datapath/tunnel.h

   1 /*
   2  * Copyright (c) 2007-2012 Nicira, Inc.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of version 2 of the GNU General Public
   6  * License as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License
  14  * along with this program; if not, write to the Free Software
  15  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  16  * 02110-1301, USA
  17  */
  18
  19 #ifndef TUNNEL_H
  20 #define TUNNEL_H 1
  21
  22 #include <linux/version.h>
  23 #include <net/net_namespace.h>
  24 #include <net/netns/generic.h>
  25
  26 #include "flow.h"
  27 #include "openvswitch/tunnel.h"
  28 #include "vport.h"
  29
  30 /*
  31  * The absolute minimum fragment size.  Note that there are many other
  32  * definitions of the minimum MTU.
  33  */
  34 #define IP_MIN_MTU 68
  35
  36 /*
  37  * One of these goes in struct tnl_ops and in tnl_find_port().
  38  * These values are in the same namespace as other TNL_T_* values, so
  39  * only the least significant 10 bits are available to define protocol
  40  * identifiers.
  41  */
  42 #define TNL_T_PROTO_GRE         0
  43 #define TNL_T_PROTO_CAPWAP      1
  44
  45 /* These flags are only needed when calling tnl_find_port(). */
  46 #define TNL_T_KEY_EXACT         (1 << 10)
  47 #define TNL_T_KEY_MATCH         (1 << 11)
  48
  49 /* Private flags not exposed to userspace in this form. */
  50 #define TNL_F_IN_KEY_MATCH      (1 << 16) /* Store the key in tun_id to
  51                                            * match in flow table. */
  52 #define TNL_F_OUT_KEY_ACTION    (1 << 17) /* Get the key from a SET_TUNNEL
  53                                            * action. */
  54
  55 /* All public tunnel flags. */
  56 #define TNL_F_PUBLIC (TNL_F_CSUM | TNL_F_TOS_INHERIT | TNL_F_TTL_INHERIT | \
  57                       TNL_F_DF_INHERIT | TNL_F_DF_DEFAULT | TNL_F_PMTUD | \
  58                       TNL_F_HDR_CACHE | TNL_F_IPSEC)
  59
  60 /**
  61  * struct port_lookup_key - Tunnel port key, used as hash table key.
  62  * @in_key: Key to match on input, 0 for wildcard.
  63  * @net: Network namespace of the port.
  64  * @saddr: IPv4 source address to match, 0 to accept any source address.
  65  * @daddr: IPv4 destination of tunnel.
  66  * @tunnel_type: Set of TNL_T_* flags that define lookup.
  67  */
  68 struct port_lookup_key {
  69         __be64 in_key;
  70 #ifdef CONFIG_NET_NS
  71         struct net *net;
  72 #endif
  73         __be32 saddr;
  74         __be32 daddr;
  75         u32    tunnel_type;
  76 };
  77
  78 #define PORT_KEY_LEN    (offsetof(struct port_lookup_key, tunnel_type) + \
  79                          FIELD_SIZEOF(struct port_lookup_key, tunnel_type))
  80
  81 static inline struct net *port_key_get_net(const struct port_lookup_key *key)
  82 {
  83         return read_pnet(&key->net);
  84 }
  85
  86 static inline void port_key_set_net(struct port_lookup_key *key, struct net *net)
  87 {
  88         write_pnet(&key->net, net);
  89 }
  90
  91 /**
  92  * struct tnl_mutable_config - modifiable configuration for a tunnel.
  93  * @key: Used as key for tunnel port.  Configured via OVS_TUNNEL_ATTR_*
  94  * attributes.
  95  * @rcu: RCU callback head for deferred destruction.
  96  * @seq: Sequence number for distinguishing configuration versions.
  97  * @tunnel_hlen: Tunnel header length.
  98  * @eth_addr: Source address for packets generated by tunnel itself
  99  * (e.g. ICMP fragmentation needed messages).
 100  * @out_key: Key to use on output, 0 if this tunnel has no fixed output key.
 101  * @flags: TNL_F_* flags.
 102  * @tos: IPv4 TOS value to use for tunnel, 0 if no fixed TOS.
 103  * @ttl: IPv4 TTL value to use for tunnel, 0 if no fixed TTL.
 104  */
 105 struct tnl_mutable_config {
 106         struct port_lookup_key key;
 107         struct rcu_head rcu;
 108
 109         unsigned seq;
 110
 111         unsigned tunnel_hlen;
 112
 113         unsigned char eth_addr[ETH_ALEN];
 114
 115         /* Configured via OVS_TUNNEL_ATTR_* attributes. */
 116         __be64  out_key;
 117         u32     flags;
 118         u8      tos;
 119         u8      ttl;
 120
 121         /* Multicast configuration. */
 122         int     mlink;
 123 };
 124
 125 struct tnl_ops {
 126         u32 tunnel_type;        /* Put the TNL_T_PROTO_* type in here. */
 127         u8 ipproto;             /* The IP protocol for the tunnel. */
 128
 129         /*
 130          * Returns the length of the tunnel header that will be added in
 131          * build_header() (i.e. excludes the IP header).  Returns a negative
 132          * error code if the configuration is invalid.
 133          */
 134         int (*hdr_len)(const struct tnl_mutable_config *);
 135
 136         /*
 137          * Builds the static portion of the tunnel header, which is stored in
 138          * the header cache.  In general the performance of this function is
 139          * not too important as we try to only call it when building the cache
 140          * so it is preferable to shift as much work as possible here.  However,
 141          * in some circumstances caching is disabled and this function will be
 142          * called for every packet, so try not to make it too slow.
 143          */
 144         void (*build_header)(const struct vport *,
 145                              const struct tnl_mutable_config *, void *header);
 146
 147         /*
 148          * Updates the cached header of a packet to match the actual packet
 149          * data.  Typical things that might need to be updated are length,
 150          * checksum, etc.  The IP header will have already been updated and this
 151          * is the final step before transmission.  Returns a linked list of
 152          * completed SKBs (multiple packets may be generated in the event
 153          * of fragmentation).
 154          */
 155         struct sk_buff *(*update_header)(const struct vport *,
 156                                          const struct tnl_mutable_config *,
 157                                          struct dst_entry *, struct sk_buff *);
 158 };
 159
 160 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)
 161 /*
 162  * On these kernels we have a fast mechanism to tell if the ARP cache for a
 163  * particular destination has changed.
 164  */
 165 #define HAVE_HH_SEQ
 166 #endif
 167 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27)
 168 /*
 169  * On these kernels we have a fast mechanism to tell if the routing table
 170  * has changed.
 171  */
 172 #define HAVE_RT_GENID
 173 #endif
 174 #if !defined(HAVE_HH_SEQ) || !defined(HAVE_RT_GENID)
 175 /* If we can't detect all system changes directly we need to use a timeout. */
 176 #define NEED_CACHE_TIMEOUT
 177 #endif
 178 struct tnl_cache {
 179         struct rcu_head rcu;
 180
 181         int len;                /* Length of data to be memcpy'd from cache. */
 182         int hh_len;             /* Hardware hdr length, cached from hh_cache. */
 183
 184         /* Sequence number of mutable->seq from which this cache was
 185          * generated. */
 186         unsigned mutable_seq;
 187
 188 #ifdef HAVE_HH_SEQ
 189         /*
 190          * The sequence number from the seqlock protecting the hardware header
 191          * cache (in the ARP cache).  Since every write increments the counter
 192          * this gives us an easy way to tell if it has changed.
 193          */
 194         unsigned hh_seq;
 195 #endif
 196
 197 #ifdef NEED_CACHE_TIMEOUT
 198         /*
 199          * If we don't have direct mechanisms to detect all important changes in
 200          * the system fall back to an expiration time.  This expiration time
 201          * can be relatively short since at high rates there will be millions of
 202          * packets per second, so we'll still get plenty of benefit from the
 203          * cache.  Note that if something changes we may blackhole packets
 204          * until the expiration time (depending on what changed and the kernel
 205          * version we may be able to detect the change sooner).  Expiration is
 206          * expressed as a time in jiffies.
 207          */
 208         unsigned long expiration;
 209 #endif
 210
 211         /*
 212          * The routing table entry that is the result of looking up the tunnel
 213          * endpoints.  It also contains a sequence number (called a generation
 214          * ID) that can be compared to a global sequence to tell if the routing
 215          * table has changed (and therefore there is a potential that this
 216          * cached route has been invalidated).
 217          */
 218         struct rtable *rt;
 219
 220         /*
 221          * If the output device for tunnel traffic is an OVS internal device,
 222          * the flow of that datapath.  Since all tunnel traffic will have the
 223          * same headers this allows us to cache the flow lookup.  NULL if the
 224          * output device is not OVS or if there is no flow installed.
 225          */
 226         struct sw_flow *flow;
 227
 228         /* The cached header follows after padding for alignment. */
 229 };
 230
 231 struct tnl_vport {
 232         struct rcu_head rcu;
 233         struct hlist_node hash_node;
 234
 235         char name[IFNAMSIZ];
 236         const struct tnl_ops *tnl_ops;
 237
 238         struct tnl_mutable_config __rcu *mutable;
 239
 240         /*
 241          * ID of last fragment sent (for tunnel protocols with direct support
 242          * fragmentation).  If the protocol relies on IP fragmentation then
 243          * this is not needed.
 244          */
 245         atomic_t frag_id;
 246
 247         spinlock_t cache_lock;
 248         struct tnl_cache __rcu *cache;  /* Protected by RCU/cache_lock. */
 249
 250 #ifdef NEED_CACHE_TIMEOUT
 251         /*
 252          * If we must rely on expiration time to invalidate the cache, this is
 253          * the interval.  It is randomized within a range (defined by
 254          * MAX_CACHE_EXP in tunnel.c) to avoid synchronized expirations caused
 255          * by creation of a large number of tunnels at a one time.
 256          */
 257         unsigned long cache_exp_interval;
 258 #endif
 259 };
 260
 261 struct vport *ovs_tnl_create(const struct vport_parms *, const struct vport_ops *,
 262                              const struct tnl_ops *);
 263 void ovs_tnl_destroy(struct vport *);
 264
 265 int ovs_tnl_set_options(struct vport *, struct nlattr *);
 266 int ovs_tnl_get_options(const struct vport *, struct sk_buff *);
 267
 268 int ovs_tnl_set_addr(struct vport *vport, const unsigned char *addr);
 269 const char *ovs_tnl_get_name(const struct vport *vport);
 270 const unsigned char *ovs_tnl_get_addr(const struct vport *vport);
 271 int ovs_tnl_send(struct vport *vport, struct sk_buff *skb);
 272 void ovs_tnl_rcv(struct vport *vport, struct sk_buff *skb, u8 tos);
 273
 274 struct vport *ovs_tnl_find_port(struct net *net, __be32 saddr, __be32 daddr,
 275                                 __be64 key, int tunnel_type,
 276                                 const struct tnl_mutable_config **mutable);
 277 bool ovs_tnl_frag_needed(struct vport *vport,
 278                          const struct tnl_mutable_config *mutable,
 279                          struct sk_buff *skb, unsigned int mtu, __be64 flow_key);
 280 void ovs_tnl_free_linked_skbs(struct sk_buff *skb);
 281
 282 int ovs_tnl_init(void);
 283 void ovs_tnl_exit(void);
 284 static inline struct tnl_vport *tnl_vport_priv(const struct vport *vport)
 285 {
 286         return vport_priv(vport);
 287 }
 288
 289 #endif /* tunnel.h */