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"
20 * The absolute minimum fragment size. Note that there are many other
21 * definitions of the minimum MTU.
26 * One of these goes in struct tnl_ops and in tnl_find_port().
27 * These values are in the same namespace as other TNL_T_* values, so
28 * only the least significant 10 bits are available to define protocol
31 #define TNL_T_PROTO_GRE 0
32 #define TNL_T_PROTO_CAPWAP 1
34 /* These flags are only needed when calling tnl_find_port(). */
35 #define TNL_T_KEY_EXACT (1 << 10)
36 #define TNL_T_KEY_MATCH (1 << 11)
37 #define TNL_T_KEY_EITHER (TNL_T_KEY_EXACT | TNL_T_KEY_MATCH)
39 /* Private flags not exposed to userspace in this form. */
40 #define TNL_F_IN_KEY_MATCH (1 << 16) /* Store the key in tun_id to match in flow table. */
41 #define TNL_F_OUT_KEY_ACTION (1 << 17) /* Get the key from a SET_TUNNEL action. */
43 /* All public tunnel flags. */
44 #define TNL_F_PUBLIC (TNL_F_CSUM | TNL_F_TOS_INHERIT | TNL_F_TTL_INHERIT | \
45 TNL_F_DF_INHERIT | TNL_F_DF_DEFAULT | TNL_F_PMTUD | \
46 TNL_F_HDR_CACHE | TNL_F_IPSEC)
49 * struct tnl_mutable_config - modifiable configuration for a tunnel.
50 * @rcu: RCU callback head for deferred destruction.
51 * @seq: Sequence number for distinguishing configuration versions.
52 * @tunnel_type: Set of TNL_T_* flags that define lookup.
53 * @tunnel_hlen: Tunnel header length.
54 * @eth_addr: Source address for packets generated by tunnel itself
55 * (e.g. ICMP fragmentation needed messages).
56 * @in_key: Key to match on input, 0 for wildcard.
57 * @out_key: Key to use on output, 0 if this tunnel has no fixed output key.
58 * @flags: TNL_F_* flags.
59 * @saddr: IPv4 source address to match, 0 to accept any source address.
60 * @daddr: IPv4 destination of tunnel.
61 * @tos: IPv4 TOS value to use for tunnel, 0 if no fixed TOS.
62 * @ttl: IPv4 TTL value to use for tunnel, 0 if no fixed TTL.
64 struct tnl_mutable_config {
72 unsigned char eth_addr[ETH_ALEN];
74 /* Configured via ODP_TUNNEL_ATTR_* attributes. */
85 u32 tunnel_type; /* Put the TNL_T_PROTO_* type in here. */
86 u8 ipproto; /* The IP protocol for the tunnel. */
89 * Returns the length of the tunnel header that will be added in
90 * build_header() (i.e. excludes the IP header). Returns a negative
91 * error code if the configuration is invalid.
93 int (*hdr_len)(const struct tnl_mutable_config *);
96 * Builds the static portion of the tunnel header, which is stored in
97 * the header cache. In general the performance of this function is
98 * not too important as we try to only call it when building the cache
99 * so it is preferable to shift as much work as possible here. However,
100 * in some circumstances caching is disabled and this function will be
101 * called for every packet, so try not to make it too slow.
103 void (*build_header)(const struct vport *,
104 const struct tnl_mutable_config *, void *header);
107 * Updates the cached header of a packet to match the actual packet
108 * data. Typical things that might need to be updated are length,
109 * checksum, etc. The IP header will have already been updated and this
110 * is the final step before transmission. Returns a linked list of
111 * completed SKBs (multiple packets may be generated in the event
114 struct sk_buff *(*update_header)(const struct vport *,
115 const struct tnl_mutable_config *,
116 struct dst_entry *, struct sk_buff *);
119 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)
121 * On these kernels we have a fast mechanism to tell if the ARP cache for a
122 * particular destination has changed.
126 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27)
128 * On these kernels we have a fast mechanism to tell if the routing table
131 #define HAVE_RT_GENID
133 #if !defined(HAVE_HH_SEQ) || !defined(HAVE_RT_GENID)
134 /* If we can't detect all system changes directly we need to use a timeout. */
135 #define NEED_CACHE_TIMEOUT
140 int len; /* Length of data to be memcpy'd from cache. */
142 /* Sequence number of mutable->seq from which this cache was generated. */
143 unsigned mutable_seq;
147 * The sequence number from the seqlock protecting the hardware header
148 * cache (in the ARP cache). Since every write increments the counter
149 * this gives us an easy way to tell if it has changed.
154 #ifdef NEED_CACHE_TIMEOUT
156 * If we don't have direct mechanisms to detect all important changes in
157 * the system fall back to an expiration time. This expiration time
158 * can be relatively short since at high rates there will be millions of
159 * packets per second, so we'll still get plenty of benefit from the
160 * cache. Note that if something changes we may blackhole packets
161 * until the expiration time (depending on what changed and the kernel
162 * version we may be able to detect the change sooner). Expiration is
163 * expressed as a time in jiffies.
165 unsigned long expiration;
169 * The routing table entry that is the result of looking up the tunnel
170 * endpoints. It also contains a sequence number (called a generation
171 * ID) that can be compared to a global sequence to tell if the routing
172 * table has changed (and therefore there is a potential that this
173 * cached route has been invalidated).
178 * If the output device for tunnel traffic is an OVS internal device,
179 * the flow of that datapath. Since all tunnel traffic will have the
180 * same headers this allows us to cache the flow lookup. NULL if the
181 * output device is not OVS or if there is no flow installed.
183 struct sw_flow *flow;
185 /* The cached header follows after padding for alignment. */
190 struct tbl_node tbl_node;
193 const struct tnl_ops *tnl_ops;
195 struct tnl_mutable_config __rcu *mutable;
198 * ID of last fragment sent (for tunnel protocols with direct support
199 * fragmentation). If the protocol relies on IP fragmentation then
200 * this is not needed.
204 spinlock_t cache_lock;
205 struct tnl_cache __rcu *cache; /* Protected by RCU/cache_lock. */
207 #ifdef NEED_CACHE_TIMEOUT
209 * If we must rely on expiration time to invalidate the cache, this is
210 * the interval. It is randomized within a range (defined by
211 * MAX_CACHE_EXP in tunnel.c) to avoid synchronized expirations caused
212 * by creation of a large number of tunnels at a one time.
214 unsigned long cache_exp_interval;
218 struct vport *tnl_create(const struct vport_parms *, const struct vport_ops *,
219 const struct tnl_ops *);
220 int tnl_destroy(struct vport *);
222 int tnl_set_options(struct vport *, struct nlattr *);
223 int tnl_get_options(const struct vport *, struct sk_buff *);
225 int tnl_set_addr(struct vport *vport, const unsigned char *addr);
226 const char *tnl_get_name(const struct vport *vport);
227 const unsigned char *tnl_get_addr(const struct vport *vport);
228 int tnl_send(struct vport *vport, struct sk_buff *skb);
229 void tnl_rcv(struct vport *vport, struct sk_buff *skb, u8 tos);
231 struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
233 const struct tnl_mutable_config **mutable);
234 bool tnl_frag_needed(struct vport *vport,
235 const struct tnl_mutable_config *mutable,
236 struct sk_buff *skb, unsigned int mtu, __be64 flow_key);
237 void tnl_free_linked_skbs(struct sk_buff *skb);
239 static inline struct tnl_vport *tnl_vport_priv(const struct vport *vport)
241 return vport_priv(vport);
245 #endif /* tunnel.h */