[linux-2.6-block.git] / include / net / vxlan.h

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

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

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];
	refcount_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;
	u8			 offloaded:1;
	__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;
	enum ifla_vxlan_df	df;
};

struct vxlan_dev_node {
	struct hlist_node hlist;
	struct vxlan_dev *vxlan;
};

/* Pseudo network device */
struct vxlan_dev {
	struct vxlan_dev_node hlist4;	/* vni hash table for IPv4 socket */
#if IS_ENABLED(CONFIG_IPV6)
	struct vxlan_dev_node hlist6;	/* vni hash table for IPv6 socket */
#endif
	struct list_head  next;		/* vxlan's per namespace list */
	struct vxlan_sock __rcu *vn4_sock;	/* listening socket for IPv4 */
#if IS_ENABLED(CONFIG_IPV6)
	struct vxlan_sock __rcu *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 */

	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
#define VXLAN_F_IPV6_LINKLOCAL		0x8000
#define VXLAN_F_TTL_INHERIT		0x10000

/* 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_IPV6_LINKLOCAL |	\
					 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 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 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 unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
{
	return vs->sock->sk->sk_family;
}

#if IS_ENABLED(CONFIG_IPV6)

static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
{
	if (ipa->sa.sa_family == AF_INET6)
		return ipv6_addr_any(&ipa->sin6.sin6_addr);
	else
		return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
}

static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
{
	if (ipa->sa.sa_family == AF_INET6)
		return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
	else
		return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
}

#else /* !IS_ENABLED(CONFIG_IPV6) */

static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
{
	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
}

static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
{
	return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
}

#endif /* IS_ENABLED(CONFIG_IPV6) */

static inline bool netif_is_vxlan(const struct net_device *dev)
{
	return dev->rtnl_link_ops &&
	       !strcmp(dev->rtnl_link_ops->kind, "vxlan");
}

struct switchdev_notifier_vxlan_fdb_info {
	struct switchdev_notifier_info info; /* must be first */
	union vxlan_addr remote_ip;
	__be16 remote_port;
	__be32 remote_vni;
	u32 remote_ifindex;
	u8 eth_addr[ETH_ALEN];
	__be32 vni;
	bool offloaded;
	bool added_by_user;
};

#if IS_ENABLED(CONFIG_VXLAN)
int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
		      struct switchdev_notifier_vxlan_fdb_info *fdb_info);
int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
		     struct notifier_block *nb);
void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);

#else
static inline int
vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
		  struct switchdev_notifier_vxlan_fdb_info *fdb_info)
{
	return -ENOENT;
}

static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
				   struct notifier_block *nb)
{
	return -EOPNOTSUPP;
}

static inline void
vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
{
}
#endif

#endif
Commit	Line	Data
b2441318	1	/* SPDX-License-Identifier: GPL-2.0 */
012a5729 PS	2	#ifndef __NET_VXLAN_H
	3	#define __NET_VXLAN_H 1
	4
5f35227e	5	#include <linux/if_vlan.h>
86a98057	6	#include <net/udp_tunnel.h>
ee122c79	7	#include <net/dst_metadata.h>
5ff4ff4f	8	#include <net/rtnetlink.h>
9a997353	9	#include <net/switchdev.h>
012a5729	10
828788ac JB	11	/* VXLAN protocol (RFC 7348) header:
	12	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	13	* \|R\|R\|R\|R\|I\|R\|R\|R\| Reserved \|
	14	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	15	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	16	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	17	*
	18	* I = VXLAN Network Identifier (VNI) present.
	19	*/
	20	struct vxlanhdr {
	21	__be32 vx_flags;
	22	__be32 vx_vni;
	23	};
	24
	25	/* VXLAN header flags. */
54bfd872	26	#define VXLAN_HF_VNI cpu_to_be32(BIT(27))
828788ac JB	27
	28	#define VXLAN_N_VID (1u << 24)
	29	#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
54bfd872	30	#define VXLAN_VNI_MASK cpu_to_be32(VXLAN_VID_MASK << 8)
828788ac JB	31	#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
	32
	33	#define VNI_HASH_BITS 10
	34	#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
	35	#define FDB_HASH_BITS 8
	36	#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
	37
	38	/* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
	39	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	40	* \|R\|R\|R\|R\|I\|R\|R\|R\|R\|R\|C\| Reserved \|
	41	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	42	* \| VXLAN Network Identifier (VNI) \|O\| Csum start \|
	43	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	44	*
	45	* C = Remote checksum offload bit. When set indicates that the
	46	* remote checksum offload data is present.
	47	*
	48	* O = Offset bit. Indicates the checksum offset relative to
	49	* checksum start.
	50	*
	51	* Csum start = Checksum start divided by two.
	52	*
	53	* http://tools.ietf.org/html/draft-herbert-vxlan-rco
	54	*/
	55
	56	/* VXLAN-RCO header flags. */
54bfd872	57	#define VXLAN_HF_RCO cpu_to_be32(BIT(21))
828788ac JB	58
828788ac JB	59	/* Remote checksum offload header option */
54bfd872 JB	60	#define VXLAN_RCO_MASK cpu_to_be32(0x7f) /* Last byte of vni field */
	61	#define VXLAN_RCO_UDP cpu_to_be32(0x80) /* Indicate UDP RCO (TCP when not set ) /
	62	#define VXLAN_RCO_SHIFT 1 /* Left shift of start */
828788ac	63	#define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
54bfd872	64	#define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)
828788ac	65
3511494c	66	/*
828788ac	67	* VXLAN Group Based Policy Extension (VXLAN_F_GBP):
3511494c	68	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
828788ac	69	* \|G\|R\|R\|R\|I\|R\|R\|R\|R\|D\|R\|R\|A\|R\|R\|R\| Group Policy ID \|
3511494c TG	70	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	71	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	72	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	73	*
828788ac JB	74	* G = Group Policy ID present.
828788ac JB	75	*
3511494c TG	76	* D = Don't Learn bit. When set, this bit indicates that the egress
	77	* VTEP MUST NOT learn the source address of the encapsulated frame.
	78	*
	79	* A = Indicates that the group policy has already been applied to
	80	* this packet. Policies MUST NOT be applied by devices when the
	81	* A bit is set.
	82	*
828788ac	83	* https://tools.ietf.org/html/draft-smith-vxlan-group-policy
3511494c TG	84	*/
3511494c TG	85	struct vxlanhdr_gbp {
0e715d6f	86	u8 vx_flags;
3511494c	87	#ifdef __LITTLE_ENDIAN_BITFIELD
0e715d6f	88	u8 reserved_flags1:3,
3511494c TG	89	policy_applied:1,
	90	reserved_flags2:2,
	91	dont_learn:1,
	92	reserved_flags3:1;
	93	#elif defined(__BIG_ENDIAN_BITFIELD)
0e715d6f	94	u8 reserved_flags1:1,
3511494c TG	95	dont_learn:1,
	96	reserved_flags2:2,
	97	policy_applied:1,
	98	reserved_flags3:3;
	99	#else
	100	#error "Please fix <asm/byteorder.h>"
	101	#endif
	102	__be16 policy_id;
	103	__be32 vx_vni;
	104	};
	105
828788ac	106	/* VXLAN-GBP header flags. */
54bfd872	107	#define VXLAN_HF_GBP cpu_to_be32(BIT(31))
828788ac	108
54bfd872	109	#define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP \| cpu_to_be32(0xFFFFFF))
3511494c TG	110
	111	/* skb->mark mapping
	112	*
	113	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	114	* \|R\|R\|R\|R\|R\|R\|R\|R\|R\|D\|R\|R\|A\|R\|R\|R\| Group Policy ID \|
	115	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	116	*/
	117	#define VXLAN_GBP_DONT_LEARN (BIT(6) << 16)
	118	#define VXLAN_GBP_POLICY_APPLIED (BIT(3) << 16)
	119	#define VXLAN_GBP_ID_MASK (0xFFFF)
	120
e1e5314d JB	121	/*
	122	* VXLAN Generic Protocol Extension (VXLAN_F_GPE):
	123	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	124	* \|R\|R\|Ver\|I\|P\|R\|O\| Reserved \|Next Protocol \|
	125	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	126	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	127	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	128	*
	129	* Ver = Version. Indicates VXLAN GPE protocol version.
	130	*
	131	* P = Next Protocol Bit. The P bit is set to indicate that the
	132	* Next Protocol field is present.
	133	*
	134	* O = OAM Flag Bit. The O bit is set to indicate that the packet
	135	* is an OAM packet.
	136	*
	137	* Next Protocol = This 8 bit field indicates the protocol header
	138	* immediately following the VXLAN GPE header.
	139	*
	140	* https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
	141	*/
	142
	143	struct vxlanhdr_gpe {
	144	#if defined(__LITTLE_ENDIAN_BITFIELD)
	145	u8 oam_flag:1,
	146	reserved_flags1:1,
	147	np_applied:1,
	148	instance_applied:1,
	149	version:2,
f1c8d372	150	reserved_flags2:2;
e1e5314d JB	151	#elif defined(__BIG_ENDIAN_BITFIELD)
	152	u8 reserved_flags2:2,
	153	version:2,
	154	instance_applied:1,
	155	np_applied:1,
	156	reserved_flags1:1,
	157	oam_flag:1;
	158	#endif
	159	u8 reserved_flags3;
	160	u8 reserved_flags4;
	161	u8 next_protocol;
	162	__be32 vx_vni;
	163	};
	164
	165	/* VXLAN-GPE header flags. */
	166	#define VXLAN_HF_VER cpu_to_be32(BIT(29) \| BIT(28))
	167	#define VXLAN_HF_NP cpu_to_be32(BIT(26))
	168	#define VXLAN_HF_OAM cpu_to_be32(BIT(24))
	169
	170	#define VXLAN_GPE_USED_BITS (VXLAN_HF_VER \| VXLAN_HF_NP \| VXLAN_HF_OAM \| \
	171	cpu_to_be32(0xff))
	172
3511494c	173	struct vxlan_metadata {
3511494c TG	174	u32 gbp;
	175	};
	176
012a5729 PS	177	/* per UDP socket information */
	178	struct vxlan_sock {
	179	struct hlist_node hlist;
012a5729	180	struct socket *sock;
012a5729	181	struct hlist_head vni_list[VNI_HASH_SIZE];
66af846f	182	refcount_t refcnt;
dfd8645e	183	u32 flags;
012a5729 PS	184	};
012a5729 PS	185
0dfbdf41 TG	186	union vxlan_addr {
	187	struct sockaddr_in sin;
	188	struct sockaddr_in6 sin6;
	189	struct sockaddr sa;
	190	};
	191
	192	struct vxlan_rdst {
	193	union vxlan_addr remote_ip;
	194	__be16 remote_port;
0efe1173	195	u8 offloaded:1;
54bfd872	196	__be32 remote_vni;
0dfbdf41 TG	197	u32 remote_ifindex;
	198	struct list_head list;
	199	struct rcu_head rcu;
0c1d70af	200	struct dst_cache dst_cache;
0dfbdf41 TG	201	};
	202
	203	struct vxlan_config {
	204	union vxlan_addr remote_ip;
	205	union vxlan_addr saddr;
54bfd872	206	__be32 vni;
0dfbdf41 TG	207	int remote_ifindex;
	208	int mtu;
	209	__be16 dst_port;
0e715d6f JB	210	u16 port_min;
	211	u16 port_max;
	212	u8 tos;
	213	u8 ttl;
e7f70af1	214	__be32 label;
0dfbdf41 TG	215	u32 flags;
	216	unsigned long age_interval;
	217	unsigned int addrmax;
	218	bool no_share;
b4d30697	219	enum ifla_vxlan_df df;
0dfbdf41 TG	220	};
0dfbdf41 TG	221
69e76661 JB	222	struct vxlan_dev_node {
	223	struct hlist_node hlist;
	224	struct vxlan_dev *vxlan;
	225	};
	226
0dfbdf41 TG	227	/* Pseudo network device */
0dfbdf41 TG	228	struct vxlan_dev {
69e76661 JB	229	struct vxlan_dev_node hlist4; /* vni hash table for IPv4 socket */
	230	#if IS_ENABLED(CONFIG_IPV6)
	231	struct vxlan_dev_node hlist6; /* vni hash table for IPv6 socket */
	232	#endif
0dfbdf41	233	struct list_head next; /* vxlan's per namespace list */
c6fcc4fc	234	struct vxlan_sock __rcu vn4_sock; / listening socket for IPv4 */
b1be00a6	235	#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc	236	struct vxlan_sock __rcu vn6_sock; / listening socket for IPv6 */
b1be00a6	237	#endif
0dfbdf41 TG	238	struct net_device *dev;
	239	struct net net; / netns for packet i/o */
	240	struct vxlan_rdst default_dst; /* default destination */
0dfbdf41 TG	241
	242	struct timer_list age_timer;
	243	spinlock_t hash_lock;
	244	unsigned int addrcnt;
58ce31cc	245	struct gro_cells gro_cells;
0dfbdf41 TG	246
	247	struct vxlan_config cfg;
	248
	249	struct hlist_head fdb_head[FDB_HASH_SIZE];
	250	};
	251
359a0ea9 TH	252	#define VXLAN_F_LEARN 0x01
	253	#define VXLAN_F_PROXY 0x02
	254	#define VXLAN_F_RSC 0x04
	255	#define VXLAN_F_L2MISS 0x08
	256	#define VXLAN_F_L3MISS 0x10
	257	#define VXLAN_F_IPV6 0x20
6ceb31ca	258	#define VXLAN_F_UDP_ZERO_CSUM_TX 0x40
359a0ea9 TH	259	#define VXLAN_F_UDP_ZERO_CSUM6_TX 0x80
359a0ea9 TH	260	#define VXLAN_F_UDP_ZERO_CSUM6_RX 0x100
dfd8645e TH	261	#define VXLAN_F_REMCSUM_TX 0x200
dfd8645e TH	262	#define VXLAN_F_REMCSUM_RX 0x400
3511494c	263	#define VXLAN_F_GBP 0x800
0ace2ca8	264	#define VXLAN_F_REMCSUM_NOPARTIAL 0x1000
ee122c79	265	#define VXLAN_F_COLLECT_METADATA 0x2000
e1e5314d	266	#define VXLAN_F_GPE 0x4000
0f22a3c6	267	#define VXLAN_F_IPV6_LINKLOCAL 0x8000
72f6d71e	268	#define VXLAN_F_TTL_INHERIT 0x10000
359a0ea9	269
d299ce14	270	/* Flags that are used in the receive path. These flags must match in
af33c1ad TH	271	* order for a socket to be shareable
	272	*/
	273	#define VXLAN_F_RCV_FLAGS (VXLAN_F_GBP \| \
e1e5314d	274	VXLAN_F_GPE \| \
af33c1ad	275	VXLAN_F_UDP_ZERO_CSUM6_RX \| \
0ace2ca8	276	VXLAN_F_REMCSUM_RX \| \
ee122c79	277	VXLAN_F_REMCSUM_NOPARTIAL \| \
da8b43c0	278	VXLAN_F_COLLECT_METADATA)
ac5132d1	279
e1e5314d JB	280	/* Flags that can be set together with VXLAN_F_GPE. */
	281	#define VXLAN_F_ALLOWED_GPE (VXLAN_F_GPE \| \
	282	VXLAN_F_IPV6 \| \
0f22a3c6	283	VXLAN_F_IPV6_LINKLOCAL \| \
e1e5314d JB	284	VXLAN_F_UDP_ZERO_CSUM_TX \| \
	285	VXLAN_F_UDP_ZERO_CSUM6_TX \| \
	286	VXLAN_F_UDP_ZERO_CSUM6_RX \| \
	287	VXLAN_F_COLLECT_METADATA)
	288
0dfbdf41 TG	289	struct net_device vxlan_dev_create(struct net net, const char *name,
	290	u8 name_assign_type, struct vxlan_config *conf);
	291
5f35227e JG	292	static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
5f35227e JG	293	netdev_features_t features)
11bf7828	294	{
5f35227e JG	295	u8 l4_hdr = 0;
	296
	297	if (!skb->encapsulation)
	298	return features;
	299
	300	switch (vlan_get_protocol(skb)) {
	301	case htons(ETH_P_IP):
	302	l4_hdr = ip_hdr(skb)->protocol;
	303	break;
	304	case htons(ETH_P_IPV6):
	305	l4_hdr = ipv6_hdr(skb)->nexthdr;
	306	break;
	307	default:
5ef7e0ba	308	return features;
5f35227e JG	309	}
	310
	311	if ((l4_hdr == IPPROTO_UDP) &&
11bf7828 JS	312	(skb->inner_protocol_type != ENCAP_TYPE_ETHER \|\|
	313	skb->inner_protocol != htons(ETH_P_TEB) \|\|
	314	(skb_inner_mac_header(skb) - skb_transport_header(skb) !=
af67eb9e AD	315	sizeof(struct udphdr) + sizeof(struct vxlanhdr)) \|\|
	316	(skb->ip_summed != CHECKSUM_NONE &&
	317	!can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
a188222b	318	return features & ~(NETIF_F_CSUM_MASK \| NETIF_F_GSO_MASK);
11bf7828	319
5f35227e	320	return features;
11bf7828	321	}
23e62de3	322
e6cd988c JG	323	/* IP header + UDP + VXLAN + Ethernet header */
	324	#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
	325	/* IPv6 header + UDP + VXLAN + Ethernet header */
	326	#define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
	327
d4ac05ff JB	328	static inline struct vxlanhdr vxlan_hdr(struct sk_buff skb)
	329	{
	330	return (struct vxlanhdr *)(udp_hdr(skb) + 1);
	331	}
	332
54bfd872 JB	333	static inline __be32 vxlan_vni(__be32 vni_field)
	334	{
	335	#if defined(__BIG_ENDIAN)
5692d7ea	336	return (__force __be32)((__force u32)vni_field >> 8);
54bfd872	337	#else
5692d7ea	338	return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
54bfd872 JB	339	#endif
	340	}
	341
	342	static inline __be32 vxlan_vni_field(__be32 vni)
	343	{
	344	#if defined(__BIG_ENDIAN)
5692d7ea	345	return (__force __be32)((__force u32)vni << 8);
54bfd872	346	#else
5692d7ea	347	return (__force __be32)((__force u32)vni >> 8);
54bfd872 JB	348	#endif
	349	}
	350
54bfd872 JB	351	static inline size_t vxlan_rco_start(__be32 vni_field)
	352	{
	353	return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
	354	}
	355
	356	static inline size_t vxlan_rco_offset(__be32 vni_field)
	357	{
	358	return (vni_field & VXLAN_RCO_UDP) ?
	359	offsetof(struct udphdr, check) :
	360	offsetof(struct tcphdr, check);
	361	}
	362
	363	static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
	364	{
	365	__be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
	366
	367	if (offset == offsetof(struct udphdr, check))
	368	vni_field \|= VXLAN_RCO_UDP;
	369	return vni_field;
	370	}
	371
705cc62f JB	372	static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
	373	{
	374	return vs->sock->sk->sk_family;
	375	}
48e92c44	376
cca45e05 IS	377	#if IS_ENABLED(CONFIG_IPV6)
	378
	379	static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
	380	{
	381	if (ipa->sa.sa_family == AF_INET6)
	382	return ipv6_addr_any(&ipa->sin6.sin6_addr);
	383	else
	384	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
	385	}
	386
	387	static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
	388	{
	389	if (ipa->sa.sa_family == AF_INET6)
	390	return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
	391	else
	392	return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
	393	}
	394
	395	#else /* !IS_ENABLED(CONFIG_IPV6) */
	396
	397	static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
	398	{
	399	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
	400	}
	401
	402	static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
	403	{
	404	return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
	405	}
	406
	407	#endif /* IS_ENABLED(CONFIG_IPV6) */
	408
5ff4ff4f IS	409	static inline bool netif_is_vxlan(const struct net_device *dev)
	410	{
	411	return dev->rtnl_link_ops &&
	412	!strcmp(dev->rtnl_link_ops->kind, "vxlan");
	413	}
	414
9a997353 PM	415	struct switchdev_notifier_vxlan_fdb_info {
	416	struct switchdev_notifier_info info; /* must be first */
	417	union vxlan_addr remote_ip;
	418	__be16 remote_port;
	419	__be32 remote_vni;
	420	u32 remote_ifindex;
	421	u8 eth_addr[ETH_ALEN];
	422	__be32 vni;
0efe1173	423	bool offloaded;
45598c1c	424	bool added_by_user;
9a997353 PM	425	};
9a997353 PM	426
1941f1d6 PM	427	#if IS_ENABLED(CONFIG_VXLAN)
	428	int vxlan_fdb_find_uc(struct net_device dev, const u8 mac, __be32 vni,
	429	struct switchdev_notifier_vxlan_fdb_info *fdb_info);
4f89f5b5 PM	430	int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
4f89f5b5 PM	431	struct notifier_block *nb);
e5ff4b19	432	void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);
4f89f5b5	433
1941f1d6 PM	434	#else
	435	static inline int
	436	vxlan_fdb_find_uc(struct net_device dev, const u8 mac, __be32 vni,
	437	struct switchdev_notifier_vxlan_fdb_info *fdb_info)
	438	{
	439	return -ENOENT;
	440	}
4f89f5b5 PM	441
	442	static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
	443	struct notifier_block *nb)
	444	{
	445	return -EOPNOTSUPP;
	446	}
e5ff4b19 PM	447
	448	static inline void
	449	vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
	450	{
	451	}
1941f1d6 PM	452	#endif
1941f1d6 PM	453
48e92c44	454	#endif