73ed2e951c020d28c21457d907c55764ea7b3d88
[deliverable/linux.git] / include / net / vxlan.h
1 #ifndef __NET_VXLAN_H
2 #define __NET_VXLAN_H 1
3
4 #include <linux/ip.h>
5 #include <linux/ipv6.h>
6 #include <linux/if_vlan.h>
7 #include <linux/skbuff.h>
8 #include <linux/netdevice.h>
9 #include <linux/udp.h>
10 #include <net/dst_metadata.h>
11
12 /* VXLAN protocol (RFC 7348) header:
13 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
14 * |R|R|R|R|I|R|R|R| Reserved |
15 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
16 * | VXLAN Network Identifier (VNI) | Reserved |
17 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
18 *
19 * I = VXLAN Network Identifier (VNI) present.
20 */
21 struct vxlanhdr {
22 __be32 vx_flags;
23 __be32 vx_vni;
24 };
25
26 /* VXLAN header flags. */
27 #define VXLAN_HF_VNI cpu_to_be32(BIT(27))
28
29 #define VXLAN_N_VID (1u << 24)
30 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
31 #define VXLAN_VNI_MASK cpu_to_be32(VXLAN_VID_MASK << 8)
32 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
33
34 #define VNI_HASH_BITS 10
35 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
36 #define FDB_HASH_BITS 8
37 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
38
39 /* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
40 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
41 * |R|R|R|R|I|R|R|R|R|R|C| Reserved |
42 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
43 * | VXLAN Network Identifier (VNI) |O| Csum start |
44 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
45 *
46 * C = Remote checksum offload bit. When set indicates that the
47 * remote checksum offload data is present.
48 *
49 * O = Offset bit. Indicates the checksum offset relative to
50 * checksum start.
51 *
52 * Csum start = Checksum start divided by two.
53 *
54 * http://tools.ietf.org/html/draft-herbert-vxlan-rco
55 */
56
57 /* VXLAN-RCO header flags. */
58 #define VXLAN_HF_RCO cpu_to_be32(BIT(21))
59
60 /* Remote checksum offload header option */
61 #define VXLAN_RCO_MASK cpu_to_be32(0x7f) /* Last byte of vni field */
62 #define VXLAN_RCO_UDP cpu_to_be32(0x80) /* Indicate UDP RCO (TCP when not set *) */
63 #define VXLAN_RCO_SHIFT 1 /* Left shift of start */
64 #define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
65 #define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)
66
67 /*
68 * VXLAN Group Based Policy Extension (VXLAN_F_GBP):
69 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
70 * |G|R|R|R|I|R|R|R|R|D|R|R|A|R|R|R| Group Policy ID |
71 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
72 * | VXLAN Network Identifier (VNI) | Reserved |
73 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
74 *
75 * G = Group Policy ID present.
76 *
77 * D = Don't Learn bit. When set, this bit indicates that the egress
78 * VTEP MUST NOT learn the source address of the encapsulated frame.
79 *
80 * A = Indicates that the group policy has already been applied to
81 * this packet. Policies MUST NOT be applied by devices when the
82 * A bit is set.
83 *
84 * https://tools.ietf.org/html/draft-smith-vxlan-group-policy
85 */
86 struct vxlanhdr_gbp {
87 u8 vx_flags;
88 #ifdef __LITTLE_ENDIAN_BITFIELD
89 u8 reserved_flags1:3,
90 policy_applied:1,
91 reserved_flags2:2,
92 dont_learn:1,
93 reserved_flags3:1;
94 #elif defined(__BIG_ENDIAN_BITFIELD)
95 u8 reserved_flags1:1,
96 dont_learn:1,
97 reserved_flags2:2,
98 policy_applied:1,
99 reserved_flags3:3;
100 #else
101 #error "Please fix <asm/byteorder.h>"
102 #endif
103 __be16 policy_id;
104 __be32 vx_vni;
105 };
106
107 /* VXLAN-GBP header flags. */
108 #define VXLAN_HF_GBP cpu_to_be32(BIT(31))
109
110 #define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP | cpu_to_be32(0xFFFFFF))
111
112 /* skb->mark mapping
113 *
114 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115 * |R|R|R|R|R|R|R|R|R|D|R|R|A|R|R|R| Group Policy ID |
116 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117 */
118 #define VXLAN_GBP_DONT_LEARN (BIT(6) << 16)
119 #define VXLAN_GBP_POLICY_APPLIED (BIT(3) << 16)
120 #define VXLAN_GBP_ID_MASK (0xFFFF)
121
122 struct vxlan_metadata {
123 u32 gbp;
124 };
125
126 /* per UDP socket information */
127 struct vxlan_sock {
128 struct hlist_node hlist;
129 struct work_struct del_work;
130 struct socket *sock;
131 struct rcu_head rcu;
132 struct hlist_head vni_list[VNI_HASH_SIZE];
133 atomic_t refcnt;
134 struct udp_offload udp_offloads;
135 u32 flags;
136 };
137
138 union vxlan_addr {
139 struct sockaddr_in sin;
140 struct sockaddr_in6 sin6;
141 struct sockaddr sa;
142 };
143
144 struct vxlan_rdst {
145 union vxlan_addr remote_ip;
146 __be16 remote_port;
147 __be32 remote_vni;
148 u32 remote_ifindex;
149 struct list_head list;
150 struct rcu_head rcu;
151 struct dst_cache dst_cache;
152 };
153
154 struct vxlan_config {
155 union vxlan_addr remote_ip;
156 union vxlan_addr saddr;
157 __be32 vni;
158 int remote_ifindex;
159 int mtu;
160 __be16 dst_port;
161 u16 port_min;
162 u16 port_max;
163 u8 tos;
164 u8 ttl;
165 __be32 label;
166 u32 flags;
167 unsigned long age_interval;
168 unsigned int addrmax;
169 bool no_share;
170 };
171
172 /* Pseudo network device */
173 struct vxlan_dev {
174 struct hlist_node hlist; /* vni hash table */
175 struct list_head next; /* vxlan's per namespace list */
176 struct vxlan_sock *vn4_sock; /* listening socket for IPv4 */
177 #if IS_ENABLED(CONFIG_IPV6)
178 struct vxlan_sock *vn6_sock; /* listening socket for IPv6 */
179 #endif
180 struct net_device *dev;
181 struct net *net; /* netns for packet i/o */
182 struct vxlan_rdst default_dst; /* default destination */
183 u32 flags; /* VXLAN_F_* in vxlan.h */
184
185 struct timer_list age_timer;
186 spinlock_t hash_lock;
187 unsigned int addrcnt;
188 struct gro_cells gro_cells;
189
190 struct vxlan_config cfg;
191
192 struct hlist_head fdb_head[FDB_HASH_SIZE];
193 };
194
195 #define VXLAN_F_LEARN 0x01
196 #define VXLAN_F_PROXY 0x02
197 #define VXLAN_F_RSC 0x04
198 #define VXLAN_F_L2MISS 0x08
199 #define VXLAN_F_L3MISS 0x10
200 #define VXLAN_F_IPV6 0x20
201 #define VXLAN_F_UDP_ZERO_CSUM_TX 0x40
202 #define VXLAN_F_UDP_ZERO_CSUM6_TX 0x80
203 #define VXLAN_F_UDP_ZERO_CSUM6_RX 0x100
204 #define VXLAN_F_REMCSUM_TX 0x200
205 #define VXLAN_F_REMCSUM_RX 0x400
206 #define VXLAN_F_GBP 0x800
207 #define VXLAN_F_REMCSUM_NOPARTIAL 0x1000
208 #define VXLAN_F_COLLECT_METADATA 0x2000
209
210 /* Flags that are used in the receive path. These flags must match in
211 * order for a socket to be shareable
212 */
213 #define VXLAN_F_RCV_FLAGS (VXLAN_F_GBP | \
214 VXLAN_F_UDP_ZERO_CSUM6_RX | \
215 VXLAN_F_REMCSUM_RX | \
216 VXLAN_F_REMCSUM_NOPARTIAL | \
217 VXLAN_F_COLLECT_METADATA)
218
219 struct net_device *vxlan_dev_create(struct net *net, const char *name,
220 u8 name_assign_type, struct vxlan_config *conf);
221
222 static inline __be16 vxlan_dev_dst_port(struct vxlan_dev *vxlan,
223 unsigned short family)
224 {
225 #if IS_ENABLED(CONFIG_IPV6)
226 if (family == AF_INET6)
227 return inet_sk(vxlan->vn6_sock->sock->sk)->inet_sport;
228 #endif
229 return inet_sk(vxlan->vn4_sock->sock->sk)->inet_sport;
230 }
231
232 static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
233 netdev_features_t features)
234 {
235 u8 l4_hdr = 0;
236
237 if (!skb->encapsulation)
238 return features;
239
240 switch (vlan_get_protocol(skb)) {
241 case htons(ETH_P_IP):
242 l4_hdr = ip_hdr(skb)->protocol;
243 break;
244 case htons(ETH_P_IPV6):
245 l4_hdr = ipv6_hdr(skb)->nexthdr;
246 break;
247 default:
248 return features;;
249 }
250
251 if ((l4_hdr == IPPROTO_UDP) &&
252 (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
253 skb->inner_protocol != htons(ETH_P_TEB) ||
254 (skb_inner_mac_header(skb) - skb_transport_header(skb) !=
255 sizeof(struct udphdr) + sizeof(struct vxlanhdr))))
256 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
257
258 return features;
259 }
260
261 /* IP header + UDP + VXLAN + Ethernet header */
262 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
263 /* IPv6 header + UDP + VXLAN + Ethernet header */
264 #define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
265
266 static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
267 {
268 return (struct vxlanhdr *)(udp_hdr(skb) + 1);
269 }
270
271 static inline __be32 vxlan_vni(__be32 vni_field)
272 {
273 #if defined(__BIG_ENDIAN)
274 return (__force __be32)((__force u32)vni_field >> 8);
275 #else
276 return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
277 #endif
278 }
279
280 static inline __be32 vxlan_vni_field(__be32 vni)
281 {
282 #if defined(__BIG_ENDIAN)
283 return (__force __be32)((__force u32)vni << 8);
284 #else
285 return (__force __be32)((__force u32)vni >> 8);
286 #endif
287 }
288
289 static inline __be32 vxlan_tun_id_to_vni(__be64 tun_id)
290 {
291 #if defined(__BIG_ENDIAN)
292 return (__force __be32)tun_id;
293 #else
294 return (__force __be32)((__force u64)tun_id >> 32);
295 #endif
296 }
297
298 static inline __be64 vxlan_vni_to_tun_id(__be32 vni)
299 {
300 #if defined(__BIG_ENDIAN)
301 return (__force __be64)vni;
302 #else
303 return (__force __be64)((u64)(__force u32)vni << 32);
304 #endif
305 }
306
307 static inline size_t vxlan_rco_start(__be32 vni_field)
308 {
309 return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
310 }
311
312 static inline size_t vxlan_rco_offset(__be32 vni_field)
313 {
314 return (vni_field & VXLAN_RCO_UDP) ?
315 offsetof(struct udphdr, check) :
316 offsetof(struct tcphdr, check);
317 }
318
319 static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
320 {
321 __be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
322
323 if (offset == offsetof(struct udphdr, check))
324 vni_field |= VXLAN_RCO_UDP;
325 return vni_field;
326 }
327
328 #if IS_ENABLED(CONFIG_VXLAN)
329 void vxlan_get_rx_port(struct net_device *netdev);
330 #else
331 static inline void vxlan_get_rx_port(struct net_device *netdev)
332 {
333 }
334 #endif
335
336 static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
337 {
338 return vs->sock->sk->sk_family;
339 }
340
341 #endif
This page took 0.037216 seconds and 4 git commands to generate.