[mirror_ubuntu-jammy-kernel.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>
#include <net/nexthop.h>

#define IANA_VXLAN_UDP_PORT     4789

/* 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)

#define VXLAN_GBP_MASK (VXLAN_GBP_DONT_LEARN | VXLAN_GBP_POLICY_APPLIED | \
			VXLAN_GBP_ID_MASK)

/*
 * 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 net_device	 *remote_dev;
	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[FDB_HASH_SIZE];
	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 ipv4_is_multicast(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 ipv4_is_multicast(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,
		     struct netlink_ext_ack *extack);
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,
				   struct netlink_ext_ack *extack)
{
	return -EOPNOTSUPP;
}

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

static inline void vxlan_flag_attr_error(int attrtype,
					 struct netlink_ext_ack *extack)
{
#define VXLAN_FLAG(flg) \
	case IFLA_VXLAN_##flg: \
		NL_SET_ERR_MSG_MOD(extack, \
				   "cannot change " #flg " flag"); \
		break
	switch (attrtype) {
	VXLAN_FLAG(TTL_INHERIT);
	VXLAN_FLAG(LEARNING);
	VXLAN_FLAG(PROXY);
	VXLAN_FLAG(RSC);
	VXLAN_FLAG(L2MISS);
	VXLAN_FLAG(L3MISS);
	VXLAN_FLAG(COLLECT_METADATA);
	VXLAN_FLAG(UDP_ZERO_CSUM6_TX);
	VXLAN_FLAG(UDP_ZERO_CSUM6_RX);
	VXLAN_FLAG(REMCSUM_TX);
	VXLAN_FLAG(REMCSUM_RX);
	VXLAN_FLAG(GBP);
	VXLAN_FLAG(GPE);
	VXLAN_FLAG(REMCSUM_NOPARTIAL);
	default:
		NL_SET_ERR_MSG_MOD(extack, \
				   "cannot change flag");
		break;
	}
#undef VXLAN_FLAG
}

static inline bool vxlan_fdb_nh_path_select(struct nexthop *nh,
					    int hash,
					    struct vxlan_rdst *rdst)
{
	struct fib_nh_common *nhc;

	nhc = nexthop_path_fdb_result(nh, hash);
	if (unlikely(!nhc))
		return false;

	switch (nhc->nhc_gw_family) {
	case AF_INET:
		rdst->remote_ip.sin.sin_addr.s_addr = nhc->nhc_gw.ipv4;
		rdst->remote_ip.sa.sa_family = AF_INET;
		break;
	case AF_INET6:
		rdst->remote_ip.sin6.sin6_addr = nhc->nhc_gw.ipv6;
		rdst->remote_ip.sa.sa_family = AF_INET6;
		break;
	}

	return true;
}

#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>
1274e1cc	10	#include <net/nexthop.h>
012a5729	11
bea96410 MS	12	#define IANA_VXLAN_UDP_PORT 4789
bea96410 MS	13
828788ac JB	14	/* VXLAN protocol (RFC 7348) header:
	15	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	16	* \|R\|R\|R\|R\|I\|R\|R\|R\| Reserved \|
	17	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	18	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	19	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	20	*
	21	* I = VXLAN Network Identifier (VNI) present.
	22	*/
	23	struct vxlanhdr {
	24	__be32 vx_flags;
	25	__be32 vx_vni;
	26	};
	27
	28	/* VXLAN header flags. */
54bfd872	29	#define VXLAN_HF_VNI cpu_to_be32(BIT(27))
828788ac JB	30
	31	#define VXLAN_N_VID (1u << 24)
	32	#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
54bfd872	33	#define VXLAN_VNI_MASK cpu_to_be32(VXLAN_VID_MASK << 8)
828788ac JB	34	#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
	35
	36	#define VNI_HASH_BITS 10
	37	#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
	38	#define FDB_HASH_BITS 8
	39	#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
	40
	41	/* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
	42	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	43	* \|R\|R\|R\|R\|I\|R\|R\|R\|R\|R\|C\| Reserved \|
	44	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	45	* \| VXLAN Network Identifier (VNI) \|O\| Csum start \|
	46	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	47	*
	48	* C = Remote checksum offload bit. When set indicates that the
	49	* remote checksum offload data is present.
	50	*
	51	* O = Offset bit. Indicates the checksum offset relative to
	52	* checksum start.
	53	*
	54	* Csum start = Checksum start divided by two.
	55	*
	56	* http://tools.ietf.org/html/draft-herbert-vxlan-rco
	57	*/
	58
	59	/* VXLAN-RCO header flags. */
54bfd872	60	#define VXLAN_HF_RCO cpu_to_be32(BIT(21))
828788ac JB	61
828788ac JB	62	/* Remote checksum offload header option */
54bfd872 JB	63	#define VXLAN_RCO_MASK cpu_to_be32(0x7f) /* Last byte of vni field */
	64	#define VXLAN_RCO_UDP cpu_to_be32(0x80) /* Indicate UDP RCO (TCP when not set ) /
	65	#define VXLAN_RCO_SHIFT 1 /* Left shift of start */
828788ac	66	#define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
54bfd872	67	#define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)
828788ac	68
3511494c	69	/*
828788ac	70	* VXLAN Group Based Policy Extension (VXLAN_F_GBP):
3511494c	71	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
828788ac	72	* \|G\|R\|R\|R\|I\|R\|R\|R\|R\|D\|R\|R\|A\|R\|R\|R\| Group Policy ID \|
3511494c TG	73	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	74	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	75	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	76	*
828788ac JB	77	* G = Group Policy ID present.
828788ac JB	78	*
3511494c TG	79	* D = Don't Learn bit. When set, this bit indicates that the egress
	80	* VTEP MUST NOT learn the source address of the encapsulated frame.
	81	*
	82	* A = Indicates that the group policy has already been applied to
	83	* this packet. Policies MUST NOT be applied by devices when the
	84	* A bit is set.
	85	*
828788ac	86	* https://tools.ietf.org/html/draft-smith-vxlan-group-policy
3511494c TG	87	*/
3511494c TG	88	struct vxlanhdr_gbp {
0e715d6f	89	u8 vx_flags;
3511494c	90	#ifdef __LITTLE_ENDIAN_BITFIELD
0e715d6f	91	u8 reserved_flags1:3,
3511494c TG	92	policy_applied:1,
	93	reserved_flags2:2,
	94	dont_learn:1,
	95	reserved_flags3:1;
	96	#elif defined(__BIG_ENDIAN_BITFIELD)
0e715d6f	97	u8 reserved_flags1:1,
3511494c TG	98	dont_learn:1,
	99	reserved_flags2:2,
	100	policy_applied:1,
	101	reserved_flags3:3;
	102	#else
	103	#error "Please fix <asm/byteorder.h>"
	104	#endif
	105	__be16 policy_id;
	106	__be32 vx_vni;
	107	};
	108
828788ac	109	/* VXLAN-GBP header flags. */
54bfd872	110	#define VXLAN_HF_GBP cpu_to_be32(BIT(31))
828788ac	111
54bfd872	112	#define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP \| cpu_to_be32(0xFFFFFF))
3511494c TG	113
	114	/* skb->mark mapping
	115	*
	116	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	117	* \|R\|R\|R\|R\|R\|R\|R\|R\|R\|D\|R\|R\|A\|R\|R\|R\| Group Policy ID \|
	118	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	119	*/
	120	#define VXLAN_GBP_DONT_LEARN (BIT(6) << 16)
	121	#define VXLAN_GBP_POLICY_APPLIED (BIT(3) << 16)
	122	#define VXLAN_GBP_ID_MASK (0xFFFF)
	123
13e6ce98 XL	124	#define VXLAN_GBP_MASK (VXLAN_GBP_DONT_LEARN \| VXLAN_GBP_POLICY_APPLIED \| \
	125	VXLAN_GBP_ID_MASK)
	126
e1e5314d JB	127	/*
	128	* VXLAN Generic Protocol Extension (VXLAN_F_GPE):
	129	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	130	* \|R\|R\|Ver\|I\|P\|R\|O\| Reserved \|Next Protocol \|
	131	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	132	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	133	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	134	*
	135	* Ver = Version. Indicates VXLAN GPE protocol version.
	136	*
	137	* P = Next Protocol Bit. The P bit is set to indicate that the
	138	* Next Protocol field is present.
	139	*
	140	* O = OAM Flag Bit. The O bit is set to indicate that the packet
	141	* is an OAM packet.
	142	*
	143	* Next Protocol = This 8 bit field indicates the protocol header
	144	* immediately following the VXLAN GPE header.
	145	*
	146	* https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
	147	*/
	148
	149	struct vxlanhdr_gpe {
	150	#if defined(__LITTLE_ENDIAN_BITFIELD)
	151	u8 oam_flag:1,
	152	reserved_flags1:1,
	153	np_applied:1,
	154	instance_applied:1,
	155	version:2,
f1c8d372	156	reserved_flags2:2;
e1e5314d JB	157	#elif defined(__BIG_ENDIAN_BITFIELD)
	158	u8 reserved_flags2:2,
	159	version:2,
	160	instance_applied:1,
	161	np_applied:1,
	162	reserved_flags1:1,
	163	oam_flag:1;
	164	#endif
	165	u8 reserved_flags3;
	166	u8 reserved_flags4;
	167	u8 next_protocol;
	168	__be32 vx_vni;
	169	};
	170
	171	/* VXLAN-GPE header flags. */
	172	#define VXLAN_HF_VER cpu_to_be32(BIT(29) \| BIT(28))
	173	#define VXLAN_HF_NP cpu_to_be32(BIT(26))
	174	#define VXLAN_HF_OAM cpu_to_be32(BIT(24))
	175
	176	#define VXLAN_GPE_USED_BITS (VXLAN_HF_VER \| VXLAN_HF_NP \| VXLAN_HF_OAM \| \
	177	cpu_to_be32(0xff))
	178
3511494c	179	struct vxlan_metadata {
3511494c TG	180	u32 gbp;
	181	};
	182
012a5729 PS	183	/* per UDP socket information */
	184	struct vxlan_sock {
	185	struct hlist_node hlist;
012a5729	186	struct socket *sock;
012a5729	187	struct hlist_head vni_list[VNI_HASH_SIZE];
66af846f	188	refcount_t refcnt;
dfd8645e	189	u32 flags;
012a5729 PS	190	};
012a5729 PS	191
0dfbdf41 TG	192	union vxlan_addr {
	193	struct sockaddr_in sin;
	194	struct sockaddr_in6 sin6;
	195	struct sockaddr sa;
	196	};
	197
	198	struct vxlan_rdst {
	199	union vxlan_addr remote_ip;
	200	__be16 remote_port;
0efe1173	201	u8 offloaded:1;
54bfd872	202	__be32 remote_vni;
0dfbdf41	203	u32 remote_ifindex;
0ce1822c	204	struct net_device *remote_dev;
0dfbdf41 TG	205	struct list_head list;
0dfbdf41 TG	206	struct rcu_head rcu;
0c1d70af	207	struct dst_cache dst_cache;
0dfbdf41 TG	208	};
	209
	210	struct vxlan_config {
	211	union vxlan_addr remote_ip;
	212	union vxlan_addr saddr;
54bfd872	213	__be32 vni;
0dfbdf41 TG	214	int remote_ifindex;
	215	int mtu;
	216	__be16 dst_port;
0e715d6f JB	217	u16 port_min;
	218	u16 port_max;
	219	u8 tos;
	220	u8 ttl;
e7f70af1	221	__be32 label;
0dfbdf41 TG	222	u32 flags;
	223	unsigned long age_interval;
	224	unsigned int addrmax;
	225	bool no_share;
b4d30697	226	enum ifla_vxlan_df df;
0dfbdf41 TG	227	};
0dfbdf41 TG	228
69e76661 JB	229	struct vxlan_dev_node {
	230	struct hlist_node hlist;
	231	struct vxlan_dev *vxlan;
	232	};
	233
0dfbdf41 TG	234	/* Pseudo network device */
0dfbdf41 TG	235	struct vxlan_dev {
69e76661 JB	236	struct vxlan_dev_node hlist4; /* vni hash table for IPv4 socket */
	237	#if IS_ENABLED(CONFIG_IPV6)
	238	struct vxlan_dev_node hlist6; /* vni hash table for IPv6 socket */
	239	#endif
0dfbdf41	240	struct list_head next; /* vxlan's per namespace list */
c6fcc4fc	241	struct vxlan_sock __rcu vn4_sock; / listening socket for IPv4 */
b1be00a6	242	#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc	243	struct vxlan_sock __rcu vn6_sock; / listening socket for IPv6 */
b1be00a6	244	#endif
0dfbdf41 TG	245	struct net_device *dev;
	246	struct net net; / netns for packet i/o */
	247	struct vxlan_rdst default_dst; /* default destination */
0dfbdf41 TG	248
0dfbdf41 TG	249	struct timer_list age_timer;
fe1e0713	250	spinlock_t hash_lock[FDB_HASH_SIZE];
0dfbdf41	251	unsigned int addrcnt;
58ce31cc	252	struct gro_cells gro_cells;
0dfbdf41 TG	253
	254	struct vxlan_config cfg;
	255
	256	struct hlist_head fdb_head[FDB_HASH_SIZE];
	257	};
	258
359a0ea9 TH	259	#define VXLAN_F_LEARN 0x01
	260	#define VXLAN_F_PROXY 0x02
	261	#define VXLAN_F_RSC 0x04
	262	#define VXLAN_F_L2MISS 0x08
	263	#define VXLAN_F_L3MISS 0x10
	264	#define VXLAN_F_IPV6 0x20
6ceb31ca	265	#define VXLAN_F_UDP_ZERO_CSUM_TX 0x40
359a0ea9 TH	266	#define VXLAN_F_UDP_ZERO_CSUM6_TX 0x80
359a0ea9 TH	267	#define VXLAN_F_UDP_ZERO_CSUM6_RX 0x100
dfd8645e TH	268	#define VXLAN_F_REMCSUM_TX 0x200
dfd8645e TH	269	#define VXLAN_F_REMCSUM_RX 0x400
3511494c	270	#define VXLAN_F_GBP 0x800
0ace2ca8	271	#define VXLAN_F_REMCSUM_NOPARTIAL 0x1000
ee122c79	272	#define VXLAN_F_COLLECT_METADATA 0x2000
e1e5314d	273	#define VXLAN_F_GPE 0x4000
0f22a3c6	274	#define VXLAN_F_IPV6_LINKLOCAL 0x8000
72f6d71e	275	#define VXLAN_F_TTL_INHERIT 0x10000
359a0ea9	276
d299ce14	277	/* Flags that are used in the receive path. These flags must match in
af33c1ad TH	278	* order for a socket to be shareable
	279	*/
	280	#define VXLAN_F_RCV_FLAGS (VXLAN_F_GBP \| \
e1e5314d	281	VXLAN_F_GPE \| \
af33c1ad	282	VXLAN_F_UDP_ZERO_CSUM6_RX \| \
0ace2ca8	283	VXLAN_F_REMCSUM_RX \| \
ee122c79	284	VXLAN_F_REMCSUM_NOPARTIAL \| \
da8b43c0	285	VXLAN_F_COLLECT_METADATA)
ac5132d1	286
e1e5314d JB	287	/* Flags that can be set together with VXLAN_F_GPE. */
	288	#define VXLAN_F_ALLOWED_GPE (VXLAN_F_GPE \| \
	289	VXLAN_F_IPV6 \| \
0f22a3c6	290	VXLAN_F_IPV6_LINKLOCAL \| \
e1e5314d JB	291	VXLAN_F_UDP_ZERO_CSUM_TX \| \
	292	VXLAN_F_UDP_ZERO_CSUM6_TX \| \
	293	VXLAN_F_UDP_ZERO_CSUM6_RX \| \
	294	VXLAN_F_COLLECT_METADATA)
	295
0dfbdf41 TG	296	struct net_device vxlan_dev_create(struct net net, const char *name,
	297	u8 name_assign_type, struct vxlan_config *conf);
	298
5f35227e JG	299	static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
5f35227e JG	300	netdev_features_t features)
11bf7828	301	{
5f35227e JG	302	u8 l4_hdr = 0;
	303
	304	if (!skb->encapsulation)
	305	return features;
	306
	307	switch (vlan_get_protocol(skb)) {
	308	case htons(ETH_P_IP):
	309	l4_hdr = ip_hdr(skb)->protocol;
	310	break;
	311	case htons(ETH_P_IPV6):
	312	l4_hdr = ipv6_hdr(skb)->nexthdr;
	313	break;
	314	default:
5ef7e0ba	315	return features;
5f35227e JG	316	}
	317
	318	if ((l4_hdr == IPPROTO_UDP) &&
11bf7828 JS	319	(skb->inner_protocol_type != ENCAP_TYPE_ETHER \|\|
	320	skb->inner_protocol != htons(ETH_P_TEB) \|\|
	321	(skb_inner_mac_header(skb) - skb_transport_header(skb) !=
af67eb9e AD	322	sizeof(struct udphdr) + sizeof(struct vxlanhdr)) \|\|
	323	(skb->ip_summed != CHECKSUM_NONE &&
	324	!can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
a188222b	325	return features & ~(NETIF_F_CSUM_MASK \| NETIF_F_GSO_MASK);
11bf7828	326
5f35227e	327	return features;
11bf7828	328	}
23e62de3	329
e6cd988c JG	330	/* IP header + UDP + VXLAN + Ethernet header */
	331	#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
	332	/* IPv6 header + UDP + VXLAN + Ethernet header */
	333	#define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
	334
d4ac05ff JB	335	static inline struct vxlanhdr vxlan_hdr(struct sk_buff skb)
	336	{
	337	return (struct vxlanhdr *)(udp_hdr(skb) + 1);
	338	}
	339
54bfd872 JB	340	static inline __be32 vxlan_vni(__be32 vni_field)
	341	{
	342	#if defined(__BIG_ENDIAN)
5692d7ea	343	return (__force __be32)((__force u32)vni_field >> 8);
54bfd872	344	#else
5692d7ea	345	return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
54bfd872 JB	346	#endif
	347	}
	348
	349	static inline __be32 vxlan_vni_field(__be32 vni)
	350	{
	351	#if defined(__BIG_ENDIAN)
5692d7ea	352	return (__force __be32)((__force u32)vni << 8);
54bfd872	353	#else
5692d7ea	354	return (__force __be32)((__force u32)vni >> 8);
54bfd872 JB	355	#endif
	356	}
	357
54bfd872 JB	358	static inline size_t vxlan_rco_start(__be32 vni_field)
	359	{
	360	return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
	361	}
	362
	363	static inline size_t vxlan_rco_offset(__be32 vni_field)
	364	{
	365	return (vni_field & VXLAN_RCO_UDP) ?
	366	offsetof(struct udphdr, check) :
	367	offsetof(struct tcphdr, check);
	368	}
	369
	370	static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
	371	{
	372	__be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
	373
	374	if (offset == offsetof(struct udphdr, check))
	375	vni_field \|= VXLAN_RCO_UDP;
	376	return vni_field;
	377	}
	378
705cc62f JB	379	static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
	380	{
	381	return vs->sock->sk->sk_family;
	382	}
48e92c44	383
cca45e05 IS	384	#if IS_ENABLED(CONFIG_IPV6)
	385
	386	static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
	387	{
	388	if (ipa->sa.sa_family == AF_INET6)
	389	return ipv6_addr_any(&ipa->sin6.sin6_addr);
	390	else
	391	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
	392	}
	393
	394	static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
	395	{
	396	if (ipa->sa.sa_family == AF_INET6)
	397	return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
	398	else
842841ec	399	return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
cca45e05 IS	400	}
	401
	402	#else /* !IS_ENABLED(CONFIG_IPV6) */
	403
	404	static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
	405	{
	406	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
	407	}
	408
	409	static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
	410	{
842841ec	411	return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
cca45e05 IS	412	}
	413
	414	#endif /* IS_ENABLED(CONFIG_IPV6) */
	415
5ff4ff4f IS	416	static inline bool netif_is_vxlan(const struct net_device *dev)
	417	{
	418	return dev->rtnl_link_ops &&
	419	!strcmp(dev->rtnl_link_ops->kind, "vxlan");
	420	}
	421
9a997353 PM	422	struct switchdev_notifier_vxlan_fdb_info {
	423	struct switchdev_notifier_info info; /* must be first */
	424	union vxlan_addr remote_ip;
	425	__be16 remote_port;
	426	__be32 remote_vni;
	427	u32 remote_ifindex;
	428	u8 eth_addr[ETH_ALEN];
	429	__be32 vni;
0efe1173	430	bool offloaded;
45598c1c	431	bool added_by_user;
9a997353 PM	432	};
9a997353 PM	433
1941f1d6 PM	434	#if IS_ENABLED(CONFIG_VXLAN)
	435	int vxlan_fdb_find_uc(struct net_device dev, const u8 mac, __be32 vni,
	436	struct switchdev_notifier_vxlan_fdb_info *fdb_info);
4f89f5b5	437	int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
4c59b7d1 PM	438	struct notifier_block *nb,
4c59b7d1 PM	439	struct netlink_ext_ack *extack);
e5ff4b19	440	void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);
4f89f5b5	441
1941f1d6 PM	442	#else
	443	static inline int
	444	vxlan_fdb_find_uc(struct net_device dev, const u8 mac, __be32 vni,
	445	struct switchdev_notifier_vxlan_fdb_info *fdb_info)
	446	{
	447	return -ENOENT;
	448	}
4f89f5b5 PM	449
4f89f5b5 PM	450	static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
4c59b7d1 PM	451	struct notifier_block *nb,
4c59b7d1 PM	452	struct netlink_ext_ack *extack)
4f89f5b5 PM	453	{
	454	return -EOPNOTSUPP;
	455	}
e5ff4b19 PM	456
	457	static inline void
	458	vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
	459	{
	460	}
1941f1d6 PM	461	#endif
1941f1d6 PM	462
70fb0828 RP	463	static inline void vxlan_flag_attr_error(int attrtype,
	464	struct netlink_ext_ack *extack)
	465	{
	466	#define VXLAN_FLAG(flg) \
	467	case IFLA_VXLAN_##flg: \
	468	NL_SET_ERR_MSG_MOD(extack, \
	469	"cannot change " #flg " flag"); \
	470	break
	471	switch (attrtype) {
	472	VXLAN_FLAG(TTL_INHERIT);
	473	VXLAN_FLAG(LEARNING);
	474	VXLAN_FLAG(PROXY);
	475	VXLAN_FLAG(RSC);
	476	VXLAN_FLAG(L2MISS);
	477	VXLAN_FLAG(L3MISS);
	478	VXLAN_FLAG(COLLECT_METADATA);
	479	VXLAN_FLAG(UDP_ZERO_CSUM6_TX);
	480	VXLAN_FLAG(UDP_ZERO_CSUM6_RX);
	481	VXLAN_FLAG(REMCSUM_TX);
	482	VXLAN_FLAG(REMCSUM_RX);
	483	VXLAN_FLAG(GBP);
	484	VXLAN_FLAG(GPE);
	485	VXLAN_FLAG(REMCSUM_NOPARTIAL);
	486	default:
	487	NL_SET_ERR_MSG_MOD(extack, \
	488	"cannot change flag");
	489	break;
	490	}
	491	#undef VXLAN_FLAG
	492	}
	493
1274e1cc RP	494	static inline bool vxlan_fdb_nh_path_select(struct nexthop *nh,
	495	int hash,
	496	struct vxlan_rdst *rdst)
	497	{
	498	struct fib_nh_common *nhc;
	499
	500	nhc = nexthop_path_fdb_result(nh, hash);
	501	if (unlikely(!nhc))
	502	return false;
	503
	504	switch (nhc->nhc_gw_family) {
	505	case AF_INET:
	506	rdst->remote_ip.sin.sin_addr.s_addr = nhc->nhc_gw.ipv4;
	507	rdst->remote_ip.sa.sa_family = AF_INET;
	508	break;
	509	case AF_INET6:
	510	rdst->remote_ip.sin6.sin6_addr = nhc->nhc_gw.ipv6;
	511	rdst->remote_ip.sa.sa_family = AF_INET6;
	512	break;
	513	}
	514
	515	return true;
	516	}
	517
48e92c44	518	#endif