xref: /openbmc/linux/include/net/vxlan.h (revision 95298d63)
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 #include <net/nexthop.h>
11 
12 #define IANA_VXLAN_UDP_PORT     4789
13 
14 /* VXLAN protocol (RFC 7348) header:
15  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
16  * |R|R|R|R|I|R|R|R|               Reserved                        |
17  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
18  * |                VXLAN Network Identifier (VNI) |   Reserved    |
19  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
20  *
21  * I = VXLAN Network Identifier (VNI) present.
22  */
23 struct vxlanhdr {
24 	__be32 vx_flags;
25 	__be32 vx_vni;
26 };
27 
28 /* VXLAN header flags. */
29 #define VXLAN_HF_VNI	cpu_to_be32(BIT(27))
30 
31 #define VXLAN_N_VID     (1u << 24)
32 #define VXLAN_VID_MASK  (VXLAN_N_VID - 1)
33 #define VXLAN_VNI_MASK	cpu_to_be32(VXLAN_VID_MASK << 8)
34 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
35 
36 #define VNI_HASH_BITS	10
37 #define VNI_HASH_SIZE	(1<<VNI_HASH_BITS)
38 #define FDB_HASH_BITS	8
39 #define FDB_HASH_SIZE	(1<<FDB_HASH_BITS)
40 
41 /* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
42  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
43  * |R|R|R|R|I|R|R|R|R|R|C|              Reserved                   |
44  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
45  * |           VXLAN Network Identifier (VNI)      |O| Csum start  |
46  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
47  *
48  * C = Remote checksum offload bit. When set indicates that the
49  *     remote checksum offload data is present.
50  *
51  * O = Offset bit. Indicates the checksum offset relative to
52  *     checksum start.
53  *
54  * Csum start = Checksum start divided by two.
55  *
56  * http://tools.ietf.org/html/draft-herbert-vxlan-rco
57  */
58 
59 /* VXLAN-RCO header flags. */
60 #define VXLAN_HF_RCO	cpu_to_be32(BIT(21))
61 
62 /* Remote checksum offload header option */
63 #define VXLAN_RCO_MASK	cpu_to_be32(0x7f)  /* Last byte of vni field */
64 #define VXLAN_RCO_UDP	cpu_to_be32(0x80)  /* Indicate UDP RCO (TCP when not set *) */
65 #define VXLAN_RCO_SHIFT	1		   /* Left shift of start */
66 #define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
67 #define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)
68 
69 /*
70  * VXLAN Group Based Policy Extension (VXLAN_F_GBP):
71  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
72  * |G|R|R|R|I|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
73  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
74  * |                VXLAN Network Identifier (VNI) |   Reserved    |
75  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
76  *
77  * G = Group Policy ID present.
78  *
79  * D = Don't Learn bit. When set, this bit indicates that the egress
80  *     VTEP MUST NOT learn the source address of the encapsulated frame.
81  *
82  * A = Indicates that the group policy has already been applied to
83  *     this packet. Policies MUST NOT be applied by devices when the
84  *     A bit is set.
85  *
86  * https://tools.ietf.org/html/draft-smith-vxlan-group-policy
87  */
88 struct vxlanhdr_gbp {
89 	u8	vx_flags;
90 #ifdef __LITTLE_ENDIAN_BITFIELD
91 	u8	reserved_flags1:3,
92 		policy_applied:1,
93 		reserved_flags2:2,
94 		dont_learn:1,
95 		reserved_flags3:1;
96 #elif defined(__BIG_ENDIAN_BITFIELD)
97 	u8	reserved_flags1:1,
98 		dont_learn:1,
99 		reserved_flags2:2,
100 		policy_applied:1,
101 		reserved_flags3:3;
102 #else
103 #error	"Please fix <asm/byteorder.h>"
104 #endif
105 	__be16	policy_id;
106 	__be32	vx_vni;
107 };
108 
109 /* VXLAN-GBP header flags. */
110 #define VXLAN_HF_GBP	cpu_to_be32(BIT(31))
111 
112 #define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP | cpu_to_be32(0xFFFFFF))
113 
114 /* skb->mark mapping
115  *
116  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
117  * |R|R|R|R|R|R|R|R|R|D|R|R|A|R|R|R|        Group Policy ID        |
118  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119  */
120 #define VXLAN_GBP_DONT_LEARN		(BIT(6) << 16)
121 #define VXLAN_GBP_POLICY_APPLIED	(BIT(3) << 16)
122 #define VXLAN_GBP_ID_MASK		(0xFFFF)
123 
124 /*
125  * VXLAN Generic Protocol Extension (VXLAN_F_GPE):
126  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
127  * |R|R|Ver|I|P|R|O|       Reserved                |Next Protocol  |
128  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
129  * |                VXLAN Network Identifier (VNI) |   Reserved    |
130  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
131  *
132  * Ver = Version. Indicates VXLAN GPE protocol version.
133  *
134  * P = Next Protocol Bit. The P bit is set to indicate that the
135  *     Next Protocol field is present.
136  *
137  * O = OAM Flag Bit. The O bit is set to indicate that the packet
138  *     is an OAM packet.
139  *
140  * Next Protocol = This 8 bit field indicates the protocol header
141  * immediately following the VXLAN GPE header.
142  *
143  * https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
144  */
145 
146 struct vxlanhdr_gpe {
147 #if defined(__LITTLE_ENDIAN_BITFIELD)
148 	u8	oam_flag:1,
149 		reserved_flags1:1,
150 		np_applied:1,
151 		instance_applied:1,
152 		version:2,
153 		reserved_flags2:2;
154 #elif defined(__BIG_ENDIAN_BITFIELD)
155 	u8	reserved_flags2:2,
156 		version:2,
157 		instance_applied:1,
158 		np_applied:1,
159 		reserved_flags1:1,
160 		oam_flag:1;
161 #endif
162 	u8	reserved_flags3;
163 	u8	reserved_flags4;
164 	u8	next_protocol;
165 	__be32	vx_vni;
166 };
167 
168 /* VXLAN-GPE header flags. */
169 #define VXLAN_HF_VER	cpu_to_be32(BIT(29) | BIT(28))
170 #define VXLAN_HF_NP	cpu_to_be32(BIT(26))
171 #define VXLAN_HF_OAM	cpu_to_be32(BIT(24))
172 
173 #define VXLAN_GPE_USED_BITS (VXLAN_HF_VER | VXLAN_HF_NP | VXLAN_HF_OAM | \
174 			     cpu_to_be32(0xff))
175 
176 struct vxlan_metadata {
177 	u32		gbp;
178 };
179 
180 /* per UDP socket information */
181 struct vxlan_sock {
182 	struct hlist_node hlist;
183 	struct socket	 *sock;
184 	struct hlist_head vni_list[VNI_HASH_SIZE];
185 	refcount_t	  refcnt;
186 	u32		  flags;
187 };
188 
189 union vxlan_addr {
190 	struct sockaddr_in sin;
191 	struct sockaddr_in6 sin6;
192 	struct sockaddr sa;
193 };
194 
195 struct vxlan_rdst {
196 	union vxlan_addr	 remote_ip;
197 	__be16			 remote_port;
198 	u8			 offloaded:1;
199 	__be32			 remote_vni;
200 	u32			 remote_ifindex;
201 	struct net_device	 *remote_dev;
202 	struct list_head	 list;
203 	struct rcu_head		 rcu;
204 	struct dst_cache	 dst_cache;
205 };
206 
207 struct vxlan_config {
208 	union vxlan_addr	remote_ip;
209 	union vxlan_addr	saddr;
210 	__be32			vni;
211 	int			remote_ifindex;
212 	int			mtu;
213 	__be16			dst_port;
214 	u16			port_min;
215 	u16			port_max;
216 	u8			tos;
217 	u8			ttl;
218 	__be32			label;
219 	u32			flags;
220 	unsigned long		age_interval;
221 	unsigned int		addrmax;
222 	bool			no_share;
223 	enum ifla_vxlan_df	df;
224 };
225 
226 struct vxlan_dev_node {
227 	struct hlist_node hlist;
228 	struct vxlan_dev *vxlan;
229 };
230 
231 /* Pseudo network device */
232 struct vxlan_dev {
233 	struct vxlan_dev_node hlist4;	/* vni hash table for IPv4 socket */
234 #if IS_ENABLED(CONFIG_IPV6)
235 	struct vxlan_dev_node hlist6;	/* vni hash table for IPv6 socket */
236 #endif
237 	struct list_head  next;		/* vxlan's per namespace list */
238 	struct vxlan_sock __rcu *vn4_sock;	/* listening socket for IPv4 */
239 #if IS_ENABLED(CONFIG_IPV6)
240 	struct vxlan_sock __rcu *vn6_sock;	/* listening socket for IPv6 */
241 #endif
242 	struct net_device *dev;
243 	struct net	  *net;		/* netns for packet i/o */
244 	struct vxlan_rdst default_dst;	/* default destination */
245 
246 	struct timer_list age_timer;
247 	spinlock_t	  hash_lock[FDB_HASH_SIZE];
248 	unsigned int	  addrcnt;
249 	struct gro_cells  gro_cells;
250 
251 	struct vxlan_config	cfg;
252 
253 	struct hlist_head fdb_head[FDB_HASH_SIZE];
254 };
255 
256 #define VXLAN_F_LEARN			0x01
257 #define VXLAN_F_PROXY			0x02
258 #define VXLAN_F_RSC			0x04
259 #define VXLAN_F_L2MISS			0x08
260 #define VXLAN_F_L3MISS			0x10
261 #define VXLAN_F_IPV6			0x20
262 #define VXLAN_F_UDP_ZERO_CSUM_TX	0x40
263 #define VXLAN_F_UDP_ZERO_CSUM6_TX	0x80
264 #define VXLAN_F_UDP_ZERO_CSUM6_RX	0x100
265 #define VXLAN_F_REMCSUM_TX		0x200
266 #define VXLAN_F_REMCSUM_RX		0x400
267 #define VXLAN_F_GBP			0x800
268 #define VXLAN_F_REMCSUM_NOPARTIAL	0x1000
269 #define VXLAN_F_COLLECT_METADATA	0x2000
270 #define VXLAN_F_GPE			0x4000
271 #define VXLAN_F_IPV6_LINKLOCAL		0x8000
272 #define VXLAN_F_TTL_INHERIT		0x10000
273 
274 /* Flags that are used in the receive path. These flags must match in
275  * order for a socket to be shareable
276  */
277 #define VXLAN_F_RCV_FLAGS		(VXLAN_F_GBP |			\
278 					 VXLAN_F_GPE |			\
279 					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
280 					 VXLAN_F_REMCSUM_RX |		\
281 					 VXLAN_F_REMCSUM_NOPARTIAL |	\
282 					 VXLAN_F_COLLECT_METADATA)
283 
284 /* Flags that can be set together with VXLAN_F_GPE. */
285 #define VXLAN_F_ALLOWED_GPE		(VXLAN_F_GPE |			\
286 					 VXLAN_F_IPV6 |			\
287 					 VXLAN_F_IPV6_LINKLOCAL |	\
288 					 VXLAN_F_UDP_ZERO_CSUM_TX |	\
289 					 VXLAN_F_UDP_ZERO_CSUM6_TX |	\
290 					 VXLAN_F_UDP_ZERO_CSUM6_RX |	\
291 					 VXLAN_F_COLLECT_METADATA)
292 
293 struct net_device *vxlan_dev_create(struct net *net, const char *name,
294 				    u8 name_assign_type, struct vxlan_config *conf);
295 
296 static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
297 						     netdev_features_t features)
298 {
299 	u8 l4_hdr = 0;
300 
301 	if (!skb->encapsulation)
302 		return features;
303 
304 	switch (vlan_get_protocol(skb)) {
305 	case htons(ETH_P_IP):
306 		l4_hdr = ip_hdr(skb)->protocol;
307 		break;
308 	case htons(ETH_P_IPV6):
309 		l4_hdr = ipv6_hdr(skb)->nexthdr;
310 		break;
311 	default:
312 		return features;
313 	}
314 
315 	if ((l4_hdr == IPPROTO_UDP) &&
316 	    (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
317 	     skb->inner_protocol != htons(ETH_P_TEB) ||
318 	     (skb_inner_mac_header(skb) - skb_transport_header(skb) !=
319 	      sizeof(struct udphdr) + sizeof(struct vxlanhdr)) ||
320 	     (skb->ip_summed != CHECKSUM_NONE &&
321 	      !can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
322 		return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
323 
324 	return features;
325 }
326 
327 /* IP header + UDP + VXLAN + Ethernet header */
328 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
329 /* IPv6 header + UDP + VXLAN + Ethernet header */
330 #define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
331 
332 static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
333 {
334 	return (struct vxlanhdr *)(udp_hdr(skb) + 1);
335 }
336 
337 static inline __be32 vxlan_vni(__be32 vni_field)
338 {
339 #if defined(__BIG_ENDIAN)
340 	return (__force __be32)((__force u32)vni_field >> 8);
341 #else
342 	return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
343 #endif
344 }
345 
346 static inline __be32 vxlan_vni_field(__be32 vni)
347 {
348 #if defined(__BIG_ENDIAN)
349 	return (__force __be32)((__force u32)vni << 8);
350 #else
351 	return (__force __be32)((__force u32)vni >> 8);
352 #endif
353 }
354 
355 static inline size_t vxlan_rco_start(__be32 vni_field)
356 {
357 	return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
358 }
359 
360 static inline size_t vxlan_rco_offset(__be32 vni_field)
361 {
362 	return (vni_field & VXLAN_RCO_UDP) ?
363 		offsetof(struct udphdr, check) :
364 		offsetof(struct tcphdr, check);
365 }
366 
367 static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
368 {
369 	__be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
370 
371 	if (offset == offsetof(struct udphdr, check))
372 		vni_field |= VXLAN_RCO_UDP;
373 	return vni_field;
374 }
375 
376 static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
377 {
378 	return vs->sock->sk->sk_family;
379 }
380 
381 #if IS_ENABLED(CONFIG_IPV6)
382 
383 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
384 {
385 	if (ipa->sa.sa_family == AF_INET6)
386 		return ipv6_addr_any(&ipa->sin6.sin6_addr);
387 	else
388 		return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
389 }
390 
391 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
392 {
393 	if (ipa->sa.sa_family == AF_INET6)
394 		return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
395 	else
396 		return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
397 }
398 
399 #else /* !IS_ENABLED(CONFIG_IPV6) */
400 
401 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
402 {
403 	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
404 }
405 
406 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
407 {
408 	return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
409 }
410 
411 #endif /* IS_ENABLED(CONFIG_IPV6) */
412 
413 static inline bool netif_is_vxlan(const struct net_device *dev)
414 {
415 	return dev->rtnl_link_ops &&
416 	       !strcmp(dev->rtnl_link_ops->kind, "vxlan");
417 }
418 
419 struct switchdev_notifier_vxlan_fdb_info {
420 	struct switchdev_notifier_info info; /* must be first */
421 	union vxlan_addr remote_ip;
422 	__be16 remote_port;
423 	__be32 remote_vni;
424 	u32 remote_ifindex;
425 	u8 eth_addr[ETH_ALEN];
426 	__be32 vni;
427 	bool offloaded;
428 	bool added_by_user;
429 };
430 
431 #if IS_ENABLED(CONFIG_VXLAN)
432 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
433 		      struct switchdev_notifier_vxlan_fdb_info *fdb_info);
434 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
435 		     struct notifier_block *nb,
436 		     struct netlink_ext_ack *extack);
437 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);
438 
439 #else
440 static inline int
441 vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
442 		  struct switchdev_notifier_vxlan_fdb_info *fdb_info)
443 {
444 	return -ENOENT;
445 }
446 
447 static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
448 				   struct notifier_block *nb,
449 				   struct netlink_ext_ack *extack)
450 {
451 	return -EOPNOTSUPP;
452 }
453 
454 static inline void
455 vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
456 {
457 }
458 #endif
459 
460 static inline void vxlan_flag_attr_error(int attrtype,
461 					 struct netlink_ext_ack *extack)
462 {
463 #define VXLAN_FLAG(flg) \
464 	case IFLA_VXLAN_##flg: \
465 		NL_SET_ERR_MSG_MOD(extack, \
466 				   "cannot change " #flg " flag"); \
467 		break
468 	switch (attrtype) {
469 	VXLAN_FLAG(TTL_INHERIT);
470 	VXLAN_FLAG(LEARNING);
471 	VXLAN_FLAG(PROXY);
472 	VXLAN_FLAG(RSC);
473 	VXLAN_FLAG(L2MISS);
474 	VXLAN_FLAG(L3MISS);
475 	VXLAN_FLAG(COLLECT_METADATA);
476 	VXLAN_FLAG(UDP_ZERO_CSUM6_TX);
477 	VXLAN_FLAG(UDP_ZERO_CSUM6_RX);
478 	VXLAN_FLAG(REMCSUM_TX);
479 	VXLAN_FLAG(REMCSUM_RX);
480 	VXLAN_FLAG(GBP);
481 	VXLAN_FLAG(GPE);
482 	VXLAN_FLAG(REMCSUM_NOPARTIAL);
483 	default:
484 		NL_SET_ERR_MSG_MOD(extack, \
485 				   "cannot change flag");
486 		break;
487 	}
488 #undef VXLAN_FLAG
489 }
490 
491 static inline bool vxlan_fdb_nh_path_select(struct nexthop *nh,
492 					    int hash,
493 					    struct vxlan_rdst *rdst)
494 {
495 	struct fib_nh_common *nhc;
496 
497 	nhc = nexthop_path_fdb_result(nh, hash);
498 	if (unlikely(!nhc))
499 		return false;
500 
501 	switch (nhc->nhc_gw_family) {
502 	case AF_INET:
503 		rdst->remote_ip.sin.sin_addr.s_addr = nhc->nhc_gw.ipv4;
504 		rdst->remote_ip.sa.sa_family = AF_INET;
505 		break;
506 	case AF_INET6:
507 		rdst->remote_ip.sin6.sin6_addr = nhc->nhc_gw.ipv6;
508 		rdst->remote_ip.sa.sa_family = AF_INET6;
509 		break;
510 	}
511 
512 	return true;
513 }
514 
515 #endif
516