xref: /linux/include/net/vxlan.h (revision 3bdab16c55f57a24245c97d707241dd9b48d1a91)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __NET_VXLAN_H
3 #define __NET_VXLAN_H 1
4 
5 #include <linux/if_vlan.h>
6 #include <net/udp_tunnel.h>
7 #include <net/dst_metadata.h>
8 #include <net/rtnetlink.h>
9 #include <net/switchdev.h>
10 
11 #define IANA_VXLAN_UDP_PORT     4789
12 
13 /* VXLAN protocol (RFC 7348) header:
14  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
15  * |R|R|R|R|I|R|R|R|               Reserved                        |
16  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
17  * |                VXLAN Network Identifier (VNI) |   Reserved    |
18  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
19  *
20  * I = VXLAN Network Identifier (VNI) present.
21  */
22 struct vxlanhdr {
23 	__be32 vx_flags;
24 	__be32 vx_vni;
25 };
26 
27 /* VXLAN header flags. */
28 #define VXLAN_HF_VNI	cpu_to_be32(BIT(27))
29 
30 #define VXLAN_N_VID     (1u << 24)
31 #define VXLAN_VID_MASK  (VXLAN_N_VID - 1)
32 #define VXLAN_VNI_MASK	cpu_to_be32(VXLAN_VID_MASK << 8)
33 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
34 
35 #define VNI_HASH_BITS	10
36 #define VNI_HASH_SIZE	(1<<VNI_HASH_BITS)
37 #define FDB_HASH_BITS	8
38 #define FDB_HASH_SIZE	(1<<FDB_HASH_BITS)
39 
40 /* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
41  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
42  * |R|R|R|R|I|R|R|R|R|R|C|              Reserved                   |
43  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
44  * |           VXLAN Network Identifier (VNI)      |O| Csum start  |
45  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
46  *
47  * C = Remote checksum offload bit. When set indicates that the
48  *     remote checksum offload data is present.
49  *
50  * O = Offset bit. Indicates the checksum offset relative to
51  *     checksum start.
52  *
53  * Csum start = Checksum start divided by two.
54  *
55  * http://tools.ietf.org/html/draft-herbert-vxlan-rco
56  */
57 
58 /* VXLAN-RCO header flags. */
59 #define VXLAN_HF_RCO	cpu_to_be32(BIT(21))
60 
61 /* Remote checksum offload header option */
62 #define VXLAN_RCO_MASK	cpu_to_be32(0x7f)  /* Last byte of vni field */
63 #define VXLAN_RCO_UDP	cpu_to_be32(0x80)  /* Indicate UDP RCO (TCP when not set *) */
64 #define VXLAN_RCO_SHIFT	1		   /* Left shift of start */
65 #define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
66 #define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)
67 
68 /*
69  * VXLAN Group Based Policy Extension (VXLAN_F_GBP):
70  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
71  * |G|R|R|R|I|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
72  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
73  * |                VXLAN Network Identifier (VNI) |   Reserved    |
74  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
75  *
76  * G = Group Policy ID present.
77  *
78  * D = Don't Learn bit. When set, this bit indicates that the egress
79  *     VTEP MUST NOT learn the source address of the encapsulated frame.
80  *
81  * A = Indicates that the group policy has already been applied to
82  *     this packet. Policies MUST NOT be applied by devices when the
83  *     A bit is set.
84  *
85  * https://tools.ietf.org/html/draft-smith-vxlan-group-policy
86  */
87 struct vxlanhdr_gbp {
88 	u8	vx_flags;
89 #ifdef __LITTLE_ENDIAN_BITFIELD
90 	u8	reserved_flags1:3,
91 		policy_applied:1,
92 		reserved_flags2:2,
93 		dont_learn:1,
94 		reserved_flags3:1;
95 #elif defined(__BIG_ENDIAN_BITFIELD)
96 	u8	reserved_flags1:1,
97 		dont_learn:1,
98 		reserved_flags2:2,
99 		policy_applied:1,
100 		reserved_flags3:3;
101 #else
102 #error	"Please fix <asm/byteorder.h>"
103 #endif
104 	__be16	policy_id;
105 	__be32	vx_vni;
106 };
107 
108 /* VXLAN-GBP header flags. */
109 #define VXLAN_HF_GBP	cpu_to_be32(BIT(31))
110 
111 #define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP | cpu_to_be32(0xFFFFFF))
112 
113 /* skb->mark mapping
114  *
115  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
116  * |R|R|R|R|R|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
117  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
118  */
119 #define VXLAN_GBP_DONT_LEARN		(BIT(6) << 16)
120 #define VXLAN_GBP_POLICY_APPLIED	(BIT(3) << 16)
121 #define VXLAN_GBP_ID_MASK		(0xFFFF)
122 
123 /*
124  * VXLAN Generic Protocol Extension (VXLAN_F_GPE):
125  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
126  * |R|R|Ver|I|P|R|O|       Reserved                |Next Protocol  |
127  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
128  * |                VXLAN Network Identifier (VNI) |   Reserved    |
129  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
130  *
131  * Ver = Version. Indicates VXLAN GPE protocol version.
132  *
133  * P = Next Protocol Bit. The P bit is set to indicate that the
134  *     Next Protocol field is present.
135  *
136  * O = OAM Flag Bit. The O bit is set to indicate that the packet
137  *     is an OAM packet.
138  *
139  * Next Protocol = This 8 bit field indicates the protocol header
140  * immediately following the VXLAN GPE header.
141  *
142  * https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
143  */
144 
145 struct vxlanhdr_gpe {
146 #if defined(__LITTLE_ENDIAN_BITFIELD)
147 	u8	oam_flag:1,
148 		reserved_flags1:1,
149 		np_applied:1,
150 		instance_applied:1,
151 		version:2,
152 		reserved_flags2:2;
153 #elif defined(__BIG_ENDIAN_BITFIELD)
154 	u8	reserved_flags2:2,
155 		version:2,
156 		instance_applied:1,
157 		np_applied:1,
158 		reserved_flags1:1,
159 		oam_flag:1;
160 #endif
161 	u8	reserved_flags3;
162 	u8	reserved_flags4;
163 	u8	next_protocol;
164 	__be32	vx_vni;
165 };
166 
167 /* VXLAN-GPE header flags. */
168 #define VXLAN_HF_VER	cpu_to_be32(BIT(29) | BIT(28))
169 #define VXLAN_HF_NP	cpu_to_be32(BIT(26))
170 #define VXLAN_HF_OAM	cpu_to_be32(BIT(24))
171 
172 #define VXLAN_GPE_USED_BITS (VXLAN_HF_VER | VXLAN_HF_NP | VXLAN_HF_OAM | \
173 			     cpu_to_be32(0xff))
174 
175 struct vxlan_metadata {
176 	u32		gbp;
177 };
178 
179 /* per UDP socket information */
180 struct vxlan_sock {
181 	struct hlist_node hlist;
182 	struct socket	 *sock;
183 	struct hlist_head vni_list[VNI_HASH_SIZE];
184 	refcount_t	  refcnt;
185 	u32		  flags;
186 };
187 
188 union vxlan_addr {
189 	struct sockaddr_in sin;
190 	struct sockaddr_in6 sin6;
191 	struct sockaddr sa;
192 };
193 
194 struct vxlan_rdst {
195 	union vxlan_addr	 remote_ip;
196 	__be16			 remote_port;
197 	u8			 offloaded:1;
198 	__be32			 remote_vni;
199 	u32			 remote_ifindex;
200 	struct list_head	 list;
201 	struct rcu_head		 rcu;
202 	struct dst_cache	 dst_cache;
203 };
204 
205 struct vxlan_config {
206 	union vxlan_addr	remote_ip;
207 	union vxlan_addr	saddr;
208 	__be32			vni;
209 	int			remote_ifindex;
210 	int			mtu;
211 	__be16			dst_port;
212 	u16			port_min;
213 	u16			port_max;
214 	u8			tos;
215 	u8			ttl;
216 	__be32			label;
217 	u32			flags;
218 	unsigned long		age_interval;
219 	unsigned int		addrmax;
220 	bool			no_share;
221 	enum ifla_vxlan_df	df;
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 #define VXLAN_F_TTL_INHERIT		0x10000
271 
272 /* Flags that are used in the receive path. These flags must match in
273  * order for a socket to be shareable
274  */
275 #define VXLAN_F_RCV_FLAGS		(VXLAN_F_GBP |			\
276 					 VXLAN_F_GPE |			\
277 					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
278 					 VXLAN_F_REMCSUM_RX |		\
279 					 VXLAN_F_REMCSUM_NOPARTIAL |	\
280 					 VXLAN_F_COLLECT_METADATA)
281 
282 /* Flags that can be set together with VXLAN_F_GPE. */
283 #define VXLAN_F_ALLOWED_GPE		(VXLAN_F_GPE |			\
284 					 VXLAN_F_IPV6 |			\
285 					 VXLAN_F_IPV6_LINKLOCAL |	\
286 					 VXLAN_F_UDP_ZERO_CSUM_TX |	\
287 					 VXLAN_F_UDP_ZERO_CSUM6_TX |	\
288 					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
289 					 VXLAN_F_COLLECT_METADATA)
290 
291 struct net_device *vxlan_dev_create(struct net *net, const char *name,
292 				    u8 name_assign_type, struct vxlan_config *conf);
293 
294 static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
295 						     netdev_features_t features)
296 {
297 	u8 l4_hdr = 0;
298 
299 	if (!skb->encapsulation)
300 		return features;
301 
302 	switch (vlan_get_protocol(skb)) {
303 	case htons(ETH_P_IP):
304 		l4_hdr = ip_hdr(skb)->protocol;
305 		break;
306 	case htons(ETH_P_IPV6):
307 		l4_hdr = ipv6_hdr(skb)->nexthdr;
308 		break;
309 	default:
310 		return features;
311 	}
312 
313 	if ((l4_hdr == IPPROTO_UDP) &&
314 	    (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
315 	     skb->inner_protocol != htons(ETH_P_TEB) ||
316 	     (skb_inner_mac_header(skb) - skb_transport_header(skb) !=
317 	      sizeof(struct udphdr) + sizeof(struct vxlanhdr)) ||
318 	     (skb->ip_summed != CHECKSUM_NONE &&
319 	      !can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
320 		return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
321 
322 	return features;
323 }
324 
325 /* IP header + UDP + VXLAN + Ethernet header */
326 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
327 /* IPv6 header + UDP + VXLAN + Ethernet header */
328 #define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
329 
330 static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
331 {
332 	return (struct vxlanhdr *)(udp_hdr(skb) + 1);
333 }
334 
335 static inline __be32 vxlan_vni(__be32 vni_field)
336 {
337 #if defined(__BIG_ENDIAN)
338 	return (__force __be32)((__force u32)vni_field >> 8);
339 #else
340 	return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
341 #endif
342 }
343 
344 static inline __be32 vxlan_vni_field(__be32 vni)
345 {
346 #if defined(__BIG_ENDIAN)
347 	return (__force __be32)((__force u32)vni << 8);
348 #else
349 	return (__force __be32)((__force u32)vni >> 8);
350 #endif
351 }
352 
353 static inline size_t vxlan_rco_start(__be32 vni_field)
354 {
355 	return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
356 }
357 
358 static inline size_t vxlan_rco_offset(__be32 vni_field)
359 {
360 	return (vni_field & VXLAN_RCO_UDP) ?
361 		offsetof(struct udphdr, check) :
362 		offsetof(struct tcphdr, check);
363 }
364 
365 static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
366 {
367 	__be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
368 
369 	if (offset == offsetof(struct udphdr, check))
370 		vni_field |= VXLAN_RCO_UDP;
371 	return vni_field;
372 }
373 
374 static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
375 {
376 	return vs->sock->sk->sk_family;
377 }
378 
379 #if IS_ENABLED(CONFIG_IPV6)
380 
381 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
382 {
383 	if (ipa->sa.sa_family == AF_INET6)
384 		return ipv6_addr_any(&ipa->sin6.sin6_addr);
385 	else
386 		return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
387 }
388 
389 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
390 {
391 	if (ipa->sa.sa_family == AF_INET6)
392 		return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
393 	else
394 		return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
395 }
396 
397 #else /* !IS_ENABLED(CONFIG_IPV6) */
398 
399 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
400 {
401 	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
402 }
403 
404 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
405 {
406 	return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
407 }
408 
409 #endif /* IS_ENABLED(CONFIG_IPV6) */
410 
411 static inline bool netif_is_vxlan(const struct net_device *dev)
412 {
413 	return dev->rtnl_link_ops &&
414 	       !strcmp(dev->rtnl_link_ops->kind, "vxlan");
415 }
416 
417 struct switchdev_notifier_vxlan_fdb_info {
418 	struct switchdev_notifier_info info; /* must be first */
419 	union vxlan_addr remote_ip;
420 	__be16 remote_port;
421 	__be32 remote_vni;
422 	u32 remote_ifindex;
423 	u8 eth_addr[ETH_ALEN];
424 	__be32 vni;
425 	bool offloaded;
426 	bool added_by_user;
427 };
428 
429 #if IS_ENABLED(CONFIG_VXLAN)
430 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
431 		      struct switchdev_notifier_vxlan_fdb_info *fdb_info);
432 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
433 		     struct notifier_block *nb,
434 		     struct netlink_ext_ack *extack);
435 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);
436 
437 #else
438 static inline int
439 vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
440 		  struct switchdev_notifier_vxlan_fdb_info *fdb_info)
441 {
442 	return -ENOENT;
443 }
444 
445 static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
446 				   struct notifier_block *nb,
447 				   struct netlink_ext_ack *extack)
448 {
449 	return -EOPNOTSUPP;
450 }
451 
452 static inline void
453 vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
454 {
455 }
456 #endif
457 
458 static inline void vxlan_flag_attr_error(int attrtype,
459 					 struct netlink_ext_ack *extack)
460 {
461 #define VXLAN_FLAG(flg) \
462 	case IFLA_VXLAN_##flg: \
463 		NL_SET_ERR_MSG_MOD(extack, \
464 				   "cannot change " #flg " flag"); \
465 		break
466 	switch (attrtype) {
467 	VXLAN_FLAG(TTL_INHERIT);
468 	VXLAN_FLAG(LEARNING);
469 	VXLAN_FLAG(PROXY);
470 	VXLAN_FLAG(RSC);
471 	VXLAN_FLAG(L2MISS);
472 	VXLAN_FLAG(L3MISS);
473 	VXLAN_FLAG(COLLECT_METADATA);
474 	VXLAN_FLAG(UDP_ZERO_CSUM6_TX);
475 	VXLAN_FLAG(UDP_ZERO_CSUM6_RX);
476 	VXLAN_FLAG(REMCSUM_TX);
477 	VXLAN_FLAG(REMCSUM_RX);
478 	VXLAN_FLAG(GBP);
479 	VXLAN_FLAG(GPE);
480 	VXLAN_FLAG(REMCSUM_NOPARTIAL);
481 	default:
482 		NL_SET_ERR_MSG_MOD(extack, \
483 				   "cannot change flag");
484 		break;
485 	}
486 #undef VXLAN_FLAG
487 }
488 
489 #endif
490