[deliverable/linux.git] / include / net / vxlan.h

#ifndef __NET_VXLAN_H
#define __NET_VXLAN_H 1

#include <linux/ip.h>
#include <linux/ipv6.h>
#include <linux/if_vlan.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/udp.h>
#include <net/dst_metadata.h>

#define VNI_HASH_BITS	10
#define VNI_HASH_SIZE	(1<<VNI_HASH_BITS)

/*
 * VXLAN Group Based Policy Extension:
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |1|-|-|-|1|-|-|-|R|D|R|R|A|R|R|R|        Group Policy ID        |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |                VXLAN Network Identifier (VNI) |   Reserved    |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *
 * 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.
 *
 * [0] 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;
};

#define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP | 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 protocol header:
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |G|R|R|R|I|R|R|C|               Reserved                        |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 * |                VXLAN Network Identifier (VNI) |   Reserved    |
 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 *
 * G = 1	Group Policy (VXLAN-GBP)
 * I = 1	VXLAN Network Identifier (VNI) present
 * C = 1	Remote checksum offload (RCO)
 */
struct vxlanhdr {
	__be32 vx_flags;
	__be32 vx_vni;
};

/* VXLAN header flags. */
#define VXLAN_HF_RCO BIT(24)
#define VXLAN_HF_VNI BIT(27)
#define VXLAN_HF_GBP BIT(31)

/* Remote checksum offload header option */
#define VXLAN_RCO_MASK  0x7f    /* Last byte of vni field */
#define VXLAN_RCO_UDP   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 (VXLAN_RCO_MASK << VXLAN_RCO_SHIFT)

#define VXLAN_N_VID     (1u << 24)
#define VXLAN_VID_MASK  (VXLAN_N_VID - 1)
#define VXLAN_VNI_MASK  (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)

struct vxlan_metadata {
	u32		gbp;
};

/* per UDP socket information */
struct vxlan_sock {
	struct hlist_node hlist;
	struct work_struct del_work;
	struct socket	 *sock;
	struct rcu_head	  rcu;
	struct hlist_head vni_list[VNI_HASH_SIZE];
	atomic_t	  refcnt;
	struct udp_offload udp_offloads;
	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;
	u32			 remote_vni;
	u32			 remote_ifindex;
	struct list_head	 list;
	struct rcu_head		 rcu;
};

struct vxlan_config {
	union vxlan_addr	remote_ip;
	union vxlan_addr	saddr;
	u32			vni;
	int			remote_ifindex;
	int			mtu;
	__be16			dst_port;
	__u16			port_min;
	__u16			port_max;
	__u8			tos;
	__u8			ttl;
	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_CSUM		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

/* 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_UDP_ZERO_CSUM6_RX |	\
					 VXLAN_F_REMCSUM_RX |		\
					 VXLAN_F_REMCSUM_NOPARTIAL |	\
					 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))))
		return features & ~(NETIF_F_ALL_CSUM | 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)

#if IS_ENABLED(CONFIG_VXLAN)
void vxlan_get_rx_port(struct net_device *netdev);
#else
static inline void vxlan_get_rx_port(struct net_device *netdev)
{
}
#endif

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 JG	4	#include <linux/ip.h>
	5	#include <linux/ipv6.h>
	6	#include <linux/if_vlan.h>
012a5729 PS	7	#include <linux/skbuff.h>
	8	#include <linux/netdevice.h>
	9	#include <linux/udp.h>
ee122c79	10	#include <net/dst_metadata.h>
012a5729 PS	11
	12	#define VNI_HASH_BITS 10
	13	#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
	14
3511494c TG	15	/*
	16	* VXLAN Group Based Policy Extension:
	17	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	18	* \|1\|-\|-\|-\|1\|-\|-\|-\|R\|D\|R\|R\|A\|R\|R\|R\| Group Policy ID \|
	19	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	20	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	21	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	22	*
	23	* D = Don't Learn bit. When set, this bit indicates that the egress
	24	* VTEP MUST NOT learn the source address of the encapsulated frame.
	25	*
	26	* A = Indicates that the group policy has already been applied to
	27	* this packet. Policies MUST NOT be applied by devices when the
	28	* A bit is set.
	29	*
	30	* [0] https://tools.ietf.org/html/draft-smith-vxlan-group-policy
	31	*/
	32	struct vxlanhdr_gbp {
	33	__u8 vx_flags;
	34	#ifdef __LITTLE_ENDIAN_BITFIELD
	35	__u8 reserved_flags1:3,
	36	policy_applied:1,
	37	reserved_flags2:2,
	38	dont_learn:1,
	39	reserved_flags3:1;
	40	#elif defined(__BIG_ENDIAN_BITFIELD)
	41	__u8 reserved_flags1:1,
	42	dont_learn:1,
	43	reserved_flags2:2,
	44	policy_applied:1,
	45	reserved_flags3:3;
	46	#else
	47	#error "Please fix <asm/byteorder.h>"
	48	#endif
	49	__be16 policy_id;
	50	__be32 vx_vni;
	51	};
	52
	53	#define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP \| 0xFFFFFF)
	54
	55	/* skb->mark mapping
	56	*
	57	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	58	* \|R\|R\|R\|R\|R\|R\|R\|R\|R\|D\|R\|R\|A\|R\|R\|R\| Group Policy ID \|
	59	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	60	*/
	61	#define VXLAN_GBP_DONT_LEARN (BIT(6) << 16)
	62	#define VXLAN_GBP_POLICY_APPLIED (BIT(3) << 16)
	63	#define VXLAN_GBP_ID_MASK (0xFFFF)
	64
	65	/* VXLAN protocol header:
	66	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	67	* \|G\|R\|R\|R\|I\|R\|R\|C\| Reserved \|
	68	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	69	* \| VXLAN Network Identifier (VNI) \| Reserved \|
	70	* +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
	71	*
	72	* G = 1 Group Policy (VXLAN-GBP)
	73	* I = 1 VXLAN Network Identifier (VNI) present
	74	* C = 1 Remote checksum offload (RCO)
	75	*/
11bf7828 JS	76	struct vxlanhdr {
	77	__be32 vx_flags;
	78	__be32 vx_vni;
	79	};
	80
3bf39475	81	/* VXLAN header flags. */
3511494c TG	82	#define VXLAN_HF_RCO BIT(24)
	83	#define VXLAN_HF_VNI BIT(27)
	84	#define VXLAN_HF_GBP BIT(31)
dfd8645e TH	85
	86	/* Remote checksum offload header option */
	87	#define VXLAN_RCO_MASK 0x7f /* Last byte of vni field */
	88	#define VXLAN_RCO_UDP 0x80 /* Indicate UDP RCO (TCP when not set ) /
	89	#define VXLAN_RCO_SHIFT 1 /* Left shift of start */
	90	#define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
	91	#define VXLAN_MAX_REMCSUM_START (VXLAN_RCO_MASK << VXLAN_RCO_SHIFT)
3bf39475 TH	92
	93	#define VXLAN_N_VID (1u << 24)
	94	#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
40fb70f3	95	#define VXLAN_VNI_MASK (VXLAN_VID_MASK << 8)
3bf39475 TH	96	#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
3bf39475 TH	97
0dfbdf41 TG	98	#define VNI_HASH_BITS 10
	99	#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
	100	#define FDB_HASH_BITS 8
	101	#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
	102
3511494c	103	struct vxlan_metadata {
3511494c TG	104	u32 gbp;
	105	};
	106
012a5729 PS	107	/* per UDP socket information */
	108	struct vxlan_sock {
	109	struct hlist_node hlist;
012a5729 PS	110	struct work_struct del_work;
	111	struct socket *sock;
	112	struct rcu_head rcu;
	113	struct hlist_head vni_list[VNI_HASH_SIZE];
	114	atomic_t refcnt;
dc01e7d3	115	struct udp_offload udp_offloads;
dfd8645e	116	u32 flags;
012a5729 PS	117	};
012a5729 PS	118
0dfbdf41 TG	119	union vxlan_addr {
	120	struct sockaddr_in sin;
	121	struct sockaddr_in6 sin6;
	122	struct sockaddr sa;
	123	};
	124
	125	struct vxlan_rdst {
	126	union vxlan_addr remote_ip;
	127	__be16 remote_port;
	128	u32 remote_vni;
	129	u32 remote_ifindex;
	130	struct list_head list;
	131	struct rcu_head rcu;
	132	};
	133
	134	struct vxlan_config {
	135	union vxlan_addr remote_ip;
	136	union vxlan_addr saddr;
	137	u32 vni;
	138	int remote_ifindex;
	139	int mtu;
	140	__be16 dst_port;
	141	__u16 port_min;
	142	__u16 port_max;
	143	__u8 tos;
	144	__u8 ttl;
	145	u32 flags;
	146	unsigned long age_interval;
	147	unsigned int addrmax;
	148	bool no_share;
	149	};
	150
	151	/* Pseudo network device */
	152	struct vxlan_dev {
	153	struct hlist_node hlist; /* vni hash table */
	154	struct list_head next; /* vxlan's per namespace list */
b1be00a6 JB	155	struct vxlan_sock vn4_sock; / listening socket for IPv4 */
	156	#if IS_ENABLED(CONFIG_IPV6)
	157	struct vxlan_sock vn6_sock; / listening socket for IPv6 */
	158	#endif
0dfbdf41 TG	159	struct net_device *dev;
	160	struct net net; / netns for packet i/o */
	161	struct vxlan_rdst default_dst; /* default destination */
	162	u32 flags; /* VXLAN_F_* in vxlan.h */
	163
	164	struct timer_list age_timer;
	165	spinlock_t hash_lock;
	166	unsigned int addrcnt;
58ce31cc	167	struct gro_cells gro_cells;
0dfbdf41 TG	168
	169	struct vxlan_config cfg;
	170
	171	struct hlist_head fdb_head[FDB_HASH_SIZE];
	172	};
	173
359a0ea9 TH	174	#define VXLAN_F_LEARN 0x01
	175	#define VXLAN_F_PROXY 0x02
	176	#define VXLAN_F_RSC 0x04
	177	#define VXLAN_F_L2MISS 0x08
	178	#define VXLAN_F_L3MISS 0x10
	179	#define VXLAN_F_IPV6 0x20
	180	#define VXLAN_F_UDP_CSUM 0x40
	181	#define VXLAN_F_UDP_ZERO_CSUM6_TX 0x80
	182	#define VXLAN_F_UDP_ZERO_CSUM6_RX 0x100
dfd8645e TH	183	#define VXLAN_F_REMCSUM_TX 0x200
dfd8645e TH	184	#define VXLAN_F_REMCSUM_RX 0x400
3511494c	185	#define VXLAN_F_GBP 0x800
0ace2ca8	186	#define VXLAN_F_REMCSUM_NOPARTIAL 0x1000
ee122c79	187	#define VXLAN_F_COLLECT_METADATA 0x2000
359a0ea9	188
d299ce14	189	/* Flags that are used in the receive path. These flags must match in
af33c1ad TH	190	* order for a socket to be shareable
	191	*/
	192	#define VXLAN_F_RCV_FLAGS (VXLAN_F_GBP \| \
	193	VXLAN_F_UDP_ZERO_CSUM6_RX \| \
0ace2ca8	194	VXLAN_F_REMCSUM_RX \| \
ee122c79	195	VXLAN_F_REMCSUM_NOPARTIAL \| \
da8b43c0	196	VXLAN_F_COLLECT_METADATA)
ac5132d1	197
0dfbdf41 TG	198	struct net_device vxlan_dev_create(struct net net, const char *name,
	199	u8 name_assign_type, struct vxlan_config *conf);
	200
b1be00a6 JB	201	static inline __be16 vxlan_dev_dst_port(struct vxlan_dev *vxlan,
b1be00a6 JB	202	unsigned short family)
614732ea	203	{
b1be00a6 JB	204	#if IS_ENABLED(CONFIG_IPV6)
	205	if (family == AF_INET6)
	206	return inet_sk(vxlan->vn6_sock->sock->sk)->inet_sport;
	207	#endif
	208	return inet_sk(vxlan->vn4_sock->sock->sk)->inet_sport;
614732ea	209	}
49560532	210
5f35227e JG	211	static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
5f35227e JG	212	netdev_features_t features)
11bf7828	213	{
5f35227e JG	214	u8 l4_hdr = 0;
	215
	216	if (!skb->encapsulation)
	217	return features;
	218
	219	switch (vlan_get_protocol(skb)) {
	220	case htons(ETH_P_IP):
	221	l4_hdr = ip_hdr(skb)->protocol;
	222	break;
	223	case htons(ETH_P_IPV6):
	224	l4_hdr = ipv6_hdr(skb)->nexthdr;
	225	break;
	226	default:
	227	return features;;
	228	}
	229
	230	if ((l4_hdr == IPPROTO_UDP) &&
11bf7828 JS	231	(skb->inner_protocol_type != ENCAP_TYPE_ETHER \|\|
	232	skb->inner_protocol != htons(ETH_P_TEB) \|\|
	233	(skb_inner_mac_header(skb) - skb_transport_header(skb) !=
	234	sizeof(struct udphdr) + sizeof(struct vxlanhdr))))
5f35227e	235	return features & ~(NETIF_F_ALL_CSUM \| NETIF_F_GSO_MASK);
11bf7828	236
5f35227e	237	return features;
11bf7828	238	}
23e62de3	239
e6cd988c JG	240	/* IP header + UDP + VXLAN + Ethernet header */
	241	#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
	242	/* IPv6 header + UDP + VXLAN + Ethernet header */
	243	#define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
	244
	245	#if IS_ENABLED(CONFIG_VXLAN)
53cf5275	246	void vxlan_get_rx_port(struct net_device *netdev);
e6cd988c JG	247	#else
	248	static inline void vxlan_get_rx_port(struct net_device *netdev)
	249	{
	250	}
	251	#endif
705cc62f JB	252
	253	static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
	254	{
	255	return vs->sock->sk->sk_family;
	256	}
48e92c44 JB	257
48e92c44 JB	258	#endif