]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
8024e028 TH |
2 | #ifndef __NET_UDP_TUNNEL_H |
3 | #define __NET_UDP_TUNNEL_H | |
4 | ||
6a93cc90 AZ |
5 | #include <net/ip_tunnels.h> |
6 | #include <net/udp.h> | |
7 | ||
8 | #if IS_ENABLED(CONFIG_IPV6) | |
9 | #include <net/ipv6.h> | |
3616d08b | 10 | #include <net/ipv6_stubs.h> |
6a93cc90 AZ |
11 | #endif |
12 | ||
8024e028 TH |
13 | struct udp_port_cfg { |
14 | u8 family; | |
15 | ||
16 | /* Used only for kernel-created sockets */ | |
17 | union { | |
18 | struct in_addr local_ip; | |
19 | #if IS_ENABLED(CONFIG_IPV6) | |
20 | struct in6_addr local_ip6; | |
21 | #endif | |
22 | }; | |
23 | ||
24 | union { | |
25 | struct in_addr peer_ip; | |
26 | #if IS_ENABLED(CONFIG_IPV6) | |
27 | struct in6_addr peer_ip6; | |
28 | #endif | |
29 | }; | |
30 | ||
31 | __be16 local_udp_port; | |
32 | __be16 peer_udp_port; | |
da5095d0 | 33 | int bind_ifindex; |
8024e028 TH |
34 | unsigned int use_udp_checksums:1, |
35 | use_udp6_tx_checksums:1, | |
a43a9ef6 JB |
36 | use_udp6_rx_checksums:1, |
37 | ipv6_v6only:1; | |
8024e028 TH |
38 | }; |
39 | ||
fd384412 AZ |
40 | int udp_sock_create4(struct net *net, struct udp_port_cfg *cfg, |
41 | struct socket **sockp); | |
42 | ||
43 | #if IS_ENABLED(CONFIG_IPV6) | |
44 | int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg, | |
45 | struct socket **sockp); | |
46 | #else | |
47 | static inline int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg, | |
48 | struct socket **sockp) | |
49 | { | |
50 | return 0; | |
51 | } | |
52 | #endif | |
53 | ||
54 | static inline int udp_sock_create(struct net *net, | |
55 | struct udp_port_cfg *cfg, | |
56 | struct socket **sockp) | |
57 | { | |
58 | if (cfg->family == AF_INET) | |
59 | return udp_sock_create4(net, cfg, sockp); | |
60 | ||
61 | if (cfg->family == AF_INET6) | |
62 | return udp_sock_create6(net, cfg, sockp); | |
63 | ||
64 | return -EPFNOSUPPORT; | |
65 | } | |
8024e028 | 66 | |
6a93cc90 | 67 | typedef int (*udp_tunnel_encap_rcv_t)(struct sock *sk, struct sk_buff *skb); |
a36e185e SB |
68 | typedef int (*udp_tunnel_encap_err_lookup_t)(struct sock *sk, |
69 | struct sk_buff *skb); | |
6a93cc90 | 70 | typedef void (*udp_tunnel_encap_destroy_t)(struct sock *sk); |
d4546c25 DM |
71 | typedef struct sk_buff *(*udp_tunnel_gro_receive_t)(struct sock *sk, |
72 | struct list_head *head, | |
73 | struct sk_buff *skb); | |
38fd2af2 TH |
74 | typedef int (*udp_tunnel_gro_complete_t)(struct sock *sk, struct sk_buff *skb, |
75 | int nhoff); | |
6a93cc90 AZ |
76 | |
77 | struct udp_tunnel_sock_cfg { | |
78 | void *sk_user_data; /* user data used by encap_rcv call back */ | |
79 | /* Used for setting up udp_sock fields, see udp.h for details */ | |
80 | __u8 encap_type; | |
81 | udp_tunnel_encap_rcv_t encap_rcv; | |
a36e185e | 82 | udp_tunnel_encap_err_lookup_t encap_err_lookup; |
6a93cc90 | 83 | udp_tunnel_encap_destroy_t encap_destroy; |
38fd2af2 TH |
84 | udp_tunnel_gro_receive_t gro_receive; |
85 | udp_tunnel_gro_complete_t gro_complete; | |
6a93cc90 AZ |
86 | }; |
87 | ||
88 | /* Setup the given (UDP) sock to receive UDP encapsulated packets */ | |
89 | void setup_udp_tunnel_sock(struct net *net, struct socket *sock, | |
90 | struct udp_tunnel_sock_cfg *sock_cfg); | |
91 | ||
e7b3db5e AD |
92 | /* -- List of parsable UDP tunnel types -- |
93 | * | |
94 | * Adding to this list will result in serious debate. The main issue is | |
95 | * that this list is essentially a list of workarounds for either poorly | |
96 | * designed tunnels, or poorly designed device offloads. | |
97 | * | |
98 | * The parsing supported via these types should really be used for Rx | |
99 | * traffic only as the network stack will have already inserted offsets for | |
100 | * the location of the headers in the skb. In addition any ports that are | |
101 | * pushed should be kept within the namespace without leaking to other | |
102 | * devices such as VFs or other ports on the same device. | |
103 | * | |
104 | * It is strongly encouraged to use CHECKSUM_COMPLETE for Rx to avoid the | |
105 | * need to use this for Rx checksum offload. It should not be necessary to | |
106 | * call this function to perform Tx offloads on outgoing traffic. | |
107 | */ | |
108 | enum udp_parsable_tunnel_type { | |
84a4160e JK |
109 | UDP_TUNNEL_TYPE_VXLAN = BIT(0), /* RFC 7348 */ |
110 | UDP_TUNNEL_TYPE_GENEVE = BIT(1), /* draft-ietf-nvo3-geneve */ | |
111 | UDP_TUNNEL_TYPE_VXLAN_GPE = BIT(2), /* draft-ietf-nvo3-vxlan-gpe */ | |
e7b3db5e AD |
112 | }; |
113 | ||
114 | struct udp_tunnel_info { | |
115 | unsigned short type; | |
116 | sa_family_t sa_family; | |
117 | __be16 port; | |
cc4e3835 | 118 | u8 hw_priv; |
e7b3db5e AD |
119 | }; |
120 | ||
121 | /* Notify network devices of offloadable types */ | |
122 | void udp_tunnel_push_rx_port(struct net_device *dev, struct socket *sock, | |
123 | unsigned short type); | |
296d8ee3 SD |
124 | void udp_tunnel_drop_rx_port(struct net_device *dev, struct socket *sock, |
125 | unsigned short type); | |
e7b3db5e AD |
126 | void udp_tunnel_notify_add_rx_port(struct socket *sock, unsigned short type); |
127 | void udp_tunnel_notify_del_rx_port(struct socket *sock, unsigned short type); | |
128 | ||
7c46a640 AD |
129 | static inline void udp_tunnel_get_rx_info(struct net_device *dev) |
130 | { | |
131 | ASSERT_RTNL(); | |
132 | call_netdevice_notifiers(NETDEV_UDP_TUNNEL_PUSH_INFO, dev); | |
133 | } | |
134 | ||
296d8ee3 SD |
135 | static inline void udp_tunnel_drop_rx_info(struct net_device *dev) |
136 | { | |
137 | ASSERT_RTNL(); | |
138 | call_netdevice_notifiers(NETDEV_UDP_TUNNEL_DROP_INFO, dev); | |
139 | } | |
140 | ||
6a93cc90 | 141 | /* Transmit the skb using UDP encapsulation. */ |
039f5062 PS |
142 | void udp_tunnel_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb, |
143 | __be32 src, __be32 dst, __u8 tos, __u8 ttl, | |
144 | __be16 df, __be16 src_port, __be16 dst_port, | |
145 | bool xnet, bool nocheck); | |
6a93cc90 | 146 | |
79b16aad DM |
147 | int udp_tunnel6_xmit_skb(struct dst_entry *dst, struct sock *sk, |
148 | struct sk_buff *skb, | |
d998f8ef TH |
149 | struct net_device *dev, struct in6_addr *saddr, |
150 | struct in6_addr *daddr, | |
13461144 DB |
151 | __u8 prio, __u8 ttl, __be32 label, |
152 | __be16 src_port, __be16 dst_port, bool nocheck); | |
6a93cc90 AZ |
153 | |
154 | void udp_tunnel_sock_release(struct socket *sock); | |
155 | ||
c29a70d2 PS |
156 | struct metadata_dst *udp_tun_rx_dst(struct sk_buff *skb, unsigned short family, |
157 | __be16 flags, __be64 tunnel_id, | |
158 | int md_size); | |
159 | ||
86a98057 | 160 | #ifdef CONFIG_INET |
aed069df | 161 | static inline int udp_tunnel_handle_offloads(struct sk_buff *skb, bool udp_csum) |
6a93cc90 AZ |
162 | { |
163 | int type = udp_csum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; | |
164 | ||
6fa79666 | 165 | return iptunnel_handle_offloads(skb, type); |
6a93cc90 | 166 | } |
86a98057 | 167 | #endif |
6a93cc90 AZ |
168 | |
169 | static inline void udp_tunnel_encap_enable(struct socket *sock) | |
170 | { | |
60fb9567 PA |
171 | struct udp_sock *up = udp_sk(sock->sk); |
172 | ||
173 | if (up->encap_enabled) | |
174 | return; | |
175 | ||
176 | up->encap_enabled = 1; | |
6a93cc90 AZ |
177 | #if IS_ENABLED(CONFIG_IPV6) |
178 | if (sock->sk->sk_family == PF_INET6) | |
179 | ipv6_stub->udpv6_encap_enable(); | |
180 | else | |
181 | #endif | |
182 | udp_encap_enable(); | |
183 | } | |
184 | ||
cc4e3835 JK |
185 | #define UDP_TUNNEL_NIC_MAX_TABLES 4 |
186 | ||
187 | enum udp_tunnel_nic_info_flags { | |
188 | /* Device callbacks may sleep */ | |
189 | UDP_TUNNEL_NIC_INFO_MAY_SLEEP = BIT(0), | |
190 | /* Device only supports offloads when it's open, all ports | |
191 | * will be removed before close and re-added after open. | |
192 | */ | |
193 | UDP_TUNNEL_NIC_INFO_OPEN_ONLY = BIT(1), | |
194 | /* Device supports only IPv4 tunnels */ | |
195 | UDP_TUNNEL_NIC_INFO_IPV4_ONLY = BIT(2), | |
966e5059 JK |
196 | /* Device has hard-coded the IANA VXLAN port (4789) as VXLAN. |
197 | * This port must not be counted towards n_entries of any table. | |
198 | * Driver will not receive any callback associated with port 4789. | |
199 | */ | |
200 | UDP_TUNNEL_NIC_INFO_STATIC_IANA_VXLAN = BIT(3), | |
cc4e3835 JK |
201 | }; |
202 | ||
74cc6d18 JK |
203 | struct udp_tunnel_nic; |
204 | ||
205 | #define UDP_TUNNEL_NIC_MAX_SHARING_DEVICES (U16_MAX / 2) | |
206 | ||
207 | struct udp_tunnel_nic_shared { | |
208 | struct udp_tunnel_nic *udp_tunnel_nic_info; | |
209 | ||
210 | struct list_head devices; | |
211 | }; | |
212 | ||
213 | struct udp_tunnel_nic_shared_node { | |
214 | struct net_device *dev; | |
215 | struct list_head list; | |
216 | }; | |
217 | ||
cc4e3835 JK |
218 | /** |
219 | * struct udp_tunnel_nic_info - driver UDP tunnel offload information | |
220 | * @set_port: callback for adding a new port | |
221 | * @unset_port: callback for removing a port | |
222 | * @sync_table: callback for syncing the entire port table at once | |
74cc6d18 | 223 | * @shared: reference to device global state (optional) |
cc4e3835 JK |
224 | * @flags: device flags from enum udp_tunnel_nic_info_flags |
225 | * @tables: UDP port tables this device has | |
226 | * @tables.n_entries: number of entries in this table | |
227 | * @tables.tunnel_types: types of tunnels this table accepts | |
228 | * | |
229 | * Drivers are expected to provide either @set_port and @unset_port callbacks | |
230 | * or the @sync_table callback. Callbacks are invoked with rtnl lock held. | |
231 | * | |
74cc6d18 JK |
232 | * Devices which (misguidedly) share the UDP tunnel port table across multiple |
233 | * netdevs should allocate an instance of struct udp_tunnel_nic_shared and | |
234 | * point @shared at it. | |
235 | * There must never be more than %UDP_TUNNEL_NIC_MAX_SHARING_DEVICES devices | |
236 | * sharing a table. | |
237 | * | |
cc4e3835 JK |
238 | * Known limitations: |
239 | * - UDP tunnel port notifications are fundamentally best-effort - | |
240 | * it is likely the driver will both see skbs which use a UDP tunnel port, | |
241 | * while not being a tunneled skb, and tunnel skbs from other ports - | |
242 | * drivers should only use these ports for non-critical RX-side offloads, | |
243 | * e.g. the checksum offload; | |
244 | * - none of the devices care about the socket family at present, so we don't | |
245 | * track it. Please extend this code if you care. | |
246 | */ | |
247 | struct udp_tunnel_nic_info { | |
248 | /* one-by-one */ | |
249 | int (*set_port)(struct net_device *dev, | |
250 | unsigned int table, unsigned int entry, | |
251 | struct udp_tunnel_info *ti); | |
252 | int (*unset_port)(struct net_device *dev, | |
253 | unsigned int table, unsigned int entry, | |
254 | struct udp_tunnel_info *ti); | |
255 | ||
256 | /* all at once */ | |
257 | int (*sync_table)(struct net_device *dev, unsigned int table); | |
258 | ||
74cc6d18 JK |
259 | struct udp_tunnel_nic_shared *shared; |
260 | ||
cc4e3835 JK |
261 | unsigned int flags; |
262 | ||
263 | struct udp_tunnel_nic_table_info { | |
264 | unsigned int n_entries; | |
265 | unsigned int tunnel_types; | |
266 | } tables[UDP_TUNNEL_NIC_MAX_TABLES]; | |
267 | }; | |
268 | ||
269 | /* UDP tunnel module dependencies | |
270 | * | |
271 | * Tunnel drivers are expected to have a hard dependency on the udp_tunnel | |
272 | * module. NIC drivers are not, they just attach their | |
273 | * struct udp_tunnel_nic_info to the netdev and wait for callbacks to come. | |
274 | * Loading a tunnel driver will cause the udp_tunnel module to be loaded | |
275 | * and only then will all the required state structures be allocated. | |
276 | * Since we want a weak dependency from the drivers and the core to udp_tunnel | |
277 | * we call things through the following stubs. | |
278 | */ | |
279 | struct udp_tunnel_nic_ops { | |
280 | void (*get_port)(struct net_device *dev, unsigned int table, | |
281 | unsigned int idx, struct udp_tunnel_info *ti); | |
282 | void (*set_port_priv)(struct net_device *dev, unsigned int table, | |
283 | unsigned int idx, u8 priv); | |
284 | void (*add_port)(struct net_device *dev, struct udp_tunnel_info *ti); | |
285 | void (*del_port)(struct net_device *dev, struct udp_tunnel_info *ti); | |
286 | void (*reset_ntf)(struct net_device *dev); | |
c7d759eb JK |
287 | |
288 | size_t (*dump_size)(struct net_device *dev, unsigned int table); | |
289 | int (*dump_write)(struct net_device *dev, unsigned int table, | |
290 | struct sk_buff *skb); | |
cc4e3835 JK |
291 | }; |
292 | ||
293 | #ifdef CONFIG_INET | |
294 | extern const struct udp_tunnel_nic_ops *udp_tunnel_nic_ops; | |
295 | #else | |
296 | #define udp_tunnel_nic_ops ((struct udp_tunnel_nic_ops *)NULL) | |
297 | #endif | |
298 | ||
299 | static inline void | |
300 | udp_tunnel_nic_get_port(struct net_device *dev, unsigned int table, | |
301 | unsigned int idx, struct udp_tunnel_info *ti) | |
302 | { | |
303 | /* This helper is used from .sync_table, we indicate empty entries | |
304 | * by zero'ed @ti. Drivers which need to know the details of a port | |
305 | * when it gets deleted should use the .set_port / .unset_port | |
306 | * callbacks. | |
307 | * Zero out here, otherwise !CONFIG_INET causes uninitilized warnings. | |
308 | */ | |
309 | memset(ti, 0, sizeof(*ti)); | |
310 | ||
311 | if (udp_tunnel_nic_ops) | |
312 | udp_tunnel_nic_ops->get_port(dev, table, idx, ti); | |
313 | } | |
314 | ||
315 | static inline void | |
316 | udp_tunnel_nic_set_port_priv(struct net_device *dev, unsigned int table, | |
317 | unsigned int idx, u8 priv) | |
318 | { | |
319 | if (udp_tunnel_nic_ops) | |
320 | udp_tunnel_nic_ops->set_port_priv(dev, table, idx, priv); | |
321 | } | |
322 | ||
323 | static inline void | |
324 | udp_tunnel_nic_add_port(struct net_device *dev, struct udp_tunnel_info *ti) | |
325 | { | |
326 | if (udp_tunnel_nic_ops) | |
327 | udp_tunnel_nic_ops->add_port(dev, ti); | |
328 | } | |
329 | ||
330 | static inline void | |
331 | udp_tunnel_nic_del_port(struct net_device *dev, struct udp_tunnel_info *ti) | |
332 | { | |
333 | if (udp_tunnel_nic_ops) | |
334 | udp_tunnel_nic_ops->del_port(dev, ti); | |
335 | } | |
336 | ||
337 | /** | |
338 | * udp_tunnel_nic_reset_ntf() - device-originating reset notification | |
339 | * @dev: network interface device structure | |
340 | * | |
341 | * Called by the driver to inform the core that the entire UDP tunnel port | |
342 | * state has been lost, usually due to device reset. Core will assume device | |
343 | * forgot all the ports and issue .set_port and .sync_table callbacks as | |
344 | * necessary. | |
345 | * | |
346 | * This function must be called with rtnl lock held, and will issue all | |
347 | * the callbacks before returning. | |
348 | */ | |
349 | static inline void udp_tunnel_nic_reset_ntf(struct net_device *dev) | |
350 | { | |
351 | if (udp_tunnel_nic_ops) | |
352 | udp_tunnel_nic_ops->reset_ntf(dev); | |
353 | } | |
c7d759eb JK |
354 | |
355 | static inline size_t | |
356 | udp_tunnel_nic_dump_size(struct net_device *dev, unsigned int table) | |
357 | { | |
358 | if (!udp_tunnel_nic_ops) | |
359 | return 0; | |
360 | return udp_tunnel_nic_ops->dump_size(dev, table); | |
361 | } | |
362 | ||
363 | static inline int | |
364 | udp_tunnel_nic_dump_write(struct net_device *dev, unsigned int table, | |
365 | struct sk_buff *skb) | |
366 | { | |
367 | if (!udp_tunnel_nic_ops) | |
368 | return 0; | |
369 | return udp_tunnel_nic_ops->dump_write(dev, table, skb); | |
370 | } | |
8024e028 | 371 | #endif |