2 * Copyright (c) 2010, 2011 Nicira Networks.
3 * Distributed under the terms of the GNU GPL version 2.
5 * Significant portions of this file may be copied from parts of the Linux
6 * kernel, by Linus Torvalds and others.
12 #include <linux/version.h>
15 #include "openvswitch/tunnel.h"
19 * The absolute minimum fragment size. Note that there are many other
20 * definitions of the minimum MTU.
25 * One of these goes in struct tnl_ops and in tnl_find_port().
26 * These values are in the same namespace as other TNL_T_* values, so
27 * only the least significant 10 bits are available to define protocol
30 #define TNL_T_PROTO_GRE 0
31 #define TNL_T_PROTO_CAPWAP 1
33 /* These flags are only needed when calling tnl_find_port(). */
34 #define TNL_T_KEY_EXACT (1 << 10)
35 #define TNL_T_KEY_MATCH (1 << 11)
36 #define TNL_T_KEY_EITHER (TNL_T_KEY_EXACT | TNL_T_KEY_MATCH)
38 /* Private flags not exposed to userspace in this form. */
39 #define TNL_F_IN_KEY_MATCH (1 << 16) /* Store the key in tun_id to match in flow table. */
40 #define TNL_F_OUT_KEY_ACTION (1 << 17) /* Get the key from a SET_TUNNEL action. */
42 /* All public tunnel flags. */
43 #define TNL_F_PUBLIC (TNL_F_CSUM | TNL_F_TOS_INHERIT | TNL_F_TTL_INHERIT | \
44 TNL_F_DF_INHERIT | TNL_F_DF_DEFAULT | TNL_F_PMTUD | \
45 TNL_F_HDR_CACHE | TNL_F_IPSEC)
48 * struct port_lookup_key - Tunnel port key, used as hash table key.
49 * @in_key: Key to match on input, 0 for wildcard.
50 * @saddr: IPv4 source address to match, 0 to accept any source address.
51 * @daddr: IPv4 destination of tunnel.
52 * @tunnel_type: Set of TNL_T_* flags that define lookup.
54 struct port_lookup_key {
61 #define PORT_KEY_LEN (offsetof(struct port_lookup_key, tunnel_type) + \
62 FIELD_SIZEOF(struct port_lookup_key, tunnel_type))
65 * struct tnl_mutable_config - modifiable configuration for a tunnel.
66 * @key: Used as key for tunnel port. Configured via OVS_TUNNEL_ATTR_*
68 * @rcu: RCU callback head for deferred destruction.
69 * @seq: Sequence number for distinguishing configuration versions.
70 * @tunnel_hlen: Tunnel header length.
71 * @eth_addr: Source address for packets generated by tunnel itself
72 * (e.g. ICMP fragmentation needed messages).
73 * @out_key: Key to use on output, 0 if this tunnel has no fixed output key.
74 * @flags: TNL_F_* flags.
75 * @tos: IPv4 TOS value to use for tunnel, 0 if no fixed TOS.
76 * @ttl: IPv4 TTL value to use for tunnel, 0 if no fixed TTL.
78 struct tnl_mutable_config {
79 struct port_lookup_key key;
86 unsigned char eth_addr[ETH_ALEN];
88 /* Configured via OVS_TUNNEL_ATTR_* attributes. */
96 u32 tunnel_type; /* Put the TNL_T_PROTO_* type in here. */
97 u8 ipproto; /* The IP protocol for the tunnel. */
100 * Returns the length of the tunnel header that will be added in
101 * build_header() (i.e. excludes the IP header). Returns a negative
102 * error code if the configuration is invalid.
104 int (*hdr_len)(const struct tnl_mutable_config *);
107 * Builds the static portion of the tunnel header, which is stored in
108 * the header cache. In general the performance of this function is
109 * not too important as we try to only call it when building the cache
110 * so it is preferable to shift as much work as possible here. However,
111 * in some circumstances caching is disabled and this function will be
112 * called for every packet, so try not to make it too slow.
114 void (*build_header)(const struct vport *,
115 const struct tnl_mutable_config *, void *header);
118 * Updates the cached header of a packet to match the actual packet
119 * data. Typical things that might need to be updated are length,
120 * checksum, etc. The IP header will have already been updated and this
121 * is the final step before transmission. Returns a linked list of
122 * completed SKBs (multiple packets may be generated in the event
125 struct sk_buff *(*update_header)(const struct vport *,
126 const struct tnl_mutable_config *,
127 struct dst_entry *, struct sk_buff *);
130 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)
132 * On these kernels we have a fast mechanism to tell if the ARP cache for a
133 * particular destination has changed.
137 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27)
139 * On these kernels we have a fast mechanism to tell if the routing table
142 #define HAVE_RT_GENID
144 #if !defined(HAVE_HH_SEQ) || !defined(HAVE_RT_GENID)
145 /* If we can't detect all system changes directly we need to use a timeout. */
146 #define NEED_CACHE_TIMEOUT
151 int len; /* Length of data to be memcpy'd from cache. */
153 /* Sequence number of mutable->seq from which this cache was generated. */
154 unsigned mutable_seq;
158 * The sequence number from the seqlock protecting the hardware header
159 * cache (in the ARP cache). Since every write increments the counter
160 * this gives us an easy way to tell if it has changed.
165 #ifdef NEED_CACHE_TIMEOUT
167 * If we don't have direct mechanisms to detect all important changes in
168 * the system fall back to an expiration time. This expiration time
169 * can be relatively short since at high rates there will be millions of
170 * packets per second, so we'll still get plenty of benefit from the
171 * cache. Note that if something changes we may blackhole packets
172 * until the expiration time (depending on what changed and the kernel
173 * version we may be able to detect the change sooner). Expiration is
174 * expressed as a time in jiffies.
176 unsigned long expiration;
180 * The routing table entry that is the result of looking up the tunnel
181 * endpoints. It also contains a sequence number (called a generation
182 * ID) that can be compared to a global sequence to tell if the routing
183 * table has changed (and therefore there is a potential that this
184 * cached route has been invalidated).
189 * If the output device for tunnel traffic is an OVS internal device,
190 * the flow of that datapath. Since all tunnel traffic will have the
191 * same headers this allows us to cache the flow lookup. NULL if the
192 * output device is not OVS or if there is no flow installed.
194 struct sw_flow *flow;
196 /* The cached header follows after padding for alignment. */
201 struct hlist_node hash_node;
204 const struct tnl_ops *tnl_ops;
206 struct tnl_mutable_config __rcu *mutable;
209 * ID of last fragment sent (for tunnel protocols with direct support
210 * fragmentation). If the protocol relies on IP fragmentation then
211 * this is not needed.
215 spinlock_t cache_lock;
216 struct tnl_cache __rcu *cache; /* Protected by RCU/cache_lock. */
218 #ifdef NEED_CACHE_TIMEOUT
220 * If we must rely on expiration time to invalidate the cache, this is
221 * the interval. It is randomized within a range (defined by
222 * MAX_CACHE_EXP in tunnel.c) to avoid synchronized expirations caused
223 * by creation of a large number of tunnels at a one time.
225 unsigned long cache_exp_interval;
229 struct vport *tnl_create(const struct vport_parms *, const struct vport_ops *,
230 const struct tnl_ops *);
231 void tnl_destroy(struct vport *);
233 int tnl_set_options(struct vport *, struct nlattr *);
234 int tnl_get_options(const struct vport *, struct sk_buff *);
236 int tnl_set_addr(struct vport *vport, const unsigned char *addr);
237 const char *tnl_get_name(const struct vport *vport);
238 const unsigned char *tnl_get_addr(const struct vport *vport);
239 int tnl_send(struct vport *vport, struct sk_buff *skb);
240 void tnl_rcv(struct vport *vport, struct sk_buff *skb, u8 tos);
242 struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
244 const struct tnl_mutable_config **mutable);
245 bool tnl_frag_needed(struct vport *vport,
246 const struct tnl_mutable_config *mutable,
247 struct sk_buff *skb, unsigned int mtu, __be64 flow_key);
248 void tnl_free_linked_skbs(struct sk_buff *skb);
252 static inline struct tnl_vport *tnl_vport_priv(const struct vport *vport)
254 return vport_priv(vport);
258 #endif /* tunnel.h */