xref: /kernel/linux/linux-5.10/include/net/vxlan.h (revision 8c2ecf20) 
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
330static inline int vxlan_headroom(u32 flags)
331{
332	/* VXLAN:     IP4/6 header + UDP + VXLAN + Ethernet header */
333	/* VXLAN-GPE: IP4/6 header + UDP + VXLAN */
334	return (flags & VXLAN_F_IPV6 ? sizeof(struct ipv6hdr) :
335				       sizeof(struct iphdr)) +
336	       sizeof(struct udphdr) + sizeof(struct vxlanhdr) +
337	       (flags & VXLAN_F_GPE ? 0 : ETH_HLEN);
338}
339
340static inline struct vxlanhdr *vxlan_hdr(struct sk_buff *skb)
341{
342	return (struct vxlanhdr *)(udp_hdr(skb) + 1);
343}
344
345static inline __be32 vxlan_vni(__be32 vni_field)
346{
347#if defined(__BIG_ENDIAN)
348	return (__force __be32)((__force u32)vni_field >> 8);
349#else
350	return (__force __be32)((__force u32)(vni_field & VXLAN_VNI_MASK) << 8);
351#endif
352}
353
354static inline __be32 vxlan_vni_field(__be32 vni)
355{
356#if defined(__BIG_ENDIAN)
357	return (__force __be32)((__force u32)vni << 8);
358#else
359	return (__force __be32)((__force u32)vni >> 8);
360#endif
361}
362
363static inline size_t vxlan_rco_start(__be32 vni_field)
364{
365	return be32_to_cpu(vni_field & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
366}
367
368static inline size_t vxlan_rco_offset(__be32 vni_field)
369{
370	return (vni_field & VXLAN_RCO_UDP) ?
371		offsetof(struct udphdr, check) :
372		offsetof(struct tcphdr, check);
373}
374
375static inline __be32 vxlan_compute_rco(unsigned int start, unsigned int offset)
376{
377	__be32 vni_field = cpu_to_be32(start >> VXLAN_RCO_SHIFT);
378
379	if (offset == offsetof(struct udphdr, check))
380		vni_field |= VXLAN_RCO_UDP;
381	return vni_field;
382}
383
384static inline unsigned short vxlan_get_sk_family(struct vxlan_sock *vs)
385{
386	return vs->sock->sk->sk_family;
387}
388
389#if IS_ENABLED(CONFIG_IPV6)
390
391static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
392{
393	if (ipa->sa.sa_family == AF_INET6)
394		return ipv6_addr_any(&ipa->sin6.sin6_addr);
395	else
396		return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
397}
398
399static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
400{
401	if (ipa->sa.sa_family == AF_INET6)
402		return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
403	else
404		return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
405}
406
407#else /* !IS_ENABLED(CONFIG_IPV6) */
408
409static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
410{
411	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
412}
413
414static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
415{
416	return ipv4_is_multicast(ipa->sin.sin_addr.s_addr);
417}
418
419#endif /* IS_ENABLED(CONFIG_IPV6) */
420
421static inline bool netif_is_vxlan(const struct net_device *dev)
422{
423	return dev->rtnl_link_ops &&
424	       !strcmp(dev->rtnl_link_ops->kind, "vxlan");
425}
426
427struct switchdev_notifier_vxlan_fdb_info {
428	struct switchdev_notifier_info info; /* must be first */
429	union vxlan_addr remote_ip;
430	__be16 remote_port;
431	__be32 remote_vni;
432	u32 remote_ifindex;
433	u8 eth_addr[ETH_ALEN];
434	__be32 vni;
435	bool offloaded;
436	bool added_by_user;
437};
438
439#if IS_ENABLED(CONFIG_VXLAN)
440int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
441		      struct switchdev_notifier_vxlan_fdb_info *fdb_info);
442int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
443		     struct notifier_block *nb,
444		     struct netlink_ext_ack *extack);
445void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni);
446
447#else
448static inline int
449vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
450		  struct switchdev_notifier_vxlan_fdb_info *fdb_info)
451{
452	return -ENOENT;
453}
454
455static inline int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
456				   struct notifier_block *nb,
457				   struct netlink_ext_ack *extack)
458{
459	return -EOPNOTSUPP;
460}
461
462static inline void
463vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
464{
465}
466#endif
467
468static inline void vxlan_flag_attr_error(int attrtype,
469					 struct netlink_ext_ack *extack)
470{
471#define VXLAN_FLAG(flg) \
472	case IFLA_VXLAN_##flg: \
473		NL_SET_ERR_MSG_MOD(extack, \
474				   "cannot change " #flg " flag"); \
475		break
476	switch (attrtype) {
477	VXLAN_FLAG(TTL_INHERIT);
478	VXLAN_FLAG(LEARNING);
479	VXLAN_FLAG(PROXY);
480	VXLAN_FLAG(RSC);
481	VXLAN_FLAG(L2MISS);
482	VXLAN_FLAG(L3MISS);
483	VXLAN_FLAG(COLLECT_METADATA);
484	VXLAN_FLAG(UDP_ZERO_CSUM6_TX);
485	VXLAN_FLAG(UDP_ZERO_CSUM6_RX);
486	VXLAN_FLAG(REMCSUM_TX);
487	VXLAN_FLAG(REMCSUM_RX);
488	VXLAN_FLAG(GBP);
489	VXLAN_FLAG(GPE);
490	VXLAN_FLAG(REMCSUM_NOPARTIAL);
491	default:
492		NL_SET_ERR_MSG_MOD(extack, \
493				   "cannot change flag");
494		break;
495	}
496#undef VXLAN_FLAG
497}
498
499static inline bool vxlan_fdb_nh_path_select(struct nexthop *nh,
500					    u32 hash,
501					    struct vxlan_rdst *rdst)
502{
503	struct fib_nh_common *nhc;
504
505	nhc = nexthop_path_fdb_result(nh, hash >> 1);
506	if (unlikely(!nhc))
507		return false;
508
509	switch (nhc->nhc_gw_family) {
510	case AF_INET:
511		rdst->remote_ip.sin.sin_addr.s_addr = nhc->nhc_gw.ipv4;
512		rdst->remote_ip.sa.sa_family = AF_INET;
513		break;
514	case AF_INET6:
515		rdst->remote_ip.sin6.sin6_addr = nhc->nhc_gw.ipv6;
516		rdst->remote_ip.sa.sa_family = AF_INET6;
517		break;
518	}
519
520	return true;
521}
522
523#endif
524