xref: /linux/include/net/vxlan.h (revision d060296cc0300ae8ed08004ebd3994bf325fa257)
1 #ifndef __NET_VXLAN_H
2 #define __NET_VXLAN_H 1
3 
4 #include <linux/if_vlan.h>
5 #include <net/udp_tunnel.h>
6 #include <net/dst_metadata.h>
7 #include <net/udp_tunnel.h>
8 
9 /* VXLAN protocol (RFC 7348) header:
10  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
11  * |R|R|R|R|I|R|R|R|               Reserved                        |
12  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
13  * |                VXLAN Network Identifier (VNI) |   Reserved    |
14  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
15  *
16  * I = VXLAN Network Identifier (VNI) present.
17  */
18 struct vxlanhdr {
19 	__be32 vx_flags;
20 	__be32 vx_vni;
21 };
22 
23 /* VXLAN header flags. */
24 #define VXLAN_HF_VNI	cpu_to_be32(BIT(27))
25 
26 #define VXLAN_N_VID     (1u << 24)
27 #define VXLAN_VID_MASK  (VXLAN_N_VID - 1)
28 #define VXLAN_VNI_MASK	cpu_to_be32(VXLAN_VID_MASK << 8)
29 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
30 
31 #define VNI_HASH_BITS	10
32 #define VNI_HASH_SIZE	(1<<VNI_HASH_BITS)
33 #define FDB_HASH_BITS	8
34 #define FDB_HASH_SIZE	(1<<FDB_HASH_BITS)
35 
36 /* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
37  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
38  * |R|R|R|R|I|R|R|R|R|R|C|              Reserved                   |
39  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
40  * |           VXLAN Network Identifier (VNI)      |O| Csum start  |
41  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
42  *
43  * C = Remote checksum offload bit. When set indicates that the
44  *     remote checksum offload data is present.
45  *
46  * O = Offset bit. Indicates the checksum offset relative to
47  *     checksum start.
48  *
49  * Csum start = Checksum start divided by two.
50  *
51  * http://tools.ietf.org/html/draft-herbert-vxlan-rco
52  */
53 
54 /* VXLAN-RCO header flags. */
55 #define VXLAN_HF_RCO	cpu_to_be32(BIT(21))
56 
57 /* Remote checksum offload header option */
58 #define VXLAN_RCO_MASK	cpu_to_be32(0x7f)  /* Last byte of vni field */
59 #define VXLAN_RCO_UDP	cpu_to_be32(0x80)  /* Indicate UDP RCO (TCP when not set *) */
60 #define VXLAN_RCO_SHIFT	1		   /* Left shift of start */
61 #define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
62 #define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)
63 
64 /*
65  * VXLAN Group Based Policy Extension (VXLAN_F_GBP):
66  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
67  * |G|R|R|R|I|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
68  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
69  * |                VXLAN Network Identifier (VNI) |   Reserved    |
70  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
71  *
72  * G = Group Policy ID present.
73  *
74  * D = Don't Learn bit. When set, this bit indicates that the egress
75  *     VTEP MUST NOT learn the source address of the encapsulated frame.
76  *
77  * A = Indicates that the group policy has already been applied to
78  *     this packet. Policies MUST NOT be applied by devices when the
79  *     A bit is set.
80  *
81  * https://tools.ietf.org/html/draft-smith-vxlan-group-policy
82  */
83 struct vxlanhdr_gbp {
84 	u8	vx_flags;
85 #ifdef __LITTLE_ENDIAN_BITFIELD
86 	u8	reserved_flags1:3,
87 		policy_applied:1,
88 		reserved_flags2:2,
89 		dont_learn:1,
90 		reserved_flags3:1;
91 #elif defined(__BIG_ENDIAN_BITFIELD)
92 	u8	reserved_flags1:1,
93 		dont_learn:1,
94 		reserved_flags2:2,
95 		policy_applied:1,
96 		reserved_flags3:3;
97 #else
98 #error	"Please fix <asm/byteorder.h>"
99 #endif
100 	__be16	policy_id;
101 	__be32	vx_vni;
102 };
103 
104 /* VXLAN-GBP header flags. */
105 #define VXLAN_HF_GBP	cpu_to_be32(BIT(31))
106 
107 #define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP | cpu_to_be32(0xFFFFFF))
108 
109 /* skb->mark mapping
110  *
111  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112  * |R|R|R|R|R|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
113  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114  */
115 #define VXLAN_GBP_DONT_LEARN		(BIT(6) << 16)
116 #define VXLAN_GBP_POLICY_APPLIED	(BIT(3) << 16)
117 #define VXLAN_GBP_ID_MASK		(0xFFFF)
118 
119 /*
120  * VXLAN Generic Protocol Extension (VXLAN_F_GPE):
121  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
122  * |R|R|Ver|I|P|R|O|       Reserved                |Next Protocol  |
123  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
124  * |                VXLAN Network Identifier (VNI) |   Reserved    |
125  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
126  *
127  * Ver = Version. Indicates VXLAN GPE protocol version.
128  *
129  * P = Next Protocol Bit. The P bit is set to indicate that the
130  *     Next Protocol field is present.
131  *
132  * O = OAM Flag Bit. The O bit is set to indicate that the packet
133  *     is an OAM packet.
134  *
135  * Next Protocol = This 8 bit field indicates the protocol header
136  * immediately following the VXLAN GPE header.
137  *
138  * https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
139  */
140 
141 struct vxlanhdr_gpe {
142 #if defined(__LITTLE_ENDIAN_BITFIELD)
143 	u8	oam_flag:1,
144 		reserved_flags1:1,
145 		np_applied:1,
146 		instance_applied:1,
147 		version:2,
148 reserved_flags2:2;
149 #elif defined(__BIG_ENDIAN_BITFIELD)
150 	u8	reserved_flags2:2,
151 		version:2,
152 		instance_applied:1,
153 		np_applied:1,
154 		reserved_flags1:1,
155 		oam_flag:1;
156 #endif
157 	u8	reserved_flags3;
158 	u8	reserved_flags4;
159 	u8	next_protocol;
160 	__be32	vx_vni;
161 };
162 
163 /* VXLAN-GPE header flags. */
164 #define VXLAN_HF_VER	cpu_to_be32(BIT(29) | BIT(28))
165 #define VXLAN_HF_NP	cpu_to_be32(BIT(26))
166 #define VXLAN_HF_OAM	cpu_to_be32(BIT(24))
167 
168 #define VXLAN_GPE_USED_BITS (VXLAN_HF_VER | VXLAN_HF_NP | VXLAN_HF_OAM | \
169 			     cpu_to_be32(0xff))
170 
171 /* VXLAN-GPE header Next Protocol. */
172 #define VXLAN_GPE_NP_IPV4      0x01
173 #define VXLAN_GPE_NP_IPV6      0x02
174 #define VXLAN_GPE_NP_ETHERNET  0x03
175 #define VXLAN_GPE_NP_NSH       0x04
176 
177 struct vxlan_metadata {
178 	u32		gbp;
179 };
180 
181 /* per UDP socket information */
182 struct vxlan_sock {
183 	struct hlist_node hlist;
184 	struct socket	 *sock;
185 	struct hlist_head vni_list[VNI_HASH_SIZE];
186 	refcount_t	  refcnt;
187 	u32		  flags;
188 };
189 
190 union vxlan_addr {
191 	struct sockaddr_in sin;
192 	struct sockaddr_in6 sin6;
193 	struct sockaddr sa;
194 };
195 
196 struct vxlan_rdst {
197 	union vxlan_addr	 remote_ip;
198 	__be16			 remote_port;
199 	__be32			 remote_vni;
200 	u32			 remote_ifindex;
201 	struct list_head	 list;
202 	struct rcu_head		 rcu;
203 	struct dst_cache	 dst_cache;
204 };
205 
206 struct vxlan_config {
207 	union vxlan_addr	remote_ip;
208 	union vxlan_addr	saddr;
209 	__be32			vni;
210 	int			remote_ifindex;
211 	int			mtu;
212 	__be16			dst_port;
213 	u16			port_min;
214 	u16			port_max;
215 	u8			tos;
216 	u8			ttl;
217 	__be32			label;
218 	u32			flags;
219 	unsigned long		age_interval;
220 	unsigned int		addrmax;
221 	bool			no_share;
222 };
223 
224 struct vxlan_dev_node {
225 	struct hlist_node hlist;
226 	struct vxlan_dev *vxlan;
227 };
228 
229 /* Pseudo network device */
230 struct vxlan_dev {
231 	struct vxlan_dev_node hlist4;	/* vni hash table for IPv4 socket */
232 #if IS_ENABLED(CONFIG_IPV6)
233 	struct vxlan_dev_node hlist6;	/* vni hash table for IPv6 socket */
234 #endif
235 	struct list_head  next;		/* vxlan's per namespace list */
236 	struct vxlan_sock __rcu *vn4_sock;	/* listening socket for IPv4 */
237 #if IS_ENABLED(CONFIG_IPV6)
238 	struct vxlan_sock __rcu *vn6_sock;	/* listening socket for IPv6 */
239 #endif
240 	struct net_device *dev;
241 	struct net	  *net;		/* netns for packet i/o */
242 	struct vxlan_rdst default_dst;	/* default destination */
243 
244 	struct timer_list age_timer;
245 	spinlock_t	  hash_lock;
246 	unsigned int	  addrcnt;
247 	struct gro_cells  gro_cells;
248 
249 	struct vxlan_config	cfg;
250 
251 	struct hlist_head fdb_head[FDB_HASH_SIZE];
252 };
253 
254 #define VXLAN_F_LEARN			0x01
255 #define VXLAN_F_PROXY			0x02
256 #define VXLAN_F_RSC			0x04
257 #define VXLAN_F_L2MISS			0x08
258 #define VXLAN_F_L3MISS			0x10
259 #define VXLAN_F_IPV6			0x20
260 #define VXLAN_F_UDP_ZERO_CSUM_TX	0x40
261 #define VXLAN_F_UDP_ZERO_CSUM6_TX	0x80
262 #define VXLAN_F_UDP_ZERO_CSUM6_RX	0x100
263 #define VXLAN_F_REMCSUM_TX		0x200
264 #define VXLAN_F_REMCSUM_RX		0x400
265 #define VXLAN_F_GBP			0x800
266 #define VXLAN_F_REMCSUM_NOPARTIAL	0x1000
267 #define VXLAN_F_COLLECT_METADATA	0x2000
268 #define VXLAN_F_GPE			0x4000
269 #define VXLAN_F_IPV6_LINKLOCAL		0x8000
270 
271 /* Flags that are used in the receive path. These flags must match in
272  * order for a socket to be shareable
273  */
274 #define VXLAN_F_RCV_FLAGS		(VXLAN_F_GBP |			\
275 					 VXLAN_F_GPE |			\
276 					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
277 					 VXLAN_F_REMCSUM_RX |		\
278 					 VXLAN_F_REMCSUM_NOPARTIAL |	\
279 					 VXLAN_F_COLLECT_METADATA)
280 
281 /* Flags that can be set together with VXLAN_F_GPE. */
282 #define VXLAN_F_ALLOWED_GPE		(VXLAN_F_GPE |			\
283 					 VXLAN_F_IPV6 |			\
284 					 VXLAN_F_IPV6_LINKLOCAL |	\
285 					 VXLAN_F_UDP_ZERO_CSUM_TX |	\
286 					 VXLAN_F_UDP_ZERO_CSUM6_TX |	\
287 					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
288 					 VXLAN_F_COLLECT_METADATA)
289 
290 struct net_device *vxlan_dev_create(struct net *net, const char *name,
291 				    u8 name_assign_type, struct vxlan_config *conf);
292 
293 static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
294 						     netdev_features_t features)
295 {
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) &&
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) !=
316 	      sizeof(struct udphdr) + sizeof(struct vxlanhdr)) ||
317 	     (skb->ip_summed != CHECKSUM_NONE &&
318 	      !can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
319 		return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
320 
321 	return features;
322 }
323 
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 
329 static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
330 {
331 	return (struct vxlanhdr *)(udp_hdr(skb) + 1);
332 }
333 
334 static inline __be32 vxlan_vni(__be32 vni_field)
335 {
336 #if defined(__BIG_ENDIAN)
337 	return (__force __be32)((__force u32)vni_field >> 8);
338 #else
339 	return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
340 #endif
341 }
342 
343 static inline __be32 vxlan_vni_field(__be32 vni)
344 {
345 #if defined(__BIG_ENDIAN)
346 	return (__force __be32)((__force u32)vni << 8);
347 #else
348 	return (__force __be32)((__force u32)vni >> 8);
349 #endif
350 }
351 
352 static inline size_t vxlan_rco_start(__be32 vni_field)
353 {
354 	return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
355 }
356 
357 static inline size_t vxlan_rco_offset(__be32 vni_field)
358 {
359 	return (vni_field & VXLAN_RCO_UDP) ?
360 		offsetof(struct udphdr, check) :
361 		offsetof(struct tcphdr, check);
362 }
363 
364 static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
365 {
366 	__be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
367 
368 	if (offset == offsetof(struct udphdr, check))
369 		vni_field |= VXLAN_RCO_UDP;
370 	return vni_field;
371 }
372 
373 static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
374 {
375 	return vs->sock->sk->sk_family;
376 }
377 
378 #endif
379