xref: /linux/include/net/vxlan.h (revision 18f90d372cf35b387663f1567de701e5393f6eb5)
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 /* VXLAN protocol (RFC 7348) header:
12  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
13  * |R|R|R|R|I|R|R|R|               Reserved                        |
14  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
15  * |                VXLAN Network Identifier (VNI) |   Reserved    |
16  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
17  *
18  * I = VXLAN Network Identifier (VNI) present.
19  */
20 struct vxlanhdr {
21 	__be32 vx_flags;
22 	__be32 vx_vni;
23 };
24 
25 /* VXLAN header flags. */
26 #define VXLAN_HF_VNI	cpu_to_be32(BIT(27))
27 
28 #define VXLAN_N_VID     (1u << 24)
29 #define VXLAN_VID_MASK  (VXLAN_N_VID - 1)
30 #define VXLAN_VNI_MASK	cpu_to_be32(VXLAN_VID_MASK << 8)
31 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
32 
33 #define VNI_HASH_BITS	10
34 #define VNI_HASH_SIZE	(1<<VNI_HASH_BITS)
35 #define FDB_HASH_BITS	8
36 #define FDB_HASH_SIZE	(1<<FDB_HASH_BITS)
37 
38 /* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
39  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
40  * |R|R|R|R|I|R|R|R|R|R|C|              Reserved                   |
41  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
42  * |           VXLAN Network Identifier (VNI)      |O| Csum start  |
43  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
44  *
45  * C = Remote checksum offload bit. When set indicates that the
46  *     remote checksum offload data is present.
47  *
48  * O = Offset bit. Indicates the checksum offset relative to
49  *     checksum start.
50  *
51  * Csum start = Checksum start divided by two.
52  *
53  * http://tools.ietf.org/html/draft-herbert-vxlan-rco
54  */
55 
56 /* VXLAN-RCO header flags. */
57 #define VXLAN_HF_RCO	cpu_to_be32(BIT(21))
58 
59 /* Remote checksum offload header option */
60 #define VXLAN_RCO_MASK	cpu_to_be32(0x7f)  /* Last byte of vni field */
61 #define VXLAN_RCO_UDP	cpu_to_be32(0x80)  /* Indicate UDP RCO (TCP when not set *) */
62 #define VXLAN_RCO_SHIFT	1		   /* Left shift of start */
63 #define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
64 #define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)
65 
66 /*
67  * VXLAN Group Based Policy Extension (VXLAN_F_GBP):
68  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
69  * |G|R|R|R|I|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
70  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
71  * |                VXLAN Network Identifier (VNI) |   Reserved    |
72  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
73  *
74  * G = Group Policy ID present.
75  *
76  * D = Don't Learn bit. When set, this bit indicates that the egress
77  *     VTEP MUST NOT learn the source address of the encapsulated frame.
78  *
79  * A = Indicates that the group policy has already been applied to
80  *     this packet. Policies MUST NOT be applied by devices when the
81  *     A bit is set.
82  *
83  * https://tools.ietf.org/html/draft-smith-vxlan-group-policy
84  */
85 struct vxlanhdr_gbp {
86 	u8	vx_flags;
87 #ifdef __LITTLE_ENDIAN_BITFIELD
88 	u8	reserved_flags1:3,
89 		policy_applied:1,
90 		reserved_flags2:2,
91 		dont_learn:1,
92 		reserved_flags3:1;
93 #elif defined(__BIG_ENDIAN_BITFIELD)
94 	u8	reserved_flags1:1,
95 		dont_learn:1,
96 		reserved_flags2:2,
97 		policy_applied:1,
98 		reserved_flags3:3;
99 #else
100 #error	"Please fix <asm/byteorder.h>"
101 #endif
102 	__be16	policy_id;
103 	__be32	vx_vni;
104 };
105 
106 /* VXLAN-GBP header flags. */
107 #define VXLAN_HF_GBP	cpu_to_be32(BIT(31))
108 
109 #define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP | cpu_to_be32(0xFFFFFF))
110 
111 /* skb->mark mapping
112  *
113  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
114  * |R|R|R|R|R|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
115  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
116  */
117 #define VXLAN_GBP_DONT_LEARN		(BIT(6) << 16)
118 #define VXLAN_GBP_POLICY_APPLIED	(BIT(3) << 16)
119 #define VXLAN_GBP_ID_MASK		(0xFFFF)
120 
121 /*
122  * VXLAN Generic Protocol Extension (VXLAN_F_GPE):
123  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
124  * |R|R|Ver|I|P|R|O|       Reserved                |Next Protocol  |
125  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
126  * |                VXLAN Network Identifier (VNI) |   Reserved    |
127  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
128  *
129  * Ver = Version. Indicates VXLAN GPE protocol version.
130  *
131  * P = Next Protocol Bit. The P bit is set to indicate that the
132  *     Next Protocol field is present.
133  *
134  * O = OAM Flag Bit. The O bit is set to indicate that the packet
135  *     is an OAM packet.
136  *
137  * Next Protocol = This 8 bit field indicates the protocol header
138  * immediately following the VXLAN GPE header.
139  *
140  * https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
141  */
142 
143 struct vxlanhdr_gpe {
144 #if defined(__LITTLE_ENDIAN_BITFIELD)
145 	u8	oam_flag:1,
146 		reserved_flags1:1,
147 		np_applied:1,
148 		instance_applied:1,
149 		version:2,
150 		reserved_flags2:2;
151 #elif defined(__BIG_ENDIAN_BITFIELD)
152 	u8	reserved_flags2:2,
153 		version:2,
154 		instance_applied:1,
155 		np_applied:1,
156 		reserved_flags1:1,
157 		oam_flag:1;
158 #endif
159 	u8	reserved_flags3;
160 	u8	reserved_flags4;
161 	u8	next_protocol;
162 	__be32	vx_vni;
163 };
164 
165 /* VXLAN-GPE header flags. */
166 #define VXLAN_HF_VER	cpu_to_be32(BIT(29) | BIT(28))
167 #define VXLAN_HF_NP	cpu_to_be32(BIT(26))
168 #define VXLAN_HF_OAM	cpu_to_be32(BIT(24))
169 
170 #define VXLAN_GPE_USED_BITS (VXLAN_HF_VER | VXLAN_HF_NP | VXLAN_HF_OAM | \
171 			     cpu_to_be32(0xff))
172 
173 struct vxlan_metadata {
174 	u32		gbp;
175 };
176 
177 /* per UDP socket information */
178 struct vxlan_sock {
179 	struct hlist_node hlist;
180 	struct socket	 *sock;
181 	struct hlist_head vni_list[VNI_HASH_SIZE];
182 	refcount_t	  refcnt;
183 	u32		  flags;
184 };
185 
186 union vxlan_addr {
187 	struct sockaddr_in sin;
188 	struct sockaddr_in6 sin6;
189 	struct sockaddr sa;
190 };
191 
192 struct vxlan_rdst {
193 	union vxlan_addr	 remote_ip;
194 	__be16			 remote_port;
195 	u8			 offloaded:1;
196 	__be32			 remote_vni;
197 	u32			 remote_ifindex;
198 	struct list_head	 list;
199 	struct rcu_head		 rcu;
200 	struct dst_cache	 dst_cache;
201 };
202 
203 struct vxlan_config {
204 	union vxlan_addr	remote_ip;
205 	union vxlan_addr	saddr;
206 	__be32			vni;
207 	int			remote_ifindex;
208 	int			mtu;
209 	__be16			dst_port;
210 	u16			port_min;
211 	u16			port_max;
212 	u8			tos;
213 	u8			ttl;
214 	__be32			label;
215 	u32			flags;
216 	unsigned long		age_interval;
217 	unsigned int		addrmax;
218 	bool			no_share;
219 	enum ifla_vxlan_df	df;
220 };
221 
222 struct vxlan_dev_node {
223 	struct hlist_node hlist;
224 	struct vxlan_dev *vxlan;
225 };
226 
227 /* Pseudo network device */
228 struct vxlan_dev {
229 	struct vxlan_dev_node hlist4;	/* vni hash table for IPv4 socket */
230 #if IS_ENABLED(CONFIG_IPV6)
231 	struct vxlan_dev_node hlist6;	/* vni hash table for IPv6 socket */
232 #endif
233 	struct list_head  next;		/* vxlan's per namespace list */
234 	struct vxlan_sock __rcu *vn4_sock;	/* listening socket for IPv4 */
235 #if IS_ENABLED(CONFIG_IPV6)
236 	struct vxlan_sock __rcu *vn6_sock;	/* listening socket for IPv6 */
237 #endif
238 	struct net_device *dev;
239 	struct net	  *net;		/* netns for packet i/o */
240 	struct vxlan_rdst default_dst;	/* default destination */
241 
242 	struct timer_list age_timer;
243 	spinlock_t	  hash_lock;
244 	unsigned int	  addrcnt;
245 	struct gro_cells  gro_cells;
246 
247 	struct vxlan_config	cfg;
248 
249 	struct hlist_head fdb_head[FDB_HASH_SIZE];
250 };
251 
252 #define VXLAN_F_LEARN			0x01
253 #define VXLAN_F_PROXY			0x02
254 #define VXLAN_F_RSC			0x04
255 #define VXLAN_F_L2MISS			0x08
256 #define VXLAN_F_L3MISS			0x10
257 #define VXLAN_F_IPV6			0x20
258 #define VXLAN_F_UDP_ZERO_CSUM_TX	0x40
259 #define VXLAN_F_UDP_ZERO_CSUM6_TX	0x80
260 #define VXLAN_F_UDP_ZERO_CSUM6_RX	0x100
261 #define VXLAN_F_REMCSUM_TX		0x200
262 #define VXLAN_F_REMCSUM_RX		0x400
263 #define VXLAN_F_GBP			0x800
264 #define VXLAN_F_REMCSUM_NOPARTIAL	0x1000
265 #define VXLAN_F_COLLECT_METADATA	0x2000
266 #define VXLAN_F_GPE			0x4000
267 #define VXLAN_F_IPV6_LINKLOCAL		0x8000
268 #define VXLAN_F_TTL_INHERIT		0x10000
269 
270 /* Flags that are used in the receive path. These flags must match in
271  * order for a socket to be shareable
272  */
273 #define VXLAN_F_RCV_FLAGS		(VXLAN_F_GBP |			\
274 					 VXLAN_F_GPE |			\
275 					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
276 					 VXLAN_F_REMCSUM_RX |		\
277 					 VXLAN_F_REMCSUM_NOPARTIAL |	\
278 					 VXLAN_F_COLLECT_METADATA)
279 
280 /* Flags that can be set together with VXLAN_F_GPE. */
281 #define VXLAN_F_ALLOWED_GPE		(VXLAN_F_GPE |			\
282 					 VXLAN_F_IPV6 |			\
283 					 VXLAN_F_IPV6_LINKLOCAL |	\
284 					 VXLAN_F_UDP_ZERO_CSUM_TX |	\
285 					 VXLAN_F_UDP_ZERO_CSUM6_TX |	\
286 					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
287 					 VXLAN_F_COLLECT_METADATA)
288 
289 struct net_device *vxlan_dev_create(struct net *net, const char *name,
290 				    u8 name_assign_type, struct vxlan_config *conf);
291 
292 static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
293 						     netdev_features_t features)
294 {
295 	u8 l4_hdr = 0;
296 
297 	if (!skb->encapsulation)
298 		return features;
299 
300 	switch (vlan_get_protocol(skb)) {
301 	case htons(ETH_P_IP):
302 		l4_hdr = ip_hdr(skb)->protocol;
303 		break;
304 	case htons(ETH_P_IPV6):
305 		l4_hdr = ipv6_hdr(skb)->nexthdr;
306 		break;
307 	default:
308 		return features;
309 	}
310 
311 	if ((l4_hdr == IPPROTO_UDP) &&
312 	    (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
313 	     skb->inner_protocol != htons(ETH_P_TEB) ||
314 	     (skb_inner_mac_header(skb) - skb_transport_header(skb) !=
315 	      sizeof(struct udphdr) + sizeof(struct vxlanhdr)) ||
316 	     (skb->ip_summed != CHECKSUM_NONE &&
317 	      !can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
318 		return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
319 
320 	return features;
321 }
322 
323 /* IP header + UDP + VXLAN + Ethernet header */
324 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
325 /* IPv6 header + UDP + VXLAN + Ethernet header */
326 #define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
327 
328 static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
329 {
330 	return (struct vxlanhdr *)(udp_hdr(skb) + 1);
331 }
332 
333 static inline __be32 vxlan_vni(__be32 vni_field)
334 {
335 #if defined(__BIG_ENDIAN)
336 	return (__force __be32)((__force u32)vni_field >> 8);
337 #else
338 	return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
339 #endif
340 }
341 
342 static inline __be32 vxlan_vni_field(__be32 vni)
343 {
344 #if defined(__BIG_ENDIAN)
345 	return (__force __be32)((__force u32)vni << 8);
346 #else
347 	return (__force __be32)((__force u32)vni >> 8);
348 #endif
349 }
350 
351 static inline size_t vxlan_rco_start(__be32 vni_field)
352 {
353 	return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
354 }
355 
356 static inline size_t vxlan_rco_offset(__be32 vni_field)
357 {
358 	return (vni_field & VXLAN_RCO_UDP) ?
359 		offsetof(struct udphdr, check) :
360 		offsetof(struct tcphdr, check);
361 }
362 
363 static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
364 {
365 	__be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
366 
367 	if (offset == offsetof(struct udphdr, check))
368 		vni_field |= VXLAN_RCO_UDP;
369 	return vni_field;
370 }
371 
372 static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
373 {
374 	return vs->sock->sk->sk_family;
375 }
376 
377 #if IS_ENABLED(CONFIG_IPV6)
378 
379 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
380 {
381 	if (ipa->sa.sa_family == AF_INET6)
382 		return ipv6_addr_any(&ipa->sin6.sin6_addr);
383 	else
384 		return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
385 }
386 
387 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
388 {
389 	if (ipa->sa.sa_family == AF_INET6)
390 		return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
391 	else
392 		return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
393 }
394 
395 #else /* !IS_ENABLED(CONFIG_IPV6) */
396 
397 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
398 {
399 	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
400 }
401 
402 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
403 {
404 	return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
405 }
406 
407 #endif /* IS_ENABLED(CONFIG_IPV6) */
408 
409 static inline bool netif_is_vxlan(const struct net_device *dev)
410 {
411 	return dev->rtnl_link_ops &&
412 	       !strcmp(dev->rtnl_link_ops->kind, "vxlan");
413 }
414 
415 struct switchdev_notifier_vxlan_fdb_info {
416 	struct switchdev_notifier_info info; /* must be first */
417 	union vxlan_addr remote_ip;
418 	__be16 remote_port;
419 	__be32 remote_vni;
420 	u32 remote_ifindex;
421 	u8 eth_addr[ETH_ALEN];
422 	__be32 vni;
423 	bool offloaded;
424 	bool added_by_user;
425 };
426 
427 #if IS_ENABLED(CONFIG_VXLAN)
428 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
429 		      struct switchdev_notifier_vxlan_fdb_info *fdb_info);
430 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
431 		     struct notifier_block *nb,
432 		     struct netlink_ext_ack *extack);
433 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);
434 
435 #else
436 static inline int
437 vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
438 		  struct switchdev_notifier_vxlan_fdb_info *fdb_info)
439 {
440 	return -ENOENT;
441 }
442 
443 static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
444 				   struct notifier_block *nb,
445 				   struct netlink_ext_ack *extack)
446 {
447 	return -EOPNOTSUPP;
448 }
449 
450 static inline void
451 vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
452 {
453 }
454 #endif
455 
456 static inline void vxlan_flag_attr_error(int attrtype,
457 					 struct netlink_ext_ack *extack)
458 {
459 #define VXLAN_FLAG(flg) \
460 	case IFLA_VXLAN_##flg: \
461 		NL_SET_ERR_MSG_MOD(extack, \
462 				   "cannot change " #flg " flag"); \
463 		break
464 	switch (attrtype) {
465 	VXLAN_FLAG(TTL_INHERIT);
466 	VXLAN_FLAG(LEARNING);
467 	VXLAN_FLAG(PROXY);
468 	VXLAN_FLAG(RSC);
469 	VXLAN_FLAG(L2MISS);
470 	VXLAN_FLAG(L3MISS);
471 	VXLAN_FLAG(COLLECT_METADATA);
472 	VXLAN_FLAG(UDP_ZERO_CSUM6_TX);
473 	VXLAN_FLAG(UDP_ZERO_CSUM6_RX);
474 	VXLAN_FLAG(REMCSUM_TX);
475 	VXLAN_FLAG(REMCSUM_RX);
476 	VXLAN_FLAG(GBP);
477 	VXLAN_FLAG(GPE);
478 	VXLAN_FLAG(REMCSUM_NOPARTIAL);
479 	default:
480 		NL_SET_ERR_MSG_MOD(extack, \
481 				   "cannot change flag");
482 		break;
483 	}
484 #undef VXLAN_FLAG
485 }
486 
487 #endif
488