[linux-block.git] / include / net / vxlan.h

#ifndef __NET_VXLAN_H
#define __NET_VXLAN_H 1

#include <linux/if_vlan.h>
#include <net/udp_tunnel.h>
#include <net/dst_metadata.h>

/* VXLAN protocol (RFC 7348) header:
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |R|R|R|R|I|R|R|R|               Reserved                        |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |                VXLAN Network Identifier (VNI) |   Reserved    |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *
 * I = VXLAN Network Identifier (VNI) present.
 */
struct vxlanhdr {
	__be32 vx_flags;
	__be32 vx_vni;
};

/* VXLAN header flags. */
#define VXLAN_HF_VNI	cpu_to_be32(BIT(27))

#define VXLAN_N_VID     (1u << 24)
#define VXLAN_VID_MASK  (VXLAN_N_VID - 1)
#define VXLAN_VNI_MASK	cpu_to_be32(VXLAN_VID_MASK << 8)
#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))

#define VNI_HASH_BITS	10
#define VNI_HASH_SIZE	(1<<VNI_HASH_BITS)
#define FDB_HASH_BITS	8
#define FDB_HASH_SIZE	(1<<FDB_HASH_BITS)

/* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |R|R|R|R|I|R|R|R|R|R|C|              Reserved                   |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |           VXLAN Network Identifier (VNI)      |O| Csum start  |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *
 * C = Remote checksum offload bit. When set indicates that the
 *     remote checksum offload data is present.
 *
 * O = Offset bit. Indicates the checksum offset relative to
 *     checksum start.
 *
 * Csum start = Checksum start divided by two.
 *
 * http://tools.ietf.org/html/draft-herbert-vxlan-rco
 */

/* VXLAN-RCO header flags. */
#define VXLAN_HF_RCO	cpu_to_be32(BIT(21))

/* Remote checksum offload header option */
#define VXLAN_RCO_MASK	cpu_to_be32(0x7f)  /* Last byte of vni field */
#define VXLAN_RCO_UDP	cpu_to_be32(0x80)  /* Indicate UDP RCO (TCP when not set *) */
#define VXLAN_RCO_SHIFT	1		   /* Left shift of start */
#define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
#define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)

/*
 * VXLAN Group Based Policy Extension (VXLAN_F_GBP):
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |G|R|R|R|I|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |                VXLAN Network Identifier (VNI) |   Reserved    |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *
 * G = Group Policy ID present.
 *
 * D = Don't Learn bit. When set, this bit indicates that the egress
 *     VTEP MUST NOT learn the source address of the encapsulated frame.
 *
 * A = Indicates that the group policy has already been applied to
 *     this packet. Policies MUST NOT be applied by devices when the
 *     A bit is set.
 *
 * https://tools.ietf.org/html/draft-smith-vxlan-group-policy
 */
struct vxlanhdr_gbp {
	u8	vx_flags;
#ifdef __LITTLE_ENDIAN_BITFIELD
	u8	reserved_flags1:3,
		policy_applied:1,
		reserved_flags2:2,
		dont_learn:1,
		reserved_flags3:1;
#elif defined(__BIG_ENDIAN_BITFIELD)
	u8	reserved_flags1:1,
		dont_learn:1,
		reserved_flags2:2,
		policy_applied:1,
		reserved_flags3:3;
#else
#error	"Please fix <asm/byteorder.h>"
#endif
	__be16	policy_id;
	__be32	vx_vni;
};

/* VXLAN-GBP header flags. */
#define VXLAN_HF_GBP	cpu_to_be32(BIT(31))

#define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP | cpu_to_be32(0xFFFFFF))

/* skb->mark mapping
 *
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |R|R|R|R|R|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 */
#define VXLAN_GBP_DONT_LEARN		(BIT(6) << 16)
#define VXLAN_GBP_POLICY_APPLIED	(BIT(3) << 16)
#define VXLAN_GBP_ID_MASK		(0xFFFF)

/*
 * VXLAN Generic Protocol Extension (VXLAN_F_GPE):
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |R|R|Ver|I|P|R|O|       Reserved                |Next Protocol  |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |                VXLAN Network Identifier (VNI) |   Reserved    |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *
 * Ver = Version. Indicates VXLAN GPE protocol version.
 *
 * P = Next Protocol Bit. The P bit is set to indicate that the
 *     Next Protocol field is present.
 *
 * O = OAM Flag Bit. The O bit is set to indicate that the packet
 *     is an OAM packet.
 *
 * Next Protocol = This 8 bit field indicates the protocol header
 * immediately following the VXLAN GPE header.
 *
 * https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
 */

struct vxlanhdr_gpe {
#if defined(__LITTLE_ENDIAN_BITFIELD)
	u8	oam_flag:1,
		reserved_flags1:1,
		np_applied:1,
		instance_applied:1,
		version:2,
reserved_flags2:2;
#elif defined(__BIG_ENDIAN_BITFIELD)
	u8	reserved_flags2:2,
		version:2,
		instance_applied:1,
		np_applied:1,
		reserved_flags1:1,
		oam_flag:1;
#endif
	u8	reserved_flags3;
	u8	reserved_flags4;
	u8	next_protocol;
	__be32	vx_vni;
};

/* VXLAN-GPE header flags. */
#define VXLAN_HF_VER	cpu_to_be32(BIT(29) | BIT(28))
#define VXLAN_HF_NP	cpu_to_be32(BIT(26))
#define VXLAN_HF_OAM	cpu_to_be32(BIT(24))

#define VXLAN_GPE_USED_BITS (VXLAN_HF_VER | VXLAN_HF_NP | VXLAN_HF_OAM | \
			     cpu_to_be32(0xff))

/* VXLAN-GPE header Next Protocol. */
#define VXLAN_GPE_NP_IPV4      0x01
#define VXLAN_GPE_NP_IPV6      0x02
#define VXLAN_GPE_NP_ETHERNET  0x03
#define VXLAN_GPE_NP_NSH       0x04

struct vxlan_metadata {
	u32		gbp;
};

/* per UDP socket information */
struct vxlan_sock {
	struct hlist_node hlist;
	struct socket	 *sock;
	struct hlist_head vni_list[VNI_HASH_SIZE];
	atomic_t	  refcnt;
	u32		  flags;
};

union vxlan_addr {
	struct sockaddr_in sin;
	struct sockaddr_in6 sin6;
	struct sockaddr sa;
};

struct vxlan_rdst {
	union vxlan_addr	 remote_ip;
	__be16			 remote_port;
	__be32			 remote_vni;
	u32			 remote_ifindex;
	struct list_head	 list;
	struct rcu_head		 rcu;
	struct dst_cache	 dst_cache;
};

struct vxlan_config {
	union vxlan_addr	remote_ip;
	union vxlan_addr	saddr;
	__be32			vni;
	int			remote_ifindex;
	int			mtu;
	__be16			dst_port;
	u16			port_min;
	u16			port_max;
	u8			tos;
	u8			ttl;
	__be32			label;
	u32			flags;
	unsigned long		age_interval;
	unsigned int		addrmax;
	bool			no_share;
};

/* Pseudo network device */
struct vxlan_dev {
	struct hlist_node hlist;	/* vni hash table */
	struct list_head  next;		/* vxlan's per namespace list */
	struct vxlan_sock *vn4_sock;	/* listening socket for IPv4 */
#if IS_ENABLED(CONFIG_IPV6)
	struct vxlan_sock *vn6_sock;	/* listening socket for IPv6 */
#endif
	struct net_device *dev;
	struct net	  *net;		/* netns for packet i/o */
	struct vxlan_rdst default_dst;	/* default destination */
	u32		  flags;	/* VXLAN_F_* in vxlan.h */

	struct timer_list age_timer;
	spinlock_t	  hash_lock;
	unsigned int	  addrcnt;
	struct gro_cells  gro_cells;

	struct vxlan_config	cfg;

	struct hlist_head fdb_head[FDB_HASH_SIZE];
};

#define VXLAN_F_LEARN			0x01
#define VXLAN_F_PROXY			0x02
#define VXLAN_F_RSC			0x04
#define VXLAN_F_L2MISS			0x08
#define VXLAN_F_L3MISS			0x10
#define VXLAN_F_IPV6			0x20
#define VXLAN_F_UDP_ZERO_CSUM_TX	0x40
#define VXLAN_F_UDP_ZERO_CSUM6_TX	0x80
#define VXLAN_F_UDP_ZERO_CSUM6_RX	0x100
#define VXLAN_F_REMCSUM_TX		0x200
#define VXLAN_F_REMCSUM_RX		0x400
#define VXLAN_F_GBP			0x800
#define VXLAN_F_REMCSUM_NOPARTIAL	0x1000
#define VXLAN_F_COLLECT_METADATA	0x2000
#define VXLAN_F_GPE			0x4000

/* Flags that are used in the receive path. These flags must match in
 * order for a socket to be shareable
 */
#define VXLAN_F_RCV_FLAGS		(VXLAN_F_GBP |			\
					 VXLAN_F_GPE |			\
					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
					 VXLAN_F_REMCSUM_RX |		\
					 VXLAN_F_REMCSUM_NOPARTIAL |	\
					 VXLAN_F_COLLECT_METADATA)

/* Flags that can be set together with VXLAN_F_GPE. */
#define VXLAN_F_ALLOWED_GPE		(VXLAN_F_GPE |			\
					 VXLAN_F_IPV6 |			\
					 VXLAN_F_UDP_ZERO_CSUM_TX |	\
					 VXLAN_F_UDP_ZERO_CSUM6_TX |	\
					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
					 VXLAN_F_COLLECT_METADATA)

struct net_device *vxlan_dev_create(struct net *net, const char *name,
				    u8 name_assign_type, struct vxlan_config *conf);

static inline __be16 vxlan_dev_dst_port(struct vxlan_dev *vxlan,
					unsigned short family)
{
#if IS_ENABLED(CONFIG_IPV6)
	if (family == AF_INET6)
		return inet_sk(vxlan->vn6_sock->sock->sk)->inet_sport;
#endif
	return inet_sk(vxlan->vn4_sock->sock->sk)->inet_sport;
}

static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
						     netdev_features_t features)
{
	u8 l4_hdr = 0;

	if (!skb->encapsulation)
		return features;

	switch (vlan_get_protocol(skb)) {
	case htons(ETH_P_IP):
		l4_hdr = ip_hdr(skb)->protocol;
		break;
	case htons(ETH_P_IPV6):
		l4_hdr = ipv6_hdr(skb)->nexthdr;
		break;
	default:
		return features;;
	}

	if ((l4_hdr == IPPROTO_UDP) &&
	    (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
	     skb->inner_protocol != htons(ETH_P_TEB) ||
	     (skb_inner_mac_header(skb) - skb_transport_header(skb) !=
	      sizeof(struct udphdr) + sizeof(struct vxlanhdr)) ||
	     (skb->ip_summed != CHECKSUM_NONE &&
	      !can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
		return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);

	return features;
}

/* IP header + UDP + VXLAN + Ethernet header */
#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
/* IPv6 header + UDP + VXLAN + Ethernet header */
#define VXLAN6_HEADROOM (40 + 8 + 8 + 14)

static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
{
	return (struct vxlanhdr *)(udp_hdr(skb) + 1);
}

static inline __be32 vxlan_vni(__be32 vni_field)
{
#if defined(__BIG_ENDIAN)
	return (__force __be32)((__force u32)vni_field >> 8);
#else
	return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
#endif
}

static inline __be32 vxlan_vni_field(__be32 vni)
{
#if defined(__BIG_ENDIAN)
	return (__force __be32)((__force u32)vni << 8);
#else
	return (__force __be32)((__force u32)vni >> 8);
#endif
}

static inline __be32 vxlan_tun_id_to_vni(__be64 tun_id)
{
#if defined(__BIG_ENDIAN)
	return (__force __be32)tun_id;
#else
	return (__force __be32)((__force u64)tun_id >> 32);
#endif
}

static inline __be64 vxlan_vni_to_tun_id(__be32 vni)
{
#if defined(__BIG_ENDIAN)
	return (__force __be64)vni;
#else
	return (__force __be64)((u64)(__force u32)vni << 32);
#endif
}

static inline size_t vxlan_rco_start(__be32 vni_field)
{
	return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
}

static inline size_t vxlan_rco_offset(__be32 vni_field)
{
	return (vni_field & VXLAN_RCO_UDP) ?
		offsetof(struct udphdr, check) :
		offsetof(struct tcphdr, check);
}

static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
{
	__be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);

	if (offset == offsetof(struct udphdr, check))
		vni_field |= VXLAN_RCO_UDP;
	return vni_field;
}

static inline void vxlan_get_rx_port(struct net_device *netdev)
{
	ASSERT_RTNL();
	call_netdevice_notifiers(NETDEV_OFFLOAD_PUSH_VXLAN, netdev);
}

static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
{
	return vs->sock->sk->sk_family;
}

#endif
Commit	Line	Data
012a5729 PS	1	#ifndef __NET_VXLAN_H
	2	#define __NET_VXLAN_H 1
	3
5f35227e	4	#include <linux/if_vlan.h>
86a98057	5	#include <net/udp_tunnel.h>
ee122c79	6	#include <net/dst_metadata.h>
012a5729	7
828788ac JB	8	/* VXLAN protocol (RFC 7348) header:
	9	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	10	* \|R\|R\|R\|R\|I\|R\|R\|R\| Reserved \|
	11	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	12	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	13	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	14	*
	15	* I = VXLAN Network Identifier (VNI) present.
	16	*/
	17	struct vxlanhdr {
	18	__be32 vx_flags;
	19	__be32 vx_vni;
	20	};
	21
	22	/* VXLAN header flags. */
54bfd872	23	#define VXLAN_HF_VNI cpu_to_be32(BIT(27))
828788ac JB	24
	25	#define VXLAN_N_VID (1u << 24)
	26	#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
54bfd872	27	#define VXLAN_VNI_MASK cpu_to_be32(VXLAN_VID_MASK << 8)
828788ac JB	28	#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
	29
	30	#define VNI_HASH_BITS 10
	31	#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
	32	#define FDB_HASH_BITS 8
	33	#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
	34
	35	/* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
	36	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	37	* \|R\|R\|R\|R\|I\|R\|R\|R\|R\|R\|C\| Reserved \|
	38	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	39	* \| VXLAN Network Identifier (VNI) \|O\| Csum start \|
	40	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	41	*
	42	* C = Remote checksum offload bit. When set indicates that the
	43	* remote checksum offload data is present.
	44	*
	45	* O = Offset bit. Indicates the checksum offset relative to
	46	* checksum start.
	47	*
	48	* Csum start = Checksum start divided by two.
	49	*
	50	* http://tools.ietf.org/html/draft-herbert-vxlan-rco
	51	*/
	52
	53	/* VXLAN-RCO header flags. */
54bfd872	54	#define VXLAN_HF_RCO cpu_to_be32(BIT(21))
828788ac JB	55
828788ac JB	56	/* Remote checksum offload header option */
54bfd872 JB	57	#define VXLAN_RCO_MASK cpu_to_be32(0x7f) /* Last byte of vni field */
	58	#define VXLAN_RCO_UDP cpu_to_be32(0x80) /* Indicate UDP RCO (TCP when not set ) /
	59	#define VXLAN_RCO_SHIFT 1 /* Left shift of start */
828788ac	60	#define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
54bfd872	61	#define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)
828788ac	62
3511494c	63	/*
828788ac	64	* VXLAN Group Based Policy Extension (VXLAN_F_GBP):
3511494c	65	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
828788ac	66	* \|G\|R\|R\|R\|I\|R\|R\|R\|R\|D\|R\|R\|A\|R\|R\|R\| Group Policy ID \|
3511494c TG	67	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	68	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	69	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	70	*
828788ac JB	71	* G = Group Policy ID present.
828788ac JB	72	*
3511494c TG	73	* D = Don't Learn bit. When set, this bit indicates that the egress
	74	* VTEP MUST NOT learn the source address of the encapsulated frame.
	75	*
	76	* A = Indicates that the group policy has already been applied to
	77	* this packet. Policies MUST NOT be applied by devices when the
	78	* A bit is set.
	79	*
828788ac	80	* https://tools.ietf.org/html/draft-smith-vxlan-group-policy
3511494c TG	81	*/
3511494c TG	82	struct vxlanhdr_gbp {
0e715d6f	83	u8 vx_flags;
3511494c	84	#ifdef __LITTLE_ENDIAN_BITFIELD
0e715d6f	85	u8 reserved_flags1:3,
3511494c TG	86	policy_applied:1,
	87	reserved_flags2:2,
	88	dont_learn:1,
	89	reserved_flags3:1;
	90	#elif defined(__BIG_ENDIAN_BITFIELD)
0e715d6f	91	u8 reserved_flags1:1,
3511494c TG	92	dont_learn:1,
	93	reserved_flags2:2,
	94	policy_applied:1,
	95	reserved_flags3:3;
	96	#else
	97	#error "Please fix <asm/byteorder.h>"
	98	#endif
	99	__be16 policy_id;
	100	__be32 vx_vni;
	101	};
	102
828788ac	103	/* VXLAN-GBP header flags. */
54bfd872	104	#define VXLAN_HF_GBP cpu_to_be32(BIT(31))
828788ac	105
54bfd872	106	#define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP \| cpu_to_be32(0xFFFFFF))
3511494c TG	107
	108	/* skb->mark mapping
	109	*
	110	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	111	* \|R\|R\|R\|R\|R\|R\|R\|R\|R\|D\|R\|R\|A\|R\|R\|R\| Group Policy ID \|
	112	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	113	*/
	114	#define VXLAN_GBP_DONT_LEARN (BIT(6) << 16)
	115	#define VXLAN_GBP_POLICY_APPLIED (BIT(3) << 16)
	116	#define VXLAN_GBP_ID_MASK (0xFFFF)
	117
e1e5314d JB	118	/*
	119	* VXLAN Generic Protocol Extension (VXLAN_F_GPE):
	120	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	121	* \|R\|R\|Ver\|I\|P\|R\|O\| Reserved \|Next Protocol \|
	122	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	123	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	124	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	125	*
	126	* Ver = Version. Indicates VXLAN GPE protocol version.
	127	*
	128	* P = Next Protocol Bit. The P bit is set to indicate that the
	129	* Next Protocol field is present.
	130	*
	131	* O = OAM Flag Bit. The O bit is set to indicate that the packet
	132	* is an OAM packet.
	133	*
	134	* Next Protocol = This 8 bit field indicates the protocol header
	135	* immediately following the VXLAN GPE header.
	136	*
	137	* https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
	138	*/
	139
	140	struct vxlanhdr_gpe {
	141	#if defined(__LITTLE_ENDIAN_BITFIELD)
	142	u8 oam_flag:1,
	143	reserved_flags1:1,
	144	np_applied:1,
	145	instance_applied:1,
	146	version:2,
	147	reserved_flags2:2;
	148	#elif defined(__BIG_ENDIAN_BITFIELD)
	149	u8 reserved_flags2:2,
	150	version:2,
	151	instance_applied:1,
	152	np_applied:1,
	153	reserved_flags1:1,
	154	oam_flag:1;
	155	#endif
	156	u8 reserved_flags3;
	157	u8 reserved_flags4;
	158	u8 next_protocol;
	159	__be32 vx_vni;
	160	};
	161
	162	/* VXLAN-GPE header flags. */
	163	#define VXLAN_HF_VER cpu_to_be32(BIT(29) \| BIT(28))
	164	#define VXLAN_HF_NP cpu_to_be32(BIT(26))
	165	#define VXLAN_HF_OAM cpu_to_be32(BIT(24))
	166
	167	#define VXLAN_GPE_USED_BITS (VXLAN_HF_VER \| VXLAN_HF_NP \| VXLAN_HF_OAM \| \
	168	cpu_to_be32(0xff))
	169
	170	/* VXLAN-GPE header Next Protocol. */
	171	#define VXLAN_GPE_NP_IPV4 0x01
	172	#define VXLAN_GPE_NP_IPV6 0x02
	173	#define VXLAN_GPE_NP_ETHERNET 0x03
	174	#define VXLAN_GPE_NP_NSH 0x04
	175
3511494c	176	struct vxlan_metadata {
3511494c TG	177	u32 gbp;
	178	};
	179
012a5729 PS	180	/* per UDP socket information */
	181	struct vxlan_sock {
	182	struct hlist_node hlist;
012a5729	183	struct socket *sock;
012a5729 PS	184	struct hlist_head vni_list[VNI_HASH_SIZE];
012a5729 PS	185	atomic_t refcnt;
dfd8645e	186	u32 flags;
012a5729 PS	187	};
012a5729 PS	188
0dfbdf41 TG	189	union vxlan_addr {
	190	struct sockaddr_in sin;
	191	struct sockaddr_in6 sin6;
	192	struct sockaddr sa;
	193	};
	194
	195	struct vxlan_rdst {
	196	union vxlan_addr remote_ip;
	197	__be16 remote_port;
54bfd872	198	__be32 remote_vni;
0dfbdf41 TG	199	u32 remote_ifindex;
	200	struct list_head list;
	201	struct rcu_head rcu;
0c1d70af	202	struct dst_cache dst_cache;
0dfbdf41 TG	203	};
	204
	205	struct vxlan_config {
	206	union vxlan_addr remote_ip;
	207	union vxlan_addr saddr;
54bfd872	208	__be32 vni;
0dfbdf41 TG	209	int remote_ifindex;
	210	int mtu;
	211	__be16 dst_port;
0e715d6f JB	212	u16 port_min;
	213	u16 port_max;
	214	u8 tos;
	215	u8 ttl;
e7f70af1	216	__be32 label;
0dfbdf41 TG	217	u32 flags;
	218	unsigned long age_interval;
	219	unsigned int addrmax;
	220	bool no_share;
	221	};
	222
	223	/* Pseudo network device */
	224	struct vxlan_dev {
	225	struct hlist_node hlist; /* vni hash table */
	226	struct list_head next; /* vxlan's per namespace list */
b1be00a6 JB	227	struct vxlan_sock vn4_sock; / listening socket for IPv4 */
	228	#if IS_ENABLED(CONFIG_IPV6)
	229	struct vxlan_sock vn6_sock; / listening socket for IPv6 */
	230	#endif
0dfbdf41 TG	231	struct net_device *dev;
	232	struct net net; / netns for packet i/o */
	233	struct vxlan_rdst default_dst; /* default destination */
	234	u32 flags; /* VXLAN_F_* in vxlan.h */
	235
	236	struct timer_list age_timer;
	237	spinlock_t hash_lock;
	238	unsigned int addrcnt;
58ce31cc	239	struct gro_cells gro_cells;
0dfbdf41 TG	240
	241	struct vxlan_config cfg;
	242
	243	struct hlist_head fdb_head[FDB_HASH_SIZE];
	244	};
	245
359a0ea9 TH	246	#define VXLAN_F_LEARN 0x01
	247	#define VXLAN_F_PROXY 0x02
	248	#define VXLAN_F_RSC 0x04
	249	#define VXLAN_F_L2MISS 0x08
	250	#define VXLAN_F_L3MISS 0x10
	251	#define VXLAN_F_IPV6 0x20
6ceb31ca	252	#define VXLAN_F_UDP_ZERO_CSUM_TX 0x40
359a0ea9 TH	253	#define VXLAN_F_UDP_ZERO_CSUM6_TX 0x80
359a0ea9 TH	254	#define VXLAN_F_UDP_ZERO_CSUM6_RX 0x100
dfd8645e TH	255	#define VXLAN_F_REMCSUM_TX 0x200
dfd8645e TH	256	#define VXLAN_F_REMCSUM_RX 0x400
3511494c	257	#define VXLAN_F_GBP 0x800
0ace2ca8	258	#define VXLAN_F_REMCSUM_NOPARTIAL 0x1000
ee122c79	259	#define VXLAN_F_COLLECT_METADATA 0x2000
e1e5314d	260	#define VXLAN_F_GPE 0x4000
359a0ea9	261
d299ce14	262	/* Flags that are used in the receive path. These flags must match in
af33c1ad TH	263	* order for a socket to be shareable
	264	*/
	265	#define VXLAN_F_RCV_FLAGS (VXLAN_F_GBP \| \
e1e5314d	266	VXLAN_F_GPE \| \
af33c1ad	267	VXLAN_F_UDP_ZERO_CSUM6_RX \| \
0ace2ca8	268	VXLAN_F_REMCSUM_RX \| \
ee122c79	269	VXLAN_F_REMCSUM_NOPARTIAL \| \
da8b43c0	270	VXLAN_F_COLLECT_METADATA)
ac5132d1	271
e1e5314d JB	272	/* Flags that can be set together with VXLAN_F_GPE. */
	273	#define VXLAN_F_ALLOWED_GPE (VXLAN_F_GPE \| \
	274	VXLAN_F_IPV6 \| \
	275	VXLAN_F_UDP_ZERO_CSUM_TX \| \
	276	VXLAN_F_UDP_ZERO_CSUM6_TX \| \
	277	VXLAN_F_UDP_ZERO_CSUM6_RX \| \
	278	VXLAN_F_COLLECT_METADATA)
	279
0dfbdf41 TG	280	struct net_device vxlan_dev_create(struct net net, const char *name,
	281	u8 name_assign_type, struct vxlan_config *conf);
	282
b1be00a6 JB	283	static inline __be16 vxlan_dev_dst_port(struct vxlan_dev *vxlan,
b1be00a6 JB	284	unsigned short family)
614732ea	285	{
b1be00a6 JB	286	#if IS_ENABLED(CONFIG_IPV6)
	287	if (family == AF_INET6)
	288	return inet_sk(vxlan->vn6_sock->sock->sk)->inet_sport;
	289	#endif
	290	return inet_sk(vxlan->vn4_sock->sock->sk)->inet_sport;
614732ea	291	}
49560532	292
5f35227e JG	293	static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
5f35227e JG	294	netdev_features_t features)
11bf7828	295	{
5f35227e JG	296	u8 l4_hdr = 0;
	297
	298	if (!skb->encapsulation)
	299	return features;
	300
	301	switch (vlan_get_protocol(skb)) {
	302	case htons(ETH_P_IP):
	303	l4_hdr = ip_hdr(skb)->protocol;
	304	break;
	305	case htons(ETH_P_IPV6):
	306	l4_hdr = ipv6_hdr(skb)->nexthdr;
	307	break;
	308	default:
	309	return features;;
	310	}
	311
	312	if ((l4_hdr == IPPROTO_UDP) &&
11bf7828 JS	313	(skb->inner_protocol_type != ENCAP_TYPE_ETHER \|\|
	314	skb->inner_protocol != htons(ETH_P_TEB) \|\|
	315	(skb_inner_mac_header(skb) - skb_transport_header(skb) !=
af67eb9e AD	316	sizeof(struct udphdr) + sizeof(struct vxlanhdr)) \|\|
	317	(skb->ip_summed != CHECKSUM_NONE &&
	318	!can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
a188222b	319	return features & ~(NETIF_F_CSUM_MASK \| NETIF_F_GSO_MASK);
11bf7828	320
5f35227e	321	return features;
11bf7828	322	}
23e62de3	323
e6cd988c JG	324	/* IP header + UDP + VXLAN + Ethernet header */
	325	#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
	326	/* IPv6 header + UDP + VXLAN + Ethernet header */
	327	#define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
	328
d4ac05ff JB	329	static inline struct vxlanhdr vxlan_hdr(struct sk_buff skb)
	330	{
	331	return (struct vxlanhdr *)(udp_hdr(skb) + 1);
	332	}
	333
54bfd872 JB	334	static inline __be32 vxlan_vni(__be32 vni_field)
	335	{
	336	#if defined(__BIG_ENDIAN)
5692d7ea	337	return (__force __be32)((__force u32)vni_field >> 8);
54bfd872	338	#else
5692d7ea	339	return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
54bfd872 JB	340	#endif
	341	}
	342
	343	static inline __be32 vxlan_vni_field(__be32 vni)
	344	{
	345	#if defined(__BIG_ENDIAN)
5692d7ea	346	return (__force __be32)((__force u32)vni << 8);
54bfd872	347	#else
5692d7ea	348	return (__force __be32)((__force u32)vni >> 8);
54bfd872 JB	349	#endif
	350	}
	351
	352	static inline __be32 vxlan_tun_id_to_vni(__be64 tun_id)
	353	{
	354	#if defined(__BIG_ENDIAN)
5692d7ea	355	return (__force __be32)tun_id;
54bfd872	356	#else
5692d7ea	357	return (__force __be32)((__force u64)tun_id >> 32);
54bfd872 JB	358	#endif
	359	}
	360
07dabf20 JB	361	static inline __be64 vxlan_vni_to_tun_id(__be32 vni)
	362	{
	363	#if defined(__BIG_ENDIAN)
5692d7ea	364	return (__force __be64)vni;
07dabf20	365	#else
5692d7ea	366	return (__force __be64)((u64)(__force u32)vni << 32);
07dabf20 JB	367	#endif
	368	}
	369
54bfd872 JB	370	static inline size_t vxlan_rco_start(__be32 vni_field)
	371	{
	372	return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
	373	}
	374
	375	static inline size_t vxlan_rco_offset(__be32 vni_field)
	376	{
	377	return (vni_field & VXLAN_RCO_UDP) ?
	378	offsetof(struct udphdr, check) :
	379	offsetof(struct tcphdr, check);
	380	}
	381
	382	static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
	383	{
	384	__be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
	385
	386	if (offset == offsetof(struct udphdr, check))
	387	vni_field \|= VXLAN_RCO_UDP;
	388	return vni_field;
	389	}
	390
e6cd988c JG	391	static inline void vxlan_get_rx_port(struct net_device *netdev)
e6cd988c JG	392	{
b7aade15 HFS	393	ASSERT_RTNL();
b7aade15 HFS	394	call_netdevice_notifiers(NETDEV_OFFLOAD_PUSH_VXLAN, netdev);
e6cd988c	395	}
705cc62f JB	396
	397	static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
	398	{
	399	return vs->sock->sk->sk_family;
	400	}
48e92c44 JB	401
48e92c44 JB	402	#endif