps[t->encap.type]);
	if (likely(ops && ops->build_header))
		ret = ops->build_header(skb, &t->encap, protocol, fl4);
	rcu_read_unlock();

	return ret;
}

/* Extract dsfield from inner protocol */
static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
				       const struct sk_buff *skb)
{
	if (skb->protocol == htons(ETH_P_IP))
		return iph->tos;
	else if (skb->protocol == htons(ETH_P_IPV6))
		return ipv6_get_dsfield((const struct ipv6hdr *)iph);
	else
		return 0;
}

static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
				       const struct sk_buff *skb)
{
	if (skb->protocol == htons(ETH_P_IP))
		return iph->ttl;
	else if (skb->protocol == htons(ETH_P_IPV6))
		return ((const struct ipv6hdr *)iph)->hop_limit;
	else
		return 0;
}

/* Propogate ECN bits out */
static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
				     const struct sk_buff *skb)
{
	u8 inner = ip_tunnel_get_dsfield(iph, skb);

	return INET_ECN_encapsulate(tos, inner);
}

int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
			   __be16 inner_proto, bool raw_proto, bool xnet);

static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
				       __be16 inner_proto, bool xnet)
{
	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
}

void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
		   __be32 src, __be32 dst, u8 proto,
		   u8 tos, u8 ttl, __be16 df, bool xnet);
struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md,
					     gfp_t flags);

int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);

static inline int iptunnel_pull_offloads(struct sk_buff *skb)
{
	if (skb_is_gso(skb)) {
		int err;

		err = skb_unclone(skb, GFP_ATOMIC);
		if (unlikely(err))
			return err;
		skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
					       NETIF_F_GSO_SHIFT);
	}

	skb->encapsulation = 0;
	return 0;
}

static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
{
	if (pkt_len > 0) {
		struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);

		u64_stats_update_begin(&tstats->syncp);
		tstats->tx_bytes += pkt_len;
		tstats->tx_packets++;
		u64_stats_update_end(&tstats->syncp);
		put_cpu_ptr(tstats);
	} else {
		struct net_device_stats *err_stats = &dev->stats;

		if (pkt_len < 0) {
			err_stats->tx_errors++;
			err_stats->tx_aborted_errors++;
		} else {
			err_stats->tx_dropped++;
		}
	}
}

static inline void *ip_tunnel_info_opts(struct ip_tunnel_info *info)
{
	return info + 1;
}

static inline void ip_tunnel_info_opts_get(void *to,
					   const struct ip_tunnel_info *info)
{
	memcpy(to, info + 1, info->options_len);
}

static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
					   const void *from, int len,
					   __be16 flags)
{
	memcpy(ip_tunnel_info_opts(info), from, len);
	info->options_len = len;
	info->key.tun_flags |= flags;
}

static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
{
	return (struct ip_tunnel_info *)lwtstate->data;
}

DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);

/* Returns > 0 if metadata should be collected */
static inline int ip_tunnel_collect_metadata(void)
{
	return static_branch_unlikely(&ip_tunnel_metadata_cnt);
}

void __init ip_tunnel_core_init(void);

void ip_tunnel_need_metadata(void);
void ip_tunnel_unneed_metadata(void);

#else /* CONFIG_INET */

static inline struct ip_tunnel_info *lwt_tun_info(struct lwtunnel_state *lwtstate)
{
	return NULL;
}

static inline void ip_tunnel_need_metadata(void)
{
}

static inline void ip_tunnel_unneed_metadata(void)
{
}

static inline void ip_tunnel_info_opts_get(void *to,
					   const struct ip_tunnel_info *info)
{
}

static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
					   const void *from, int len,
					   __be16 flags)
{
	info->options_len = 0;
	info->key.tun_flags |= flags;
}

#endif /* CONFIG_INET */

#endif /* __NET_IP_TUNNELS_H */
Commit	Line	Data
b2441318	1	/* SPDX-License-Identifier: GPL-2.0 */
c5441932 PS	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>
7f9562a1	8	#include <linux/socket.h>
c5441932 PS	9	#include <linux/types.h>
c5441932 PS	10	#include <linux/u64_stats_sync.h>
fca5fdf6 DB	11	#include <linux/bitops.h>
fca5fdf6 DB	12
c5441932 PS	13	#include <net/dsfield.h>
	14	#include <net/gro_cells.h>
	15	#include <net/inet_ecn.h>
56328486	16	#include <net/netns/generic.h>
c5441932	17	#include <net/rtnetlink.h>
3093fbe7	18	#include <net/lwtunnel.h>
e09acddf	19	#include <net/dst_cache.h>
c5441932 PS	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
1d8fff90	30	/* Used to memset ip_tunnel padding. */
376534a3	31	#define IP_TUNNEL_KEY_SIZE offsetofend(struct ip_tunnel_key, tp_dst)
1d8fff90	32
c1ea5d67 JB	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	(FIELD_SIZEOF(struct ip_tunnel_key, u) - \
	37	FIELD_SIZEOF(struct ip_tunnel_key, u.ipv4))
	38
1d8fff90 TG	39	struct ip_tunnel_key {
1d8fff90 TG	40	__be64 tun_id;
c1ea5d67 JB	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;
1d8fff90	51	__be16 tun_flags;
7c383fb2 JB	52	u8 tos; /* TOS for IPv4, TC for IPv6 */
7c383fb2 JB	53	u8 ttl; /* TTL for IPv4, HL for IPv6 */
13461144	54	__be32 label; /* Flow Label for IPv6 */
1d8fff90 TG	55	__be16 tp_src;
1d8fff90 TG	56	__be16 tp_dst;
ac1cf399	57	};
1d8fff90	58
46fa062a JB	59	/* Flags for ip_tunnel_info mode. */
46fa062a JB	60	#define IP_TUNNEL_INFO_TX 0x01 /* represents tx tunnel parameters */
7f9562a1	61	#define IP_TUNNEL_INFO_IPV6 0x02 /* key contains IPv6 addresses */
f35581d6	62	#define IP_TUNNEL_INFO_BRIDGE 0x04 /* represents a bridged tunnel id */
ee122c79	63
fca5fdf6 DB	64	/* Maximum tunnel options length. */
	65	#define IP_TUNNEL_OPTS_MAX \
	66	GENMASK((FIELD_SIZEOF(struct ip_tunnel_info, \
	67	options_len) * BITS_PER_BYTE) - 1, 0)
	68
1d8fff90 TG	69	struct ip_tunnel_info {
1d8fff90 TG	70	struct ip_tunnel_key key;
d71785ff PA	71	#ifdef CONFIG_DST_CACHE
	72	struct dst_cache dst_cache;
	73	#endif
1d8fff90	74	u8 options_len;
ee122c79	75	u8 mode;
1d8fff90 TG	76	};
1d8fff90 TG	77
c5441932 PS	78	/* 6rd prefix/relay information */
	79	#ifdef CONFIG_IPV6_SIT_6RD
	80	struct ip_tunnel_6rd_parm {
	81	struct in6_addr prefix;
	82	__be32 relay_prefix;
	83	u16 prefixlen;
	84	u16 relay_prefixlen;
	85	};
	86	#endif
	87
56328486	88	struct ip_tunnel_encap {
6b8847c5 JB	89	u16 type;
6b8847c5 JB	90	u16 flags;
56328486 TH	91	__be16 sport;
	92	__be16 dport;
	93	};
	94
c5441932 PS	95	struct 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
2e15ea39 PS	102	struct metadata_dst;
2e15ea39 PS	103
c5441932 PS	104	struct ip_tunnel {
	105	struct ip_tunnel __rcu *next;
	106	struct hlist_node hash_node;
	107	struct net_device *dev;
5e6700b3	108	struct net net; / netns for packet i/o */
c5441932	109
c5441932 PS	110	unsigned long err_time; /* Time when the last ICMP error
c5441932 PS	111	* arrived */
f38ba953	112	int err_count; /* Number of arrived ICMP errors */
c5441932 PS	113
c5441932 PS	114	/* These four fields used only by GRE */
6b8847c5 JB	115	u32 i_seqno; /* The last seen seqno */
6b8847c5 JB	116	u32 o_seqno; /* The last output seqno */
56328486	117	int tun_hlen; /* Precalculated header length */
c5441932	118
f551c91d	119	/* These four fields used only by ERSPAN */
84e54fe0	120	u32 index; /* ERSPAN type II index */
f551c91d WT	121	u8 erspan_ver; /* ERSPAN version */
	122	u8 dir; /* ERSPAN direction */
	123	u16 hwid; /* ERSPAN hardware ID */
84e54fe0	124
e09acddf	125	struct dst_cache dst_cache;
7d442fab	126
c5441932 PS	127	struct ip_tunnel_parm parms;
c5441932 PS	128
f38ba953	129	int mlink;
56328486	130	int encap_hlen; /* Encap header length (FOU,GUE) */
56328486	131	int hlen; /* tun_hlen + encap_hlen */
f38ba953	132	struct ip_tunnel_encap encap;
56328486	133
c5441932 PS	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 */
c7d03a00	140	unsigned int ip_tnl_net_id;
c5441932	141	struct gro_cells gro_cells;
9830ad4c	142	__u32 fwmark;
2e15ea39	143	bool collect_md;
22a59be8	144	bool ignore_df;
c5441932 PS	145	};
c5441932 PS	146
c5441932 PS	147	struct tnl_ptk_info {
	148	__be16 flags;
	149	__be16 proto;
	150	__be32 key;
	151	__be32 seq;
9b8c6d7b	152	int hdr_len;
c5441932 PS	153	};
	154
	155	#define PACKET_RCVD 0
	156	#define PACKET_REJECT 1
125372fa	157	#define PACKET_NEXT 2
c5441932	158
6261d983	159	#define IP_TNL_HASH_BITS 7
c5441932 PS	160	#define IP_TNL_HASH_SIZE (1 << IP_TNL_HASH_BITS)
	161
	162	struct ip_tunnel_net {
c5441932	163	struct net_device *fb_tunnel_dev;
79134e6c	164	struct rtnl_link_ops *rtnl_link_ops;
6261d983	165	struct hlist_head tunnels[IP_TNL_HASH_SIZE];
2e15ea39	166	struct ip_tunnel __rcu *collect_md_tun;
79134e6c	167	int type;
c5441932 PS	168	};
c5441932 PS	169
4c222798 PS	170	static inline void ip_tunnel_key_init(struct ip_tunnel_key *key,
4c222798 PS	171	__be32 saddr, __be32 daddr,
13461144	172	u8 tos, u8 ttl, __be32 label,
4c222798 PS	173	__be16 tp_src, __be16 tp_dst,
4c222798 PS	174	__be64 tun_id, __be16 tun_flags)
1d8fff90	175	{
4c222798 PS	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,
c1ea5d67	180	0, IP_TUNNEL_KEY_IPV4_PAD_LEN);
4c222798 PS	181	key->tos = tos;
4c222798 PS	182	key->ttl = ttl;
13461144	183	key->label = label;
4c222798	184	key->tun_flags = tun_flags;
1d8fff90 TG	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	*/
4c222798 PS	190	key->tp_src = tp_src;
4c222798 PS	191	key->tp_dst = tp_dst;
1d8fff90 TG	192
1d8fff90 TG	193	/* Clear struct padding. */
4c222798 PS	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);
1d8fff90 TG	197	}
1d8fff90 TG	198
db3c6139 DB	199	static inline bool
	200	ip_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
7f9562a1 JB	213	static 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
d817f432 AV	219	static 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. */
	229	static 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
5243b6ac JG	238	#ifdef CONFIG_INET
5243b6ac JG	239
b0066da5 PM	240	static inline void ip_tunnel_init_flow(struct flowi4 *fl4,
	241	int proto,
	242	__be32 daddr, __be32 saddr,
	243	__be32 key, __u8 tos, int oif,
24ba1440	244	__u32 mark, __u32 tun_inner_hash)
b0066da5 PM	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;
24ba1440	254	fl4->flowi4_multipath_hash = tun_inner_hash;
b0066da5 PM	255	}
b0066da5 PM	256
c5441932 PS	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);
1728d4fa	260	struct net ip_tunnel_get_link_net(const struct net_device dev);
1e99584b	261	int ip_tunnel_get_iflink(const struct net_device *dev);
c7d03a00	262	int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
d3b6f614	263	struct rtnl_link_ops ops, char devname);
c5441932	264
64bc1781 ED	265	void ip_tunnel_delete_nets(struct list_head *list_net, unsigned int id,
64bc1781 ED	266	struct rtnl_link_ops *ops);
c5441932 PS	267
c5441932 PS	268	void ip_tunnel_xmit(struct sk_buff skb, struct net_device dev,
bf3d6a8f	269	const struct iphdr *tnl_params, const u8 protocol);
cfc7381b	270	void ip_md_tunnel_xmit(struct sk_buff skb, struct net_device dev,
c8b34e68	271	const u8 proto, int tunnel_hlen);
c5441932	272	int ip_tunnel_ioctl(struct net_device dev, struct ip_tunnel_parm p, int cmd);
7e059158	273	int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict);
c5441932 PS	274	int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu);
c5441932 PS	275
bc1f4470	276	void ip_tunnel_get_stats64(struct net_device *dev,
bc1f4470	277	struct rtnl_link_stats64 *tot);
c5441932 PS	278	struct ip_tunnel ip_tunnel_lookup(struct ip_tunnel_net itn,
	279	int link, __be16 flags,
	280	__be32 remote, __be32 local,
	281	__be32 key);
	282
	283	int ip_tunnel_rcv(struct ip_tunnel tunnel, struct sk_buff skb,
2e15ea39 PS	284	const struct tnl_ptk_info tpi, struct metadata_dst tun_dst,
2e15ea39 PS	285	bool log_ecn_error);
c5441932	286	int ip_tunnel_changelink(struct net_device dev, struct nlattr tb[],
9830ad4c	287	struct ip_tunnel_parm *p, __u32 fwmark);
c5441932	288	int ip_tunnel_newlink(struct net_device dev, struct nlattr tb[],
9830ad4c	289	struct ip_tunnel_parm *p, __u32 fwmark);
c7d03a00	290	void ip_tunnel_setup(struct net_device *dev, unsigned int net_id);
55c2bc14 TH	291
	292	struct ip_tunnel_encap_ops {
	293	size_t (encap_hlen)(struct ip_tunnel_encap e);
	294	int (build_header)(struct sk_buff skb, struct ip_tunnel_encap *e,
	295	u8 protocol, struct flowi4 fl4);
e7cc0824	296	int (err_handler)(struct sk_buff skb, u32 info);
55c2bc14 TH	297	};
	298
	299	#define MAX_IPTUN_ENCAP_OPS 8
	300
	301	extern const struct ip_tunnel_encap_ops __rcu *
	302	iptun_encaps[MAX_IPTUN_ENCAP_OPS];
	303
	304	int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *op,
	305	unsigned int num);
	306	int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *op,
	307	unsigned int num);
	308
56328486 TH	309	int ip_tunnel_encap_setup(struct ip_tunnel *t,
56328486 TH	310	struct ip_tunnel_encap *ipencap);
c5441932	311
cb9f1b78 WB	312	static inline bool pskb_inet_may_pull(struct sk_buff *skb)
	313	{
	314	int nhlen;
	315
	316	switch (skb->protocol) {
	317	#if IS_ENABLED(CONFIG_IPV6)
	318	case htons(ETH_P_IPV6):
	319	nhlen = sizeof(struct ipv6hdr);
	320	break;
	321	#endif
	322	case htons(ETH_P_IP):
	323	nhlen = sizeof(struct iphdr);
	324	break;
	325	default:
	326	nhlen = 0;
	327	}
	328
	329	return pskb_network_may_pull(skb, nhlen);
	330	}
	331
55c2bc14 TH	332	static inline int ip_encap_hlen(struct ip_tunnel_encap *e)
	333	{
	334	const struct ip_tunnel_encap_ops *ops;
	335	int hlen = -EINVAL;
	336
	337	if (e->type == TUNNEL_ENCAP_NONE)
	338	return 0;
	339
	340	if (e->type >= MAX_IPTUN_ENCAP_OPS)
	341	return -EINVAL;
	342
	343	rcu_read_lock();
	344	ops = rcu_dereference(iptun_encaps[e->type]);
	345	if (likely(ops && ops->encap_hlen))
	346	hlen = ops->encap_hlen(e);
	347	rcu_read_unlock();
	348
	349	return hlen;
	350	}
	351
	352	static inline int ip_tunnel_encap(struct sk_buff skb, struct ip_tunnel t,
	353	u8 protocol, struct flowi4 fl4)
	354	{
	355	const struct ip_tunnel_encap_ops *ops;
	356	int ret = -EINVAL;
	357
	358	if (t->encap.type == TUNNEL_ENCAP_NONE)
	359	return 0;
	360
	361	if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
	362	return -EINVAL;
	363
	364	rcu_read_lock();
	365	ops = rcu_dereference(iptun_encaps[t->encap.type]);
	366	if (likely(ops && ops->build_header))
	367	ret = ops->build_header(skb, &t->encap, protocol, fl4);
	368	rcu_read_unlock();
	369
	370	return ret;
	371	}
	372
c5441932 PS	373	/* Extract dsfield from inner protocol */
	374	static inline u8 ip_tunnel_get_dsfield(const struct iphdr *iph,
	375	const struct sk_buff *skb)
	376	{
	377	if (skb->protocol == htons(ETH_P_IP))
	378	return iph->tos;
	379	else if (skb->protocol == htons(ETH_P_IPV6))
	380	return ipv6_get_dsfield((const struct ipv6hdr *)iph);
	381	else
	382	return 0;
	383	}
	384
72f6d71e HL	385	static inline u8 ip_tunnel_get_ttl(const struct iphdr *iph,
	386	const struct sk_buff *skb)
	387	{
	388	if (skb->protocol == htons(ETH_P_IP))
	389	return iph->ttl;
	390	else if (skb->protocol == htons(ETH_P_IPV6))
	391	return ((const struct ipv6hdr *)iph)->hop_limit;
	392	else
	393	return 0;
	394	}
	395
c5441932 PS	396	/* Propogate ECN bits out */
	397	static inline u8 ip_tunnel_ecn_encap(u8 tos, const struct iphdr *iph,
	398	const struct sk_buff *skb)
	399	{
	400	u8 inner = ip_tunnel_get_dsfield(iph, skb);
	401
	402	return INET_ECN_encapsulate(tos, inner);
	403	}
	404
a6d5bbf3 JB	405	int __iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
	406	__be16 inner_proto, bool raw_proto, bool xnet);
	407
	408	static inline int iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
	409	__be16 inner_proto, bool xnet)
	410	{
	411	return __iptunnel_pull_header(skb, hdr_len, inner_proto, false, xnet);
	412	}
	413
039f5062 PS	414	void iptunnel_xmit(struct sock sk, struct rtable rt, struct sk_buff *skb,
	415	__be32 src, __be32 dst, u8 proto,
	416	u8 tos, u8 ttl, __be16 df, bool xnet);
63d008a4 JB	417	struct metadata_dst iptunnel_metadata_reply(struct metadata_dst md,
63d008a4 JB	418	gfp_t flags);
0e6fbc5b	419
aed069df	420	int iptunnel_handle_offloads(struct sk_buff *skb, int gso_type_mask);
2d26f0a3	421
a09a4c8d JG	422	static inline int iptunnel_pull_offloads(struct sk_buff *skb)
	423	{
	424	if (skb_is_gso(skb)) {
	425	int err;
	426
	427	err = skb_unclone(skb, GFP_ATOMIC);
	428	if (unlikely(err))
	429	return err;
	430	skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >>
	431	NETIF_F_GSO_SHIFT);
	432	}
	433
	434	skb->encapsulation = 0;
	435	return 0;
	436	}
	437
039f5062	438	static inline void iptunnel_xmit_stats(struct net_device *dev, int pkt_len)
c5441932	439	{
039f5062 PS	440	if (pkt_len > 0) {
039f5062 PS	441	struct pcpu_sw_netstats *tstats = get_cpu_ptr(dev->tstats);
c5441932	442
c5441932	443	u64_stats_update_begin(&tstats->syncp);
039f5062	444	tstats->tx_bytes += pkt_len;
c5441932 PS	445	tstats->tx_packets++;
c5441932 PS	446	u64_stats_update_end(&tstats->syncp);
b4fe85f9	447	put_cpu_ptr(tstats);
c5441932	448	} else {
039f5062 PS	449	struct net_device_stats *err_stats = &dev->stats;
	450
	451	if (pkt_len < 0) {
	452	err_stats->tx_errors++;
	453	err_stats->tx_aborted_errors++;
	454	} else {
	455	err_stats->tx_dropped++;
	456	}
c5441932 PS	457	}
c5441932 PS	458	}
5243b6ac	459
4c222798	460	static inline void ip_tunnel_info_opts(struct ip_tunnel_info info)
ee122c79 TG	461	{
	462	return info + 1;
	463	}
	464
4c222798 PS	465	static inline void ip_tunnel_info_opts_get(void *to,
	466	const struct ip_tunnel_info *info)
	467	{
	468	memcpy(to, info + 1, info->options_len);
	469	}
	470
	471	static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
256c87c1 PJV	472	const void *from, int len,
256c87c1 PJV	473	__be16 flags)
4c222798 PS	474	{
	475	memcpy(ip_tunnel_info_opts(info), from, len);
	476	info->options_len = len;
256c87c1	477	info->key.tun_flags \|= flags;
4c222798 PS	478	}
4c222798 PS	479
3093fbe7 TG	480	static inline struct ip_tunnel_info lwt_tun_info(struct lwtunnel_state lwtstate)
	481	{
	482	return (struct ip_tunnel_info *)lwtstate->data;
	483	}
	484
5263a98f	485	DECLARE_STATIC_KEY_FALSE(ip_tunnel_metadata_cnt);
e7030878 TG	486
	487	/* Returns > 0 if metadata should be collected */
	488	static inline int ip_tunnel_collect_metadata(void)
	489	{
5263a98f	490	return static_branch_unlikely(&ip_tunnel_metadata_cnt);
e7030878 TG	491	}
e7030878 TG	492
045a0fa0 TG	493	void __init ip_tunnel_core_init(void);
045a0fa0 TG	494
e7030878 TG	495	void ip_tunnel_need_metadata(void);
	496	void ip_tunnel_unneed_metadata(void);
	497
05283187 TG	498	#else /* CONFIG_INET */
	499
	500	static inline struct ip_tunnel_info lwt_tun_info(struct lwtunnel_state lwtstate)
	501	{
	502	return NULL;
	503	}
	504
	505	static inline void ip_tunnel_need_metadata(void)
	506	{
	507	}
	508
	509	static inline void ip_tunnel_unneed_metadata(void)
	510	{
	511	}
	512
e28e87ed DB	513	static inline void ip_tunnel_info_opts_get(void *to,
	514	const struct ip_tunnel_info *info)
	515	{
	516	}
	517
	518	static inline void ip_tunnel_info_opts_set(struct ip_tunnel_info *info,
256c87c1 PJV	519	const void *from, int len,
256c87c1 PJV	520	__be16 flags)
e28e87ed DB	521	{
e28e87ed DB	522	info->options_len = 0;
256c87c1	523	info->key.tun_flags \|= flags;
e28e87ed DB	524	}
e28e87ed DB	525
5243b6ac JG	526	#endif /* CONFIG_INET */
5243b6ac JG	527
c5441932	528	#endif /* __NET_IP_TUNNELS_H */