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 net_device *remote_dev;
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 enum ifla_vxlan_df df;
223 };
224
225 struct vxlan_dev_node {
226 struct hlist_node hlist;
227 struct vxlan_dev *vxlan;
228 };
229
230 /* Pseudo network device */
231 struct vxlan_dev {
232 struct vxlan_dev_node hlist4; /* vni hash table for IPv4 socket */
233 #if IS_ENABLED(CONFIG_IPV6)
234 struct vxlan_dev_node hlist6; /* vni hash table for IPv6 socket */
235 #endif
236 struct list_head next; /* vxlan's per namespace list */
237 struct vxlan_sock __rcu *vn4_sock; /* listening socket for IPv4 */
238 #if IS_ENABLED(CONFIG_IPV6)
239 struct vxlan_sock __rcu *vn6_sock; /* listening socket for IPv6 */
240 #endif
241 struct net_device *dev;
242 struct net *net; /* netns for packet i/o */
243 struct vxlan_rdst default_dst; /* default destination */
244
245 struct timer_list age_timer;
246 spinlock_t hash_lock[FDB_HASH_SIZE];
247 unsigned int addrcnt;
248 struct gro_cells gro_cells;
249
250 struct vxlan_config cfg;
251
252 struct hlist_head fdb_head[FDB_HASH_SIZE];
253 };
254
255 #define VXLAN_F_LEARN 0x01
256 #define VXLAN_F_PROXY 0x02
257 #define VXLAN_F_RSC 0x04
258 #define VXLAN_F_L2MISS 0x08
259 #define VXLAN_F_L3MISS 0x10
260 #define VXLAN_F_IPV6 0x20
261 #define VXLAN_F_UDP_ZERO_CSUM_TX 0x40
262 #define VXLAN_F_UDP_ZERO_CSUM6_TX 0x80
263 #define VXLAN_F_UDP_ZERO_CSUM6_RX 0x100
264 #define VXLAN_F_REMCSUM_TX 0x200
265 #define VXLAN_F_REMCSUM_RX 0x400
266 #define VXLAN_F_GBP 0x800
267 #define VXLAN_F_REMCSUM_NOPARTIAL 0x1000
268 #define VXLAN_F_COLLECT_METADATA 0x2000
269 #define VXLAN_F_GPE 0x4000
270 #define VXLAN_F_IPV6_LINKLOCAL 0x8000
271 #define VXLAN_F_TTL_INHERIT 0x10000
272
273 /* Flags that are used in the receive path. These flags must match in
274 * order for a socket to be shareable
275 */
276 #define VXLAN_F_RCV_FLAGS (VXLAN_F_GBP | \
277 VXLAN_F_GPE | \
278 VXLAN_F_UDP_ZERO_CSUM6_RX | \
279 VXLAN_F_REMCSUM_RX | \
280 VXLAN_F_REMCSUM_NOPARTIAL | \
281 VXLAN_F_COLLECT_METADATA)
282
283 /* Flags that can be set together with VXLAN_F_GPE. */
284 #define VXLAN_F_ALLOWED_GPE (VXLAN_F_GPE | \
285 VXLAN_F_IPV6 | \
286 VXLAN_F_IPV6_LINKLOCAL | \
287 VXLAN_F_UDP_ZERO_CSUM_TX | \
288 VXLAN_F_UDP_ZERO_CSUM6_TX | \
289 VXLAN_F_UDP_ZERO_CSUM6_RX | \
290 VXLAN_F_COLLECT_METADATA)
291
292 struct net_device *vxlan_dev_create(struct net *net, const char *name,
293 u8 name_assign_type, struct vxlan_config *conf);
294
vxlan_features_check(struct sk_buff * skb,netdev_features_t features)295 static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
296 netdev_features_t features)
297 {
298 u8 l4_hdr = 0;
299
300 if (!skb->encapsulation)
301 return features;
302
303 switch (vlan_get_protocol(skb)) {
304 case htons(ETH_P_IP):
305 l4_hdr = ip_hdr(skb)->protocol;
306 break;
307 case htons(ETH_P_IPV6):
308 l4_hdr = ipv6_hdr(skb)->nexthdr;
309 break;
310 default:
311 return features;
312 }
313
314 if ((l4_hdr == IPPROTO_UDP) &&
315 (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
316 skb->inner_protocol != htons(ETH_P_TEB) ||
317 (skb_inner_mac_header(skb) - skb_transport_header(skb) !=
318 sizeof(struct udphdr) + sizeof(struct vxlanhdr)) ||
319 (skb->ip_summed != CHECKSUM_NONE &&
320 !can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
321 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
322
323 return features;
324 }
325
326 /* IP header + UDP + VXLAN + Ethernet header */
327 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
328 /* IPv6 header + UDP + VXLAN + Ethernet header */
329 #define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
330
vxlan_hdr(struct sk_buff * skb)331 static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
332 {
333 return (struct vxlanhdr *)(udp_hdr(skb) + 1);
334 }
335
vxlan_vni(__be32 vni_field)336 static inline __be32 vxlan_vni(__be32 vni_field)
337 {
338 #if defined(__BIG_ENDIAN)
339 return (__force __be32)((__force u32)vni_field >> 8);
340 #else
341 return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
342 #endif
343 }
344
vxlan_vni_field(__be32 vni)345 static inline __be32 vxlan_vni_field(__be32 vni)
346 {
347 #if defined(__BIG_ENDIAN)
348 return (__force __be32)((__force u32)vni << 8);
349 #else
350 return (__force __be32)((__force u32)vni >> 8);
351 #endif
352 }
353
vxlan_rco_start(__be32 vni_field)354 static inline size_t vxlan_rco_start(__be32 vni_field)
355 {
356 return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
357 }
358
vxlan_rco_offset(__be32 vni_field)359 static inline size_t vxlan_rco_offset(__be32 vni_field)
360 {
361 return (vni_field & VXLAN_RCO_UDP) ?
362 offsetof(struct udphdr, check) :
363 offsetof(struct tcphdr, check);
364 }
365
vxlan_compute_rco(unsigned int start,unsigned int offset)366 static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
367 {
368 __be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
369
370 if (offset == offsetof(struct udphdr, check))
371 vni_field |= VXLAN_RCO_UDP;
372 return vni_field;
373 }
374
vxlan_get_sk_family(struct vxlan_sock * vs)375 static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
376 {
377 return vs->sock->sk->sk_family;
378 }
379
380 #if IS_ENABLED(CONFIG_IPV6)
381
vxlan_addr_any(const union vxlan_addr * ipa)382 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
383 {
384 if (ipa->sa.sa_family == AF_INET6)
385 return ipv6_addr_any(&ipa->sin6.sin6_addr);
386 else
387 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
388 }
389
vxlan_addr_multicast(const union vxlan_addr * ipa)390 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
391 {
392 if (ipa->sa.sa_family == AF_INET6)
393 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
394 else
395 return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
396 }
397
398 #else /* !IS_ENABLED(CONFIG_IPV6) */
399
vxlan_addr_any(const union vxlan_addr * ipa)400 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
401 {
402 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
403 }
404
vxlan_addr_multicast(const union vxlan_addr * ipa)405 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
406 {
407 return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
408 }
409
410 #endif /* IS_ENABLED(CONFIG_IPV6) */
411
netif_is_vxlan(const struct net_device * dev)412 static inline bool netif_is_vxlan(const struct net_device *dev)
413 {
414 return dev->rtnl_link_ops &&
415 !strcmp(dev->rtnl_link_ops->kind, "vxlan");
416 }
417
418 struct switchdev_notifier_vxlan_fdb_info {
419 struct switchdev_notifier_info info; /* must be first */
420 union vxlan_addr remote_ip;
421 __be16 remote_port;
422 __be32 remote_vni;
423 u32 remote_ifindex;
424 u8 eth_addr[ETH_ALEN];
425 __be32 vni;
426 bool offloaded;
427 bool added_by_user;
428 };
429
430 #if IS_ENABLED(CONFIG_VXLAN)
431 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
432 struct switchdev_notifier_vxlan_fdb_info *fdb_info);
433 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
434 struct notifier_block *nb,
435 struct netlink_ext_ack *extack);
436 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);
437
438 #else
439 static inline int
vxlan_fdb_find_uc(struct net_device * dev,const u8 * mac,__be32 vni,struct switchdev_notifier_vxlan_fdb_info * fdb_info)440 vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
441 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
442 {
443 return -ENOENT;
444 }
445
vxlan_fdb_replay(const struct net_device * dev,__be32 vni,struct notifier_block * nb,struct netlink_ext_ack * extack)446 static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
447 struct notifier_block *nb,
448 struct netlink_ext_ack *extack)
449 {
450 return -EOPNOTSUPP;
451 }
452
453 static inline void
vxlan_fdb_clear_offload(const struct net_device * dev,__be32 vni)454 vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
455 {
456 }
457 #endif
458
vxlan_flag_attr_error(int attrtype,struct netlink_ext_ack * extack)459 static inline void vxlan_flag_attr_error(int attrtype,
460 struct netlink_ext_ack *extack)
461 {
462 #define VXLAN_FLAG(flg) \
463 case IFLA_VXLAN_##flg: \
464 NL_SET_ERR_MSG_MOD(extack, \
465 "cannot change " #flg " flag"); \
466 break
467 switch (attrtype) {
468 VXLAN_FLAG(TTL_INHERIT);
469 VXLAN_FLAG(LEARNING);
470 VXLAN_FLAG(PROXY);
471 VXLAN_FLAG(RSC);
472 VXLAN_FLAG(L2MISS);
473 VXLAN_FLAG(L3MISS);
474 VXLAN_FLAG(COLLECT_METADATA);
475 VXLAN_FLAG(UDP_ZERO_CSUM6_TX);
476 VXLAN_FLAG(UDP_ZERO_CSUM6_RX);
477 VXLAN_FLAG(REMCSUM_TX);
478 VXLAN_FLAG(REMCSUM_RX);
479 VXLAN_FLAG(GBP);
480 VXLAN_FLAG(GPE);
481 VXLAN_FLAG(REMCSUM_NOPARTIAL);
482 default:
483 NL_SET_ERR_MSG_MOD(extack, \
484 "cannot change flag");
485 break;
486 }
487 #undef VXLAN_FLAG
488 }
489
490 #endif
491
