Loading...
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 <linux/rhashtable-types.h>
7#include <net/udp_tunnel.h>
8#include <net/dst_metadata.h>
9#include <net/rtnetlink.h>
10#include <net/switchdev.h>
11#include <net/nexthop.h>
12
13#define IANA_VXLAN_UDP_PORT 4789
14#define IANA_VXLAN_GPE_UDP_PORT 4790
15
16/* VXLAN protocol (RFC 7348) header:
17 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
18 * |R|R|R|R|I|R|R|R| Reserved |
19 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
20 * | VXLAN Network Identifier (VNI) | Reserved |
21 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
22 *
23 * I = VXLAN Network Identifier (VNI) present.
24 */
25struct vxlanhdr {
26 __be32 vx_flags;
27 __be32 vx_vni;
28};
29
30/* VXLAN header flags. */
31#define VXLAN_HF_VNI cpu_to_be32(BIT(27))
32
33#define VXLAN_N_VID (1u << 24)
34#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
35#define VXLAN_VNI_MASK cpu_to_be32(VXLAN_VID_MASK << 8)
36#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
37
38#define VNI_HASH_BITS 10
39#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
40#define FDB_HASH_BITS 8
41#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
42
43/* Remote checksum offload for VXLAN (VXLAN_F_REMCSUM_[RT]X):
44 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
45 * |R|R|R|R|I|R|R|R|R|R|C| Reserved |
46 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
47 * | VXLAN Network Identifier (VNI) |O| Csum start |
48 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
49 *
50 * C = Remote checksum offload bit. When set indicates that the
51 * remote checksum offload data is present.
52 *
53 * O = Offset bit. Indicates the checksum offset relative to
54 * checksum start.
55 *
56 * Csum start = Checksum start divided by two.
57 *
58 * http://tools.ietf.org/html/draft-herbert-vxlan-rco
59 */
60
61/* VXLAN-RCO header flags. */
62#define VXLAN_HF_RCO cpu_to_be32(BIT(21))
63
64/* Remote checksum offload header option */
65#define VXLAN_RCO_MASK cpu_to_be32(0x7f) /* Last byte of vni field */
66#define VXLAN_RCO_UDP cpu_to_be32(0x80) /* Indicate UDP RCO (TCP when not set *) */
67#define VXLAN_RCO_SHIFT 1 /* Left shift of start */
68#define VXLAN_RCO_SHIFT_MASK ((1 << VXLAN_RCO_SHIFT) - 1)
69#define VXLAN_MAX_REMCSUM_START (0x7f << VXLAN_RCO_SHIFT)
70
71/*
72 * VXLAN Group Based Policy Extension (VXLAN_F_GBP):
73 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
74 * |G|R|R|R|I|R|R|R|R|D|R|R|A|R|R|R| Group Policy ID |
75 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
76 * | VXLAN Network Identifier (VNI) | Reserved |
77 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
78 *
79 * G = Group Policy ID present.
80 *
81 * D = Don't Learn bit. When set, this bit indicates that the egress
82 * VTEP MUST NOT learn the source address of the encapsulated frame.
83 *
84 * A = Indicates that the group policy has already been applied to
85 * this packet. Policies MUST NOT be applied by devices when the
86 * A bit is set.
87 *
88 * https://tools.ietf.org/html/draft-smith-vxlan-group-policy
89 */
90struct vxlanhdr_gbp {
91 u8 vx_flags;
92#ifdef __LITTLE_ENDIAN_BITFIELD
93 u8 reserved_flags1:3,
94 policy_applied:1,
95 reserved_flags2:2,
96 dont_learn:1,
97 reserved_flags3:1;
98#elif defined(__BIG_ENDIAN_BITFIELD)
99 u8 reserved_flags1:1,
100 dont_learn:1,
101 reserved_flags2:2,
102 policy_applied:1,
103 reserved_flags3:3;
104#else
105#error "Please fix <asm/byteorder.h>"
106#endif
107 __be16 policy_id;
108 __be32 vx_vni;
109};
110
111/* VXLAN-GBP header flags. */
112#define VXLAN_HF_GBP cpu_to_be32(BIT(31))
113
114#define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP | cpu_to_be32(0xFFFFFF))
115
116/* skb->mark mapping
117 *
118 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
119 * |R|R|R|R|R|R|R|R|R|D|R|R|A|R|R|R| Group Policy ID |
120 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
121 */
122#define VXLAN_GBP_DONT_LEARN (BIT(6) << 16)
123#define VXLAN_GBP_POLICY_APPLIED (BIT(3) << 16)
124#define VXLAN_GBP_ID_MASK (0xFFFF)
125
126#define VXLAN_GBP_MASK (VXLAN_GBP_DONT_LEARN | VXLAN_GBP_POLICY_APPLIED | \
127 VXLAN_GBP_ID_MASK)
128
129/*
130 * VXLAN Generic Protocol Extension (VXLAN_F_GPE):
131 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
132 * |R|R|Ver|I|P|R|O| Reserved |Next Protocol |
133 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
134 * | VXLAN Network Identifier (VNI) | Reserved |
135 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
136 *
137 * Ver = Version. Indicates VXLAN GPE protocol version.
138 *
139 * P = Next Protocol Bit. The P bit is set to indicate that the
140 * Next Protocol field is present.
141 *
142 * O = OAM Flag Bit. The O bit is set to indicate that the packet
143 * is an OAM packet.
144 *
145 * Next Protocol = This 8 bit field indicates the protocol header
146 * immediately following the VXLAN GPE header.
147 *
148 * https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
149 */
150
151struct vxlanhdr_gpe {
152#if defined(__LITTLE_ENDIAN_BITFIELD)
153 u8 oam_flag:1,
154 reserved_flags1:1,
155 np_applied:1,
156 instance_applied:1,
157 version:2,
158 reserved_flags2:2;
159#elif defined(__BIG_ENDIAN_BITFIELD)
160 u8 reserved_flags2:2,
161 version:2,
162 instance_applied:1,
163 np_applied:1,
164 reserved_flags1:1,
165 oam_flag:1;
166#endif
167 u8 reserved_flags3;
168 u8 reserved_flags4;
169 u8 next_protocol;
170 __be32 vx_vni;
171};
172
173/* VXLAN-GPE header flags. */
174#define VXLAN_HF_VER cpu_to_be32(BIT(29) | BIT(28))
175#define VXLAN_HF_NP cpu_to_be32(BIT(26))
176#define VXLAN_HF_OAM cpu_to_be32(BIT(24))
177
178#define VXLAN_GPE_USED_BITS (VXLAN_HF_VER | VXLAN_HF_NP | VXLAN_HF_OAM | \
179 cpu_to_be32(0xff))
180
181struct vxlan_metadata {
182 u32 gbp;
183};
184
185/* per UDP socket information */
186struct vxlan_sock {
187 struct hlist_node hlist;
188 struct socket *sock;
189 struct hlist_head vni_list[VNI_HASH_SIZE];
190 refcount_t refcnt;
191 u32 flags;
192};
193
194union vxlan_addr {
195 struct sockaddr_in sin;
196 struct sockaddr_in6 sin6;
197 struct sockaddr sa;
198};
199
200struct vxlan_rdst {
201 union vxlan_addr remote_ip;
202 __be16 remote_port;
203 u8 offloaded:1;
204 __be32 remote_vni;
205 u32 remote_ifindex;
206 struct net_device *remote_dev;
207 struct list_head list;
208 struct rcu_head rcu;
209 struct dst_cache dst_cache;
210};
211
212struct vxlan_config {
213 union vxlan_addr remote_ip;
214 union vxlan_addr saddr;
215 __be32 vni;
216 int remote_ifindex;
217 int mtu;
218 __be16 dst_port;
219 u16 port_min;
220 u16 port_max;
221 u8 tos;
222 u8 ttl;
223 __be32 label;
224 enum ifla_vxlan_label_policy label_policy;
225 u32 flags;
226 unsigned long age_interval;
227 unsigned int addrmax;
228 bool no_share;
229 enum ifla_vxlan_df df;
230};
231
232enum {
233 VXLAN_VNI_STATS_RX,
234 VXLAN_VNI_STATS_RX_DROPS,
235 VXLAN_VNI_STATS_RX_ERRORS,
236 VXLAN_VNI_STATS_TX,
237 VXLAN_VNI_STATS_TX_DROPS,
238 VXLAN_VNI_STATS_TX_ERRORS,
239};
240
241struct vxlan_vni_stats {
242 u64 rx_packets;
243 u64 rx_bytes;
244 u64 rx_drops;
245 u64 rx_errors;
246 u64 tx_packets;
247 u64 tx_bytes;
248 u64 tx_drops;
249 u64 tx_errors;
250};
251
252struct vxlan_vni_stats_pcpu {
253 struct vxlan_vni_stats stats;
254 struct u64_stats_sync syncp;
255};
256
257struct vxlan_dev_node {
258 struct hlist_node hlist;
259 struct vxlan_dev *vxlan;
260};
261
262struct vxlan_vni_node {
263 struct rhash_head vnode;
264 struct vxlan_dev_node hlist4; /* vni hash table for IPv4 socket */
265#if IS_ENABLED(CONFIG_IPV6)
266 struct vxlan_dev_node hlist6; /* vni hash table for IPv6 socket */
267#endif
268 struct list_head vlist;
269 __be32 vni;
270 union vxlan_addr remote_ip; /* default remote ip for this vni */
271 struct vxlan_vni_stats_pcpu __percpu *stats;
272
273 struct rcu_head rcu;
274};
275
276struct vxlan_vni_group {
277 struct rhashtable vni_hash;
278 struct list_head vni_list;
279 u32 num_vnis;
280};
281
282/* Pseudo network device */
283struct vxlan_dev {
284 struct vxlan_dev_node hlist4; /* vni hash table for IPv4 socket */
285#if IS_ENABLED(CONFIG_IPV6)
286 struct vxlan_dev_node hlist6; /* vni hash table for IPv6 socket */
287#endif
288 struct list_head next; /* vxlan's per namespace list */
289 struct vxlan_sock __rcu *vn4_sock; /* listening socket for IPv4 */
290#if IS_ENABLED(CONFIG_IPV6)
291 struct vxlan_sock __rcu *vn6_sock; /* listening socket for IPv6 */
292#endif
293 struct net_device *dev;
294 struct net *net; /* netns for packet i/o */
295 struct vxlan_rdst default_dst; /* default destination */
296
297 struct timer_list age_timer;
298 spinlock_t hash_lock[FDB_HASH_SIZE];
299 unsigned int addrcnt;
300 struct gro_cells gro_cells;
301
302 struct vxlan_config cfg;
303
304 struct vxlan_vni_group __rcu *vnigrp;
305
306 struct hlist_head fdb_head[FDB_HASH_SIZE];
307
308 struct rhashtable mdb_tbl;
309 struct hlist_head mdb_list;
310 unsigned int mdb_seq;
311};
312
313#define VXLAN_F_LEARN 0x01
314#define VXLAN_F_PROXY 0x02
315#define VXLAN_F_RSC 0x04
316#define VXLAN_F_L2MISS 0x08
317#define VXLAN_F_L3MISS 0x10
318#define VXLAN_F_IPV6 0x20
319#define VXLAN_F_UDP_ZERO_CSUM_TX 0x40
320#define VXLAN_F_UDP_ZERO_CSUM6_TX 0x80
321#define VXLAN_F_UDP_ZERO_CSUM6_RX 0x100
322#define VXLAN_F_REMCSUM_TX 0x200
323#define VXLAN_F_REMCSUM_RX 0x400
324#define VXLAN_F_GBP 0x800
325#define VXLAN_F_REMCSUM_NOPARTIAL 0x1000
326#define VXLAN_F_COLLECT_METADATA 0x2000
327#define VXLAN_F_GPE 0x4000
328#define VXLAN_F_IPV6_LINKLOCAL 0x8000
329#define VXLAN_F_TTL_INHERIT 0x10000
330#define VXLAN_F_VNIFILTER 0x20000
331#define VXLAN_F_MDB 0x40000
332#define VXLAN_F_LOCALBYPASS 0x80000
333
334/* Flags that are used in the receive path. These flags must match in
335 * order for a socket to be shareable
336 */
337#define VXLAN_F_RCV_FLAGS (VXLAN_F_GBP | \
338 VXLAN_F_GPE | \
339 VXLAN_F_UDP_ZERO_CSUM6_RX | \
340 VXLAN_F_REMCSUM_RX | \
341 VXLAN_F_REMCSUM_NOPARTIAL | \
342 VXLAN_F_COLLECT_METADATA | \
343 VXLAN_F_VNIFILTER)
344
345/* Flags that can be set together with VXLAN_F_GPE. */
346#define VXLAN_F_ALLOWED_GPE (VXLAN_F_GPE | \
347 VXLAN_F_IPV6 | \
348 VXLAN_F_IPV6_LINKLOCAL | \
349 VXLAN_F_UDP_ZERO_CSUM_TX | \
350 VXLAN_F_UDP_ZERO_CSUM6_TX | \
351 VXLAN_F_UDP_ZERO_CSUM6_RX | \
352 VXLAN_F_COLLECT_METADATA | \
353 VXLAN_F_VNIFILTER | \
354 VXLAN_F_LOCALBYPASS)
355
356struct net_device *vxlan_dev_create(struct net *net, const char *name,
357 u8 name_assign_type, struct vxlan_config *conf);
358
359static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
360 netdev_features_t features)
361{
362 u8 l4_hdr = 0;
363
364 if (!skb->encapsulation)
365 return features;
366
367 switch (vlan_get_protocol(skb)) {
368 case htons(ETH_P_IP):
369 l4_hdr = ip_hdr(skb)->protocol;
370 break;
371 case htons(ETH_P_IPV6):
372 l4_hdr = ipv6_hdr(skb)->nexthdr;
373 break;
374 default:
375 return features;
376 }
377
378 if ((l4_hdr == IPPROTO_UDP) &&
379 (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
380 skb->inner_protocol != htons(ETH_P_TEB) ||
381 (skb_inner_mac_header(skb) - skb_transport_header(skb) !=
382 sizeof(struct udphdr) + sizeof(struct vxlanhdr)) ||
383 (skb->ip_summed != CHECKSUM_NONE &&
384 !can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
385 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
386
387 return features;
388}
389
390static inline int vxlan_headroom(u32 flags)
391{
392 /* VXLAN: IP4/6 header + UDP + VXLAN + Ethernet header */
393 /* VXLAN-GPE: IP4/6 header + UDP + VXLAN */
394 return (flags & VXLAN_F_IPV6 ? sizeof(struct ipv6hdr) :
395 sizeof(struct iphdr)) +
396 sizeof(struct udphdr) + sizeof(struct vxlanhdr) +
397 (flags & VXLAN_F_GPE ? 0 : ETH_HLEN);
398}
399
400static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
401{
402 return (struct vxlanhdr *)(udp_hdr(skb) + 1);
403}
404
405static inline __be32 vxlan_vni(__be32 vni_field)
406{
407#if defined(__BIG_ENDIAN)
408 return (__force __be32)((__force u32)vni_field >> 8);
409#else
410 return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
411#endif
412}
413
414static inline __be32 vxlan_vni_field(__be32 vni)
415{
416#if defined(__BIG_ENDIAN)
417 return (__force __be32)((__force u32)vni << 8);
418#else
419 return (__force __be32)((__force u32)vni >> 8);
420#endif
421}
422
423static inline size_t vxlan_rco_start(__be32 vni_field)
424{
425 return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
426}
427
428static inline size_t vxlan_rco_offset(__be32 vni_field)
429{
430 return (vni_field & VXLAN_RCO_UDP) ?
431 offsetof(struct udphdr, check) :
432 offsetof(struct tcphdr, check);
433}
434
435static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
436{
437 __be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
438
439 if (offset == offsetof(struct udphdr, check))
440 vni_field |= VXLAN_RCO_UDP;
441 return vni_field;
442}
443
444static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
445{
446 return vs->sock->sk->sk_family;
447}
448
449#if IS_ENABLED(CONFIG_IPV6)
450
451static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
452{
453 if (ipa->sa.sa_family == AF_INET6)
454 return ipv6_addr_any(&ipa->sin6.sin6_addr);
455 else
456 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
457}
458
459static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
460{
461 if (ipa->sa.sa_family == AF_INET6)
462 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
463 else
464 return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
465}
466
467#else /* !IS_ENABLED(CONFIG_IPV6) */
468
469static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
470{
471 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
472}
473
474static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
475{
476 return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
477}
478
479#endif /* IS_ENABLED(CONFIG_IPV6) */
480
481static inline bool netif_is_vxlan(const struct net_device *dev)
482{
483 return dev->rtnl_link_ops &&
484 !strcmp(dev->rtnl_link_ops->kind, "vxlan");
485}
486
487struct switchdev_notifier_vxlan_fdb_info {
488 struct switchdev_notifier_info info; /* must be first */
489 union vxlan_addr remote_ip;
490 __be16 remote_port;
491 __be32 remote_vni;
492 u32 remote_ifindex;
493 u8 eth_addr[ETH_ALEN];
494 __be32 vni;
495 bool offloaded;
496 bool added_by_user;
497};
498
499#if IS_ENABLED(CONFIG_VXLAN)
500int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
501 struct switchdev_notifier_vxlan_fdb_info *fdb_info);
502int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
503 struct notifier_block *nb,
504 struct netlink_ext_ack *extack);
505void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);
506
507#else
508static inline int
509vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
510 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
511{
512 return -ENOENT;
513}
514
515static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
516 struct notifier_block *nb,
517 struct netlink_ext_ack *extack)
518{
519 return -EOPNOTSUPP;
520}
521
522static inline void
523vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
524{
525}
526#endif
527
528static inline void vxlan_flag_attr_error(int attrtype,
529 struct netlink_ext_ack *extack)
530{
531#define VXLAN_FLAG(flg) \
532 case IFLA_VXLAN_##flg: \
533 NL_SET_ERR_MSG_MOD(extack, \
534 "cannot change " #flg " flag"); \
535 break
536 switch (attrtype) {
537 VXLAN_FLAG(TTL_INHERIT);
538 VXLAN_FLAG(LEARNING);
539 VXLAN_FLAG(PROXY);
540 VXLAN_FLAG(RSC);
541 VXLAN_FLAG(L2MISS);
542 VXLAN_FLAG(L3MISS);
543 VXLAN_FLAG(COLLECT_METADATA);
544 VXLAN_FLAG(UDP_ZERO_CSUM6_TX);
545 VXLAN_FLAG(UDP_ZERO_CSUM6_RX);
546 VXLAN_FLAG(REMCSUM_TX);
547 VXLAN_FLAG(REMCSUM_RX);
548 VXLAN_FLAG(GBP);
549 VXLAN_FLAG(GPE);
550 VXLAN_FLAG(REMCSUM_NOPARTIAL);
551 default:
552 NL_SET_ERR_MSG_MOD(extack, \
553 "cannot change flag");
554 break;
555 }
556#undef VXLAN_FLAG
557}
558
559static inline bool vxlan_fdb_nh_path_select(struct nexthop *nh,
560 u32 hash,
561 struct vxlan_rdst *rdst)
562{
563 struct fib_nh_common *nhc;
564
565 nhc = nexthop_path_fdb_result(nh, hash >> 1);
566 if (unlikely(!nhc))
567 return false;
568
569 switch (nhc->nhc_gw_family) {
570 case AF_INET:
571 rdst->remote_ip.sin.sin_addr.s_addr = nhc->nhc_gw.ipv4;
572 rdst->remote_ip.sa.sa_family = AF_INET;
573 break;
574 case AF_INET6:
575 rdst->remote_ip.sin6.sin6_addr = nhc->nhc_gw.ipv6;
576 rdst->remote_ip.sa.sa_family = AF_INET6;
577 break;
578 }
579
580 return true;
581}
582
583static inline void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, const struct vxlan_metadata *md)
584{
585 struct vxlanhdr_gbp *gbp;
586
587 if (!md->gbp)
588 return;
589
590 gbp = (struct vxlanhdr_gbp *)vxh;
591 vxh->vx_flags |= VXLAN_HF_GBP;
592
593 if (md->gbp & VXLAN_GBP_DONT_LEARN)
594 gbp->dont_learn = 1;
595
596 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
597 gbp->policy_applied = 1;
598
599 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
600}
601
602#endif
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 */
23struct 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 */
88struct 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#define VXLAN_GBP_MASK (VXLAN_GBP_DONT_LEARN | VXLAN_GBP_POLICY_APPLIED | \
125 VXLAN_GBP_ID_MASK)
126
127/*
128 * VXLAN Generic Protocol Extension (VXLAN_F_GPE):
129 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
130 * |R|R|Ver|I|P|R|O| Reserved |Next Protocol |
131 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
132 * | VXLAN Network Identifier (VNI) | Reserved |
133 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
134 *
135 * Ver = Version. Indicates VXLAN GPE protocol version.
136 *
137 * P = Next Protocol Bit. The P bit is set to indicate that the
138 * Next Protocol field is present.
139 *
140 * O = OAM Flag Bit. The O bit is set to indicate that the packet
141 * is an OAM packet.
142 *
143 * Next Protocol = This 8 bit field indicates the protocol header
144 * immediately following the VXLAN GPE header.
145 *
146 * https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-01
147 */
148
149struct vxlanhdr_gpe {
150#if defined(__LITTLE_ENDIAN_BITFIELD)
151 u8 oam_flag:1,
152 reserved_flags1:1,
153 np_applied:1,
154 instance_applied:1,
155 version:2,
156 reserved_flags2:2;
157#elif defined(__BIG_ENDIAN_BITFIELD)
158 u8 reserved_flags2:2,
159 version:2,
160 instance_applied:1,
161 np_applied:1,
162 reserved_flags1:1,
163 oam_flag:1;
164#endif
165 u8 reserved_flags3;
166 u8 reserved_flags4;
167 u8 next_protocol;
168 __be32 vx_vni;
169};
170
171/* VXLAN-GPE header flags. */
172#define VXLAN_HF_VER cpu_to_be32(BIT(29) | BIT(28))
173#define VXLAN_HF_NP cpu_to_be32(BIT(26))
174#define VXLAN_HF_OAM cpu_to_be32(BIT(24))
175
176#define VXLAN_GPE_USED_BITS (VXLAN_HF_VER | VXLAN_HF_NP | VXLAN_HF_OAM | \
177 cpu_to_be32(0xff))
178
179struct vxlan_metadata {
180 u32 gbp;
181};
182
183/* per UDP socket information */
184struct vxlan_sock {
185 struct hlist_node hlist;
186 struct socket *sock;
187 struct hlist_head vni_list[VNI_HASH_SIZE];
188 refcount_t refcnt;
189 u32 flags;
190};
191
192union vxlan_addr {
193 struct sockaddr_in sin;
194 struct sockaddr_in6 sin6;
195 struct sockaddr sa;
196};
197
198struct vxlan_rdst {
199 union vxlan_addr remote_ip;
200 __be16 remote_port;
201 u8 offloaded:1;
202 __be32 remote_vni;
203 u32 remote_ifindex;
204 struct net_device *remote_dev;
205 struct list_head list;
206 struct rcu_head rcu;
207 struct dst_cache dst_cache;
208};
209
210struct vxlan_config {
211 union vxlan_addr remote_ip;
212 union vxlan_addr saddr;
213 __be32 vni;
214 int remote_ifindex;
215 int mtu;
216 __be16 dst_port;
217 u16 port_min;
218 u16 port_max;
219 u8 tos;
220 u8 ttl;
221 __be32 label;
222 u32 flags;
223 unsigned long age_interval;
224 unsigned int addrmax;
225 bool no_share;
226 enum ifla_vxlan_df df;
227};
228
229struct vxlan_dev_node {
230 struct hlist_node hlist;
231 struct vxlan_dev *vxlan;
232};
233
234/* Pseudo network device */
235struct vxlan_dev {
236 struct vxlan_dev_node hlist4; /* vni hash table for IPv4 socket */
237#if IS_ENABLED(CONFIG_IPV6)
238 struct vxlan_dev_node hlist6; /* vni hash table for IPv6 socket */
239#endif
240 struct list_head next; /* vxlan's per namespace list */
241 struct vxlan_sock __rcu *vn4_sock; /* listening socket for IPv4 */
242#if IS_ENABLED(CONFIG_IPV6)
243 struct vxlan_sock __rcu *vn6_sock; /* listening socket for IPv6 */
244#endif
245 struct net_device *dev;
246 struct net *net; /* netns for packet i/o */
247 struct vxlan_rdst default_dst; /* default destination */
248
249 struct timer_list age_timer;
250 spinlock_t hash_lock[FDB_HASH_SIZE];
251 unsigned int addrcnt;
252 struct gro_cells gro_cells;
253
254 struct vxlan_config cfg;
255
256 struct hlist_head fdb_head[FDB_HASH_SIZE];
257};
258
259#define VXLAN_F_LEARN 0x01
260#define VXLAN_F_PROXY 0x02
261#define VXLAN_F_RSC 0x04
262#define VXLAN_F_L2MISS 0x08
263#define VXLAN_F_L3MISS 0x10
264#define VXLAN_F_IPV6 0x20
265#define VXLAN_F_UDP_ZERO_CSUM_TX 0x40
266#define VXLAN_F_UDP_ZERO_CSUM6_TX 0x80
267#define VXLAN_F_UDP_ZERO_CSUM6_RX 0x100
268#define VXLAN_F_REMCSUM_TX 0x200
269#define VXLAN_F_REMCSUM_RX 0x400
270#define VXLAN_F_GBP 0x800
271#define VXLAN_F_REMCSUM_NOPARTIAL 0x1000
272#define VXLAN_F_COLLECT_METADATA 0x2000
273#define VXLAN_F_GPE 0x4000
274#define VXLAN_F_IPV6_LINKLOCAL 0x8000
275#define VXLAN_F_TTL_INHERIT 0x10000
276
277/* Flags that are used in the receive path. These flags must match in
278 * order for a socket to be shareable
279 */
280#define VXLAN_F_RCV_FLAGS (VXLAN_F_GBP | \
281 VXLAN_F_GPE | \
282 VXLAN_F_UDP_ZERO_CSUM6_RX | \
283 VXLAN_F_REMCSUM_RX | \
284 VXLAN_F_REMCSUM_NOPARTIAL | \
285 VXLAN_F_COLLECT_METADATA)
286
287/* Flags that can be set together with VXLAN_F_GPE. */
288#define VXLAN_F_ALLOWED_GPE (VXLAN_F_GPE | \
289 VXLAN_F_IPV6 | \
290 VXLAN_F_IPV6_LINKLOCAL | \
291 VXLAN_F_UDP_ZERO_CSUM_TX | \
292 VXLAN_F_UDP_ZERO_CSUM6_TX | \
293 VXLAN_F_UDP_ZERO_CSUM6_RX | \
294 VXLAN_F_COLLECT_METADATA)
295
296struct net_device *vxlan_dev_create(struct net *net, const char *name,
297 u8 name_assign_type, struct vxlan_config *conf);
298
299static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
300 netdev_features_t features)
301{
302 u8 l4_hdr = 0;
303
304 if (!skb->encapsulation)
305 return features;
306
307 switch (vlan_get_protocol(skb)) {
308 case htons(ETH_P_IP):
309 l4_hdr = ip_hdr(skb)->protocol;
310 break;
311 case htons(ETH_P_IPV6):
312 l4_hdr = ipv6_hdr(skb)->nexthdr;
313 break;
314 default:
315 return features;
316 }
317
318 if ((l4_hdr == IPPROTO_UDP) &&
319 (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
320 skb->inner_protocol != htons(ETH_P_TEB) ||
321 (skb_inner_mac_header(skb) - skb_transport_header(skb) !=
322 sizeof(struct udphdr) + sizeof(struct vxlanhdr)) ||
323 (skb->ip_summed != CHECKSUM_NONE &&
324 !can_checksum_protocol(features, inner_eth_hdr(skb)->h_proto))))
325 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
326
327 return features;
328}
329
330/* IP header + UDP + VXLAN + Ethernet header */
331#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
332/* IPv6 header + UDP + VXLAN + Ethernet header */
333#define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
334
335static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
336{
337 return (struct vxlanhdr *)(udp_hdr(skb) + 1);
338}
339
340static inline __be32 vxlan_vni(__be32 vni_field)
341{
342#if defined(__BIG_ENDIAN)
343 return (__force __be32)((__force u32)vni_field >> 8);
344#else
345 return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
346#endif
347}
348
349static inline __be32 vxlan_vni_field(__be32 vni)
350{
351#if defined(__BIG_ENDIAN)
352 return (__force __be32)((__force u32)vni << 8);
353#else
354 return (__force __be32)((__force u32)vni >> 8);
355#endif
356}
357
358static inline size_t vxlan_rco_start(__be32 vni_field)
359{
360 return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
361}
362
363static inline size_t vxlan_rco_offset(__be32 vni_field)
364{
365 return (vni_field & VXLAN_RCO_UDP) ?
366 offsetof(struct udphdr, check) :
367 offsetof(struct tcphdr, check);
368}
369
370static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
371{
372 __be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
373
374 if (offset == offsetof(struct udphdr, check))
375 vni_field |= VXLAN_RCO_UDP;
376 return vni_field;
377}
378
379static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
380{
381 return vs->sock->sk->sk_family;
382}
383
384#if IS_ENABLED(CONFIG_IPV6)
385
386static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
387{
388 if (ipa->sa.sa_family == AF_INET6)
389 return ipv6_addr_any(&ipa->sin6.sin6_addr);
390 else
391 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
392}
393
394static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
395{
396 if (ipa->sa.sa_family == AF_INET6)
397 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
398 else
399 return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
400}
401
402#else /* !IS_ENABLED(CONFIG_IPV6) */
403
404static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
405{
406 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
407}
408
409static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
410{
411 return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
412}
413
414#endif /* IS_ENABLED(CONFIG_IPV6) */
415
416static inline bool netif_is_vxlan(const struct net_device *dev)
417{
418 return dev->rtnl_link_ops &&
419 !strcmp(dev->rtnl_link_ops->kind, "vxlan");
420}
421
422struct switchdev_notifier_vxlan_fdb_info {
423 struct switchdev_notifier_info info; /* must be first */
424 union vxlan_addr remote_ip;
425 __be16 remote_port;
426 __be32 remote_vni;
427 u32 remote_ifindex;
428 u8 eth_addr[ETH_ALEN];
429 __be32 vni;
430 bool offloaded;
431 bool added_by_user;
432};
433
434#if IS_ENABLED(CONFIG_VXLAN)
435int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
436 struct switchdev_notifier_vxlan_fdb_info *fdb_info);
437int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
438 struct notifier_block *nb,
439 struct netlink_ext_ack *extack);
440void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);
441
442#else
443static inline int
444vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
445 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
446{
447 return -ENOENT;
448}
449
450static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
451 struct notifier_block *nb,
452 struct netlink_ext_ack *extack)
453{
454 return -EOPNOTSUPP;
455}
456
457static inline void
458vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
459{
460}
461#endif
462
463static inline void vxlan_flag_attr_error(int attrtype,
464 struct netlink_ext_ack *extack)
465{
466#define VXLAN_FLAG(flg) \
467 case IFLA_VXLAN_##flg: \
468 NL_SET_ERR_MSG_MOD(extack, \
469 "cannot change " #flg " flag"); \
470 break
471 switch (attrtype) {
472 VXLAN_FLAG(TTL_INHERIT);
473 VXLAN_FLAG(LEARNING);
474 VXLAN_FLAG(PROXY);
475 VXLAN_FLAG(RSC);
476 VXLAN_FLAG(L2MISS);
477 VXLAN_FLAG(L3MISS);
478 VXLAN_FLAG(COLLECT_METADATA);
479 VXLAN_FLAG(UDP_ZERO_CSUM6_TX);
480 VXLAN_FLAG(UDP_ZERO_CSUM6_RX);
481 VXLAN_FLAG(REMCSUM_TX);
482 VXLAN_FLAG(REMCSUM_RX);
483 VXLAN_FLAG(GBP);
484 VXLAN_FLAG(GPE);
485 VXLAN_FLAG(REMCSUM_NOPARTIAL);
486 default:
487 NL_SET_ERR_MSG_MOD(extack, \
488 "cannot change flag");
489 break;
490 }
491#undef VXLAN_FLAG
492}
493
494static inline bool vxlan_fdb_nh_path_select(struct nexthop *nh,
495 int hash,
496 struct vxlan_rdst *rdst)
497{
498 struct fib_nh_common *nhc;
499
500 nhc = nexthop_path_fdb_result(nh, hash);
501 if (unlikely(!nhc))
502 return false;
503
504 switch (nhc->nhc_gw_family) {
505 case AF_INET:
506 rdst->remote_ip.sin.sin_addr.s_addr = nhc->nhc_gw.ipv4;
507 rdst->remote_ip.sa.sa_family = AF_INET;
508 break;
509 case AF_INET6:
510 rdst->remote_ip.sin6.sin6_addr = nhc->nhc_gw.ipv6;
511 rdst->remote_ip.sa.sa_family = AF_INET6;
512 break;
513 }
514
515 return true;
516}
517
518#endif