nest/route.h

   1 /*
   2  *      BIRD Internet Routing Daemon -- Routing Table
   3  *
   4  *      (c) 1998--2000 Martin Mares <mj@ucw.cz>
   5  *
   6  *      Can be freely distributed and used under the terms of the GNU GPL.
   7  */
   8
   9 #ifndef _BIRD_ROUTE_H_
  10 #define _BIRD_ROUTE_H_
  11
  12 #include "lib/lists.h"
  13 #include "lib/bitmap.h"
  14 #include "lib/resource.h"
  15 #include "lib/net.h"
  16
  17 struct ea_list;
  18 struct protocol;
  19 struct proto;
  20 struct rte_src;
  21 struct symbol;
  22 struct timer;
  23 struct filter;
  24 struct cli;
  25
  26 /*
  27  *      Generic data structure for storing network prefixes. Also used
  28  *      for the master routing table. Currently implemented as a hash
  29  *      table.
  30  *
  31  *      Available operations:
  32  *              - insertion of new entry
  33  *              - deletion of entry
  34  *              - searching for entry by network prefix
  35  *              - asynchronous retrieval of fib contents
  36  */
  37
  38 struct fib_node {
  39   struct fib_node *next;                /* Next in hash chain */
  40   struct fib_iterator *readers;         /* List of readers of this node */
  41   net_addr addr[0];
  42 };
  43
  44 struct fib_iterator {                   /* See lib/slists.h for an explanation */
  45   struct fib_iterator *prev, *next;     /* Must be synced with struct fib_node! */
  46   byte efef;                            /* 0xff to distinguish between iterator and node */
  47   byte pad[3];
  48   struct fib_node *node;                /* Or NULL if freshly merged */
  49   uint hash;
  50 };
  51
  52 typedef void (*fib_init_fn)(void *);
  53
  54 struct fib {
  55   pool *fib_pool;                       /* Pool holding all our data */
  56   slab *fib_slab;                       /* Slab holding all fib nodes */
  57   struct fib_node **hash_table;         /* Node hash table */
  58   uint hash_size;                       /* Number of hash table entries (a power of two) */
  59   uint hash_order;                      /* Binary logarithm of hash_size */
  60   uint hash_shift;                      /* 32 - hash_order */
  61   uint addr_type;                       /* Type of address data stored in fib (NET_*) */
  62   uint node_size;                       /* FIB node size, 0 for nonuniform */
  63   uint node_offset;                     /* Offset of fib_node struct inside of user data */
  64   uint entries;                         /* Number of entries */
  65   uint entries_min, entries_max;        /* Entry count limits (else start rehashing) */
  66   fib_init_fn init;                     /* Constructor */
  67 };
  68
  69 static inline void * fib_node_to_user(struct fib *f, struct fib_node *e)
  70 { return e ? (void *) ((char *) e - f->node_offset) : NULL; }
  71
  72 static inline struct fib_node * fib_user_to_node(struct fib *f, void *e)
  73 { return e ? (void *) ((char *) e + f->node_offset) : NULL; }
  74
  75 void fib_init(struct fib *f, pool *p, uint addr_type, uint node_size, uint node_offset, uint hash_order, fib_init_fn init);
  76 void *fib_find(struct fib *, const net_addr *); /* Find or return NULL if doesn't exist */
  77 void *fib_get_chain(struct fib *f, const net_addr *a); /* Find first node in linked list from hash table */
  78 void *fib_get(struct fib *, const net_addr *);  /* Find or create new if nonexistent */
  79 void *fib_route(struct fib *, const net_addr *); /* Longest-match routing lookup */
  80 void fib_delete(struct fib *, void *);  /* Remove fib entry */
  81 void fib_free(struct fib *);            /* Destroy the fib */
  82 void fib_check(struct fib *);           /* Consistency check for debugging */
  83
  84 void fit_init(struct fib_iterator *, struct fib *); /* Internal functions, don't call */
  85 struct fib_node *fit_get(struct fib *, struct fib_iterator *);
  86 void fit_put(struct fib_iterator *, struct fib_node *);
  87 void fit_put_next(struct fib *f, struct fib_iterator *i, struct fib_node *n, uint hpos);
  88 void fit_put_end(struct fib_iterator *i);
  89 void fit_copy(struct fib *f, struct fib_iterator *dst, struct fib_iterator *src);
  90
  91
  92 #define FIB_WALK(fib, type, z) do {                             \
  93         struct fib_node *fn_, **ff_ = (fib)->hash_table;        \
  94         uint count_ = (fib)->hash_size;                         \
  95         type *z;                                                \
  96         while (count_--)                                        \
  97           for (fn_ = *ff_++; z = fib_node_to_user(fib, fn_); fn_=fn_->next)
  98
  99 #define FIB_WALK_END } while (0)
 100
 101 #define FIB_ITERATE_INIT(it, fib) fit_init(it, fib)
 102
 103 #define FIB_ITERATE_START(fib, it, type, z) do {                \
 104         struct fib_node *fn_ = fit_get(fib, it);                \
 105         uint count_ = (fib)->hash_size;                         \
 106         uint hpos_ = (it)->hash;                                \
 107         type *z;                                                \
 108         for(;;) {                                               \
 109           if (!fn_)                                             \
 110             {                                                   \
 111                if (++hpos_ >= count_)                           \
 112                  break;                                         \
 113                fn_ = (fib)->hash_table[hpos_];                  \
 114                continue;                                        \
 115             }                                                   \
 116           z = fib_node_to_user(fib, fn_);
 117
 118 #define FIB_ITERATE_END fn_ = fn_->next; } } while(0)
 119
 120 #define FIB_ITERATE_PUT(it) fit_put(it, fn_)
 121
 122 #define FIB_ITERATE_PUT_NEXT(it, fib) fit_put_next(fib, it, fn_, hpos_)
 123
 124 #define FIB_ITERATE_PUT_END(it) fit_put_end(it)
 125
 126 #define FIB_ITERATE_UNLINK(it, fib) fit_get(fib, it)
 127
 128 #define FIB_ITERATE_COPY(dst, src, fib) fit_copy(fib, dst, src)
 129
 130
 131 /*
 132  *      Master Routing Tables. Generally speaking, each of them contains a FIB
 133  *      with each entry pointing to a list of route entries representing routes
 134  *      to given network (with the selected one at the head).
 135  *
 136  *      Each of the RTE's contains variable data (the preference and protocol-dependent
 137  *      metrics) and a pointer to a route attribute block common for many routes).
 138  *
 139  *      It's guaranteed that there is at most one RTE for every (prefix,proto) pair.
 140  */
 141
 142 struct rtable_config {
 143   node n;
 144   char *name;
 145   struct rtable *table;
 146   struct proto_config *krt_attached;    /* Kernel syncer attached to this table */
 147   uint addr_type;                       /* Type of address data stored in table (NET_*) */
 148   int gc_max_ops;                       /* Maximum number of operations before GC is run */
 149   int gc_min_time;                      /* Minimum time between two consecutive GC runs */
 150   byte sorted;                          /* Routes of network are sorted according to rte_better() */
 151   byte internal;                        /* Internal table of a protocol */
 152   btime min_settle_time;                /* Minimum settle time for notifications */
 153   btime max_settle_time;                /* Maximum settle time for notifications */
 154 };
 155
 156 typedef struct rtable {
 157   resource r;
 158   node n;                               /* Node in list of all tables */
 159   pool *rp;                             /* Resource pool to allocate everything from, including itself */
 160   struct fib fib;
 161   char *name;                           /* Name of this table */
 162   list channels;                        /* List of attached channels (struct channel) */
 163   uint addr_type;                       /* Type of address data stored in table (NET_*) */
 164   int pipe_busy;                        /* Pipe loop detection */
 165   int use_count;                        /* Number of protocols using this table */
 166   u32 rt_count;                         /* Number of routes in the table */
 167   struct hmap id_map;
 168   struct hostcache *hostcache;
 169   struct rtable_config *config;         /* Configuration of this table */
 170   struct config *deleted;               /* Table doesn't exist in current configuration,
 171                                          * delete as soon as use_count becomes 0 and remove
 172                                          * obstacle from this routing table.
 173                                          */
 174   struct event *rt_event;               /* Routing table event */
 175   btime last_rt_change;                 /* Last time when route changed */
 176   btime base_settle_time;               /* Start time of rtable settling interval */
 177   btime gc_time;                        /* Time of last GC */
 178   int gc_counter;                       /* Number of operations since last GC */
 179   byte prune_state;                     /* Table prune state, 1 -> scheduled, 2-> running */
 180   byte hcu_scheduled;                   /* Hostcache update is scheduled */
 181   byte nhu_state;                       /* Next Hop Update state */
 182   struct fib_iterator prune_fit;        /* Rtable prune FIB iterator */
 183   struct fib_iterator nhu_fit;          /* Next Hop Update FIB iterator */
 184
 185   list subscribers;                     /* Subscribers for notifications */
 186   struct timer *settle_timer;           /* Settle time for notifications */
 187 } rtable;
 188
 189 struct rt_subscription {
 190   node n;
 191   rtable *tab;
 192   void (*hook)(struct rt_subscription *b);
 193   void *data;
 194 };
 195
 196 #define NHU_CLEAN       0
 197 #define NHU_SCHEDULED   1
 198 #define NHU_RUNNING     2
 199 #define NHU_DIRTY       3
 200
 201 typedef struct network {
 202   struct rte *routes;                   /* Available routes for this network */
 203   struct fib_node n;                    /* FIB flags reserved for kernel syncer */
 204 } net;
 205
 206 struct hostcache {
 207   slab *slab;                           /* Slab holding all hostentries */
 208   struct hostentry **hash_table;        /* Hash table for hostentries */
 209   unsigned hash_order, hash_shift;
 210   unsigned hash_max, hash_min;
 211   unsigned hash_items;
 212   linpool *lp;                          /* Linpool for trie */
 213   struct f_trie *trie;                  /* Trie of prefixes that might affect hostentries */
 214   list hostentries;                     /* List of all hostentries */
 215   byte update_hostcache;
 216 };
 217
 218 struct hostentry {
 219   node ln;
 220   ip_addr addr;                         /* IP address of host, part of key */
 221   ip_addr link;                         /* (link-local) IP address of host, used as gw
 222                                            if host is directly attached */
 223   struct rtable *tab;                   /* Dependent table, part of key */
 224   struct hostentry *next;               /* Next in hash chain */
 225   unsigned hash_key;                    /* Hash key */
 226   unsigned uc;                          /* Use count */
 227   struct rta *src;                      /* Source rta entry */
 228   byte dest;                            /* Chosen route destination type (RTD_...) */
 229   byte nexthop_linkable;                /* Nexthop list is completely non-device */
 230   u32 igp_metric;                       /* Chosen route IGP metric */
 231 };
 232
 233 typedef struct rte {
 234   struct rte *next;
 235   net *net;                             /* Network this RTE belongs to */
 236   struct channel *sender;               /* Channel used to send the route to the routing table */
 237   struct rta *attrs;                    /* Attributes of this route */
 238   u32 id;                               /* Table specific route id */
 239   byte flags;                           /* Flags (REF_...) */
 240   byte pflags;                          /* Protocol-specific flags */
 241   word pref;                            /* Route preference */
 242   btime lastmod;                        /* Last modified */
 243   union {                               /* Protocol-dependent data (metrics etc.) */
 244 #ifdef CONFIG_RIP
 245     struct {
 246       struct iface *from;               /* Incoming iface */
 247       u8 metric;                        /* RIP metric */
 248       u16 tag;                          /* External route tag */
 249     } rip;
 250 #endif
 251 #ifdef CONFIG_OSPF
 252     struct {
 253       u32 metric1, metric2;             /* OSPF Type 1 and Type 2 metrics */
 254       u32 tag;                          /* External route tag */
 255       u32 router_id;                    /* Router that originated this route */
 256     } ospf;
 257 #endif
 258 #ifdef CONFIG_BGP
 259     struct {
 260       u8 suppressed;                    /* Used for deterministic MED comparison */
 261       s8 stale;                         /* Route is LLGR_STALE, -1 if unknown */
 262     } bgp;
 263 #endif
 264 #ifdef CONFIG_BABEL
 265     struct {
 266       u16 seqno;                        /* Babel seqno */
 267       u16 metric;                       /* Babel metric */
 268       u64 router_id;                    /* Babel router id */
 269     } babel;
 270 #endif
 271     struct {                            /* Routes generated by krt sync (both temporary and inherited ones) */
 272       s8 src;                           /* Alleged route source (see krt.h) */
 273       u8 proto;                         /* Kernel source protocol ID */
 274       u8 seen;                          /* Seen during last scan */
 275       u8 best;                          /* Best route in network, propagated to core */
 276       u32 metric;                       /* Kernel metric */
 277     } krt;
 278   } u;
 279 } rte;
 280
 281 #define REF_COW         1               /* Copy this rte on write */
 282 #define REF_FILTERED    2               /* Route is rejected by import filter */
 283 #define REF_STALE       4               /* Route is stale in a refresh cycle */
 284 #define REF_DISCARD     8               /* Route is scheduled for discard */
 285 #define REF_MODIFY      16              /* Route is scheduled for modify */
 286
 287 /* Route is valid for propagation (may depend on other flags in the future), accepts NULL */
 288 static inline int rte_is_valid(rte *r) { return r && !(r->flags & REF_FILTERED); }
 289
 290 /* Route just has REF_FILTERED flag */
 291 static inline int rte_is_filtered(rte *r) { return !!(r->flags & REF_FILTERED); }
 292
 293
 294 /* Types of route announcement, also used as flags */
 295 #define RA_UNDEF        0               /* Undefined RA type */
 296 #define RA_OPTIMAL      1               /* Announcement of optimal route change */
 297 #define RA_ACCEPTED     2               /* Announcement of first accepted route */
 298 #define RA_ANY          3               /* Announcement of any route change */
 299 #define RA_MERGED       4               /* Announcement of optimal route merged with next ones */
 300
 301 /* Return value of preexport() callback */
 302 #define RIC_ACCEPT      1               /* Accepted by protocol */
 303 #define RIC_PROCESS     0               /* Process it through import filter */
 304 #define RIC_REJECT      -1              /* Rejected by protocol */
 305 #define RIC_DROP        -2              /* Silently dropped by protocol */
 306
 307 extern list routing_tables;
 308 struct config;
 309
 310 void rt_init(void);
 311 void rt_preconfig(struct config *);
 312 void rt_commit(struct config *new, struct config *old);
 313 void rt_lock_table(rtable *);
 314 void rt_unlock_table(rtable *);
 315 void rt_subscribe(rtable *tab, struct rt_subscription *s);
 316 void rt_unsubscribe(struct rt_subscription *s);
 317 rtable *rt_setup(pool *, struct rtable_config *);
 318 static inline void rt_shutdown(rtable *r) { rfree(r->rp); }
 319
 320 static inline net *net_find(rtable *tab, const net_addr *addr) { return (net *) fib_find(&tab->fib, addr); }
 321 static inline net *net_find_valid(rtable *tab, const net_addr *addr)
 322 { net *n = net_find(tab, addr); return (n && rte_is_valid(n->routes)) ? n : NULL; }
 323 static inline net *net_get(rtable *tab, const net_addr *addr) { return (net *) fib_get(&tab->fib, addr); }
 324 void *net_route(rtable *tab, const net_addr *n);
 325 int net_roa_check(rtable *tab, const net_addr *n, u32 asn);
 326 rte *rte_find(net *net, struct rte_src *src);
 327 rte *rte_get_temp(struct rta *);
 328 void rte_update2(struct channel *c, const net_addr *n, rte *new, struct rte_src *src);
 329 /* rte_update() moved to protocol.h to avoid dependency conflicts */
 330 int rt_examine(rtable *t, net_addr *a, struct proto *p, const struct filter *filter);
 331 rte *rt_export_merged(struct channel *c, net *net, rte **rt_free, linpool *pool, int silent);
 332 void rt_refresh_begin(rtable *t, struct channel *c);
 333 void rt_refresh_end(rtable *t, struct channel *c);
 334 void rt_modify_stale(rtable *t, struct channel *c);
 335 void rt_schedule_prune(rtable *t);
 336 void rte_dump(rte *);
 337 void rte_free(rte *);
 338 rte *rte_do_cow(rte *);
 339 static inline rte * rte_cow(rte *r) { return (r->flags & REF_COW) ? rte_do_cow(r) : r; }
 340 rte *rte_cow_rta(rte *r, linpool *lp);
 341 void rte_init_tmp_attrs(struct rte *r, linpool *lp, uint max);
 342 void rte_make_tmp_attr(struct rte *r, uint id, uint type, uintptr_t val);
 343 void rte_make_tmp_attrs(struct rte **r, struct linpool *pool, struct rta **old_attrs);
 344 uintptr_t rte_store_tmp_attr(struct rte *r, uint id);
 345 void rt_dump(rtable *);
 346 void rt_dump_all(void);
 347 int rt_feed_channel(struct channel *c);
 348 void rt_feed_channel_abort(struct channel *c);
 349 int rte_update_in(struct channel *c, const net_addr *n, rte *new, struct rte_src *src);
 350 int rt_reload_channel(struct channel *c);
 351 void rt_reload_channel_abort(struct channel *c);
 352 void rt_prune_sync(rtable *t, int all);
 353 int rte_update_out(struct channel *c, const net_addr *n, rte *new, rte *old0, int refeed);
 354 struct rtable_config *rt_new_table(struct symbol *s, uint addr_type);
 355
 356
 357 /* Default limit for ECMP next hops, defined in sysdep code */
 358 extern const int rt_default_ecmp;
 359
 360 struct rt_show_data_rtable {
 361   node n;
 362   rtable *table;
 363   struct channel *export_channel;
 364 };
 365
 366 struct rt_show_data {
 367   net_addr *addr;
 368   list tables;
 369   struct rt_show_data_rtable *tab;      /* Iterator over table list */
 370   struct rt_show_data_rtable *last_table; /* Last table in output */
 371   struct fib_iterator fit;              /* Iterator over networks in table */
 372   int verbose, tables_defined_by;
 373   const struct filter *filter;
 374   struct proto *show_protocol;
 375   struct proto *export_protocol;
 376   struct channel *export_channel;
 377   struct config *running_on_config;
 378   struct krt_proto *kernel;
 379   int export_mode, primary_only, filtered, stats, show_for;
 380
 381   int table_open;                       /* Iteration (fit) is open */
 382   int net_counter, rt_counter, show_counter, table_counter;
 383   int net_counter_last, rt_counter_last, show_counter_last;
 384 };
 385
 386 void rt_show(struct rt_show_data *);
 387 struct rt_show_data_rtable * rt_show_add_table(struct rt_show_data *d, rtable *t);
 388
 389 /* Value of table definition mode in struct rt_show_data */
 390 #define RSD_TDB_DEFAULT   0             /* no table specified */
 391 #define RSD_TDB_INDIRECT  0             /* show route ... protocol P ... */
 392 #define RSD_TDB_ALL       RSD_TDB_SET                   /* show route ... table all ... */
 393 #define RSD_TDB_DIRECT    RSD_TDB_SET | RSD_TDB_NMN     /* show route ... table X table Y ... */
 394
 395 #define RSD_TDB_SET       0x1           /* internal: show empty tables */
 396 #define RSD_TDB_NMN       0x2           /* internal: need matching net */
 397
 398 /* Value of export_mode in struct rt_show_data */
 399 #define RSEM_NONE       0               /* Export mode not used */
 400 #define RSEM_PREEXPORT  1               /* Routes ready for export, before filtering */
 401 #define RSEM_EXPORT     2               /* Routes accepted by export filter */
 402 #define RSEM_NOEXPORT   3               /* Routes rejected by export filter */
 403 #define RSEM_EXPORTED   4               /* Routes marked in export map */
 404
 405 /*
 406  *      Route Attributes
 407  *
 408  *      Beware: All standard BGP attributes must be represented here instead
 409  *      of making them local to the route. This is needed to ensure proper
 410  *      construction of BGP route attribute lists.
 411  */
 412
 413 /* Nexthop structure */
 414 struct nexthop {
 415   ip_addr gw;                           /* Next hop */
 416   struct iface *iface;                  /* Outgoing interface */
 417   struct nexthop *next;
 418   byte flags;
 419   byte weight;
 420   byte labels_orig;                     /* Number of labels before hostentry was applied */
 421   byte labels;                          /* Number of all labels */
 422   u32 label[0];
 423 };
 424
 425 #define RNF_ONLINK              0x1     /* Gateway is onlink regardless of IP ranges */
 426
 427
 428 struct rte_src {
 429   struct rte_src *next;                 /* Hash chain */
 430   struct proto *proto;                  /* Protocol the source is based on */
 431   u32 private_id;                       /* Private ID, assigned by the protocol */
 432   u32 global_id;                        /* Globally unique ID of the source */
 433   unsigned uc;                          /* Use count */
 434 };
 435
 436
 437 typedef struct rta {
 438   struct rta *next, **pprev;            /* Hash chain */
 439   u32 uc;                               /* Use count */
 440   u32 hash_key;                         /* Hash over important fields */
 441   struct ea_list *eattrs;               /* Extended Attribute chain */
 442   struct rte_src *src;                  /* Route source that created the route */
 443   struct hostentry *hostentry;          /* Hostentry for recursive next-hops */
 444   ip_addr from;                         /* Advertising router */
 445   u32 igp_metric;                       /* IGP metric to next hop (for iBGP routes) */
 446   u8 source;                            /* Route source (RTS_...) */
 447   u8 scope;                             /* Route scope (SCOPE_... -- see ip.h) */
 448   u8 dest;                              /* Route destination type (RTD_...) */
 449   u8 aflags;
 450   struct nexthop nh;                    /* Next hop */
 451 } rta;
 452
 453 #define RTS_DUMMY 0                     /* Dummy route to be removed soon */
 454 #define RTS_STATIC 1                    /* Normal static route */
 455 #define RTS_INHERIT 2                   /* Route inherited from kernel */
 456 #define RTS_DEVICE 3                    /* Device route */
 457 #define RTS_STATIC_DEVICE 4             /* Static device route */
 458 #define RTS_REDIRECT 5                  /* Learned via redirect */
 459 #define RTS_RIP 6                       /* RIP route */
 460 #define RTS_OSPF 7                      /* OSPF route */
 461 #define RTS_OSPF_IA 8                   /* OSPF inter-area route */
 462 #define RTS_OSPF_EXT1 9                 /* OSPF external route type 1 */
 463 #define RTS_OSPF_EXT2 10                /* OSPF external route type 2 */
 464 #define RTS_BGP 11                      /* BGP route */
 465 #define RTS_PIPE 12                     /* Inter-table wormhole */
 466 #define RTS_BABEL 13                    /* Babel route */
 467 #define RTS_RPKI 14                     /* Route Origin Authorization */
 468 #define RTS_PERF 15                     /* Perf checker */
 469 #define RTS_MAX 16
 470
 471 #define RTC_UNICAST 0
 472 #define RTC_BROADCAST 1
 473 #define RTC_MULTICAST 2
 474 #define RTC_ANYCAST 3                   /* IPv6 Anycast */
 475
 476 #define RTD_NONE 0                      /* Undefined next hop */
 477 #define RTD_UNICAST 1                   /* Next hop is neighbor router */
 478 #define RTD_BLACKHOLE 2                 /* Silently drop packets */
 479 #define RTD_UNREACHABLE 3               /* Reject as unreachable */
 480 #define RTD_PROHIBIT 4                  /* Administratively prohibited */
 481 #define RTD_MAX 5
 482
 483 #define RTAF_CACHED 1                   /* This is a cached rta */
 484
 485 #define IGP_METRIC_UNKNOWN 0x80000000   /* Default igp_metric used when no other
 486                                            protocol-specific metric is availabe */
 487
 488
 489 extern const char * rta_dest_names[RTD_MAX];
 490
 491 static inline const char *rta_dest_name(uint n)
 492 { return (n < RTD_MAX) ? rta_dest_names[n] : "???"; }
 493
 494 /* Route has regular, reachable nexthop (i.e. not RTD_UNREACHABLE and like) */
 495 static inline int rte_is_reachable(rte *r)
 496 { return r->attrs->dest == RTD_UNICAST; }
 497
 498
 499 /*
 500  *      Extended Route Attributes
 501  */
 502
 503 typedef struct eattr {
 504   word id;                              /* EA_CODE(PROTOCOL_..., protocol-dependent ID) */
 505   byte flags;                           /* Protocol-dependent flags */
 506   byte type;                            /* Attribute type and several flags (EAF_...) */
 507   union {
 508     u32 data;
 509     const struct adata *ptr;                    /* Attribute data elsewhere */
 510   } u;
 511 } eattr;
 512
 513
 514 #define EA_CODE(proto,id) (((proto) << 8) | (id))
 515 #define EA_ID(ea) ((ea) & 0xff)
 516 #define EA_PROTO(ea) ((ea) >> 8)
 517 #define EA_ID_FLAG(ea) (1 << EA_ID(ea))
 518 #define EA_CUSTOM(id) ((id) | EA_CUSTOM_BIT)
 519 #define EA_IS_CUSTOM(ea) ((ea) & EA_CUSTOM_BIT)
 520 #define EA_CUSTOM_ID(ea) ((ea) & ~EA_CUSTOM_BIT)
 521
 522 const char *ea_custom_name(uint ea);
 523
 524 #define EA_GEN_IGP_METRIC EA_CODE(PROTOCOL_NONE, 0)
 525
 526 #define EA_CODE_MASK 0xffff
 527 #define EA_CUSTOM_BIT 0x8000
 528 #define EA_ALLOW_UNDEF 0x10000          /* ea_find: allow EAF_TYPE_UNDEF */
 529 #define EA_BIT(n) ((n) << 24)           /* Used in bitfield accessors */
 530 #define EA_BIT_GET(ea) ((ea) >> 24)
 531
 532 #define EAF_TYPE_MASK 0x1f              /* Mask with this to get type */
 533 #define EAF_TYPE_INT 0x01               /* 32-bit unsigned integer number */
 534 #define EAF_TYPE_OPAQUE 0x02            /* Opaque byte string (not filterable) */
 535 #define EAF_TYPE_IP_ADDRESS 0x04        /* IP address */
 536 #define EAF_TYPE_ROUTER_ID 0x05         /* Router ID (IPv4 address) */
 537 #define EAF_TYPE_AS_PATH 0x06           /* BGP AS path (encoding per RFC 1771:4.3) */
 538 #define EAF_TYPE_BITFIELD 0x09          /* 32-bit embedded bitfield */
 539 #define EAF_TYPE_INT_SET 0x0a           /* Set of u32's (e.g., a community list) */
 540 #define EAF_TYPE_EC_SET 0x0e            /* Set of pairs of u32's - ext. community list */
 541 #define EAF_TYPE_LC_SET 0x12            /* Set of triplets of u32's - large community list */
 542 #define EAF_TYPE_UNDEF 0x1f             /* `force undefined' entry */
 543 #define EAF_EMBEDDED 0x01               /* Data stored in eattr.u.data (part of type spec) */
 544 #define EAF_VAR_LENGTH 0x02             /* Attribute length is variable (part of type spec) */
 545 #define EAF_ORIGINATED 0x20             /* The attribute has originated locally */
 546 #define EAF_FRESH 0x40                  /* An uncached attribute (e.g. modified in export filter) */
 547
 548 typedef struct adata {
 549   uint length;                          /* Length of data */
 550   byte data[0];
 551 } adata;
 552
 553 extern const adata null_adata;          /* adata of length 0 */
 554
 555 static inline struct adata *
 556 lp_alloc_adata(struct linpool *pool, uint len)
 557 {
 558   struct adata *ad = lp_alloc(pool, sizeof(struct adata) + len);
 559   ad->length = len;
 560   return ad;
 561 }
 562
 563 static inline int adata_same(const struct adata *a, const struct adata *b)
 564 { return (a->length == b->length && !memcmp(a->data, b->data, a->length)); }
 565
 566
 567 typedef struct ea_list {
 568   struct ea_list *next;                 /* In case we have an override list */
 569   byte flags;                           /* Flags: EALF_... */
 570   byte rfu;
 571   word count;                           /* Number of attributes */
 572   eattr attrs[0];                       /* Attribute definitions themselves */
 573 } ea_list;
 574
 575 #define EALF_SORTED 1                   /* Attributes are sorted by code */
 576 #define EALF_BISECT 2                   /* Use interval bisection for searching */
 577 #define EALF_CACHED 4                   /* Attributes belonging to cached rta */
 578 #define EALF_TEMP 8                     /* Temporary ea_list added by make_tmp_attrs hooks */
 579
 580 struct rte_src *rt_find_source(struct proto *p, u32 id);
 581 struct rte_src *rt_get_source(struct proto *p, u32 id);
 582 static inline void rt_lock_source(struct rte_src *src) { src->uc++; }
 583 static inline void rt_unlock_source(struct rte_src *src) { src->uc--; }
 584 void rt_prune_sources(void);
 585
 586 struct ea_walk_state {
 587   ea_list *eattrs;                      /* Ccurrent ea_list, initially set by caller */
 588   eattr *ea;                            /* Current eattr, initially NULL */
 589   u32 visited[4];                       /* Bitfield, limiting max to 128 */
 590 };
 591
 592 eattr *ea_find(ea_list *, unsigned ea);
 593 eattr *ea_walk(struct ea_walk_state *s, uint id, uint max);
 594 int ea_get_int(ea_list *, unsigned ea, int def);
 595 void ea_dump(ea_list *);
 596 void ea_sort(ea_list *);                /* Sort entries in all sub-lists */
 597 unsigned ea_scan(ea_list *);            /* How many bytes do we need for merged ea_list */
 598 void ea_merge(ea_list *from, ea_list *to); /* Merge sub-lists to allocated buffer */
 599 int ea_same(ea_list *x, ea_list *y);    /* Test whether two ea_lists are identical */
 600 uint ea_hash(ea_list *e);       /* Calculate 16-bit hash value */
 601 ea_list *ea_append(ea_list *to, ea_list *what);
 602 void ea_format_bitfield(const struct eattr *a, byte *buf, int bufsize, const char **names, int min, int max);
 603
 604 #define ea_normalize(ea) do { \
 605   if (ea->next) { \
 606     ea_list *t = alloca(ea_scan(ea)); \
 607     ea_merge(ea, t); \
 608     ea = t; \
 609   } \
 610   ea_sort(ea); \
 611   if (ea->count == 0) \
 612     ea = NULL; \
 613 } while(0) \
 614
 615 static inline eattr *
 616 ea_set_attr(ea_list **to, struct linpool *pool, uint id, uint flags, uint type, uintptr_t val)
 617 {
 618   ea_list *a = lp_alloc(pool, sizeof(ea_list) + sizeof(eattr));
 619   eattr *e = &a->attrs[0];
 620
 621   a->flags = EALF_SORTED;
 622   a->count = 1;
 623   a->next = *to;
 624   *to = a;
 625
 626   e->id = id;
 627   e->type = type;
 628   e->flags = flags;
 629
 630   if (type & EAF_EMBEDDED)
 631     e->u.data = (u32) val;
 632   else
 633     e->u.ptr = (struct adata *) val;
 634
 635   return e;
 636 }
 637
 638 static inline void
 639 ea_set_attr_u32(ea_list **to, struct linpool *pool, uint id, uint flags, uint type, u32 val)
 640 { ea_set_attr(to, pool, id, flags, type, (uintptr_t) val); }
 641
 642 static inline void
 643 ea_set_attr_ptr(ea_list **to, struct linpool *pool, uint id, uint flags, uint type, struct adata *val)
 644 { ea_set_attr(to, pool, id, flags, type, (uintptr_t) val); }
 645
 646 static inline void
 647 ea_set_attr_data(ea_list **to, struct linpool *pool, uint id, uint flags, uint type, void *data, uint len)
 648 {
 649   struct adata *a = lp_alloc_adata(pool, len);
 650   memcpy(a->data, data, len);
 651   ea_set_attr(to, pool, id, flags, type, (uintptr_t) a);
 652 }
 653
 654
 655 #define NEXTHOP_MAX_SIZE (sizeof(struct nexthop) + sizeof(u32)*MPLS_MAX_LABEL_STACK)
 656
 657 static inline size_t nexthop_size(const struct nexthop *nh)
 658 { return sizeof(struct nexthop) + sizeof(u32)*nh->labels; }
 659 int nexthop__same(struct nexthop *x, struct nexthop *y); /* Compare multipath nexthops */
 660 static inline int nexthop_same(struct nexthop *x, struct nexthop *y)
 661 { return (x == y) || nexthop__same(x, y); }
 662 struct nexthop *nexthop_merge(struct nexthop *x, struct nexthop *y, int rx, int ry, int max, linpool *lp);
 663 struct nexthop *nexthop_sort(struct nexthop *x);
 664 static inline void nexthop_link(struct rta *a, struct nexthop *from)
 665 { memcpy(&a->nh, from, nexthop_size(from)); }
 666 void nexthop_insert(struct nexthop **n, struct nexthop *y);
 667 int nexthop_is_sorted(struct nexthop *x);
 668
 669 void rta_init(void);
 670 static inline size_t rta_size(const rta *a) { return sizeof(rta) + sizeof(u32)*a->nh.labels; }
 671 #define RTA_MAX_SIZE (sizeof(rta) + sizeof(u32)*MPLS_MAX_LABEL_STACK)
 672 rta *rta_lookup(rta *);                 /* Get rta equivalent to this one, uc++ */
 673 static inline int rta_is_cached(rta *r) { return r->aflags & RTAF_CACHED; }
 674 static inline rta *rta_clone(rta *r) { r->uc++; return r; }
 675 void rta__free(rta *r);
 676 static inline void rta_free(rta *r) { if (r && !--r->uc) rta__free(r); }
 677 rta *rta_do_cow(rta *o, linpool *lp);
 678 static inline rta * rta_cow(rta *r, linpool *lp) { return rta_is_cached(r) ? rta_do_cow(r, lp) : r; }
 679 void rta_dump(rta *);
 680 void rta_dump_all(void);
 681 void rta_show(struct cli *, rta *);
 682
 683 u32 rt_get_igp_metric(rte *rt);
 684 struct hostentry * rt_get_hostentry(rtable *tab, ip_addr a, ip_addr ll, rtable *dep);
 685 void rta_apply_hostentry(rta *a, struct hostentry *he, mpls_label_stack *mls);
 686
 687 static inline void
 688 rta_set_recursive_next_hop(rtable *dep, rta *a, rtable *tab, ip_addr gw, ip_addr ll, mpls_label_stack *mls)
 689 {
 690   rta_apply_hostentry(a, rt_get_hostentry(tab, gw, ll, dep), mls);
 691 }
 692
 693 /*
 694  * rta_set_recursive_next_hop() acquires hostentry from hostcache and fills
 695  * rta->hostentry field.  New hostentry has zero use count. Cached rta locks its
 696  * hostentry (increases its use count), uncached rta does not lock it. Hostentry
 697  * with zero use count is removed asynchronously during host cache update,
 698  * therefore it is safe to hold such hostentry temorarily. Hostentry holds a
 699  * lock for a 'source' rta, mainly to share multipath nexthops.
 700  *
 701  * There is no need to hold a lock for hostentry->dep table, because that table
 702  * contains routes responsible for that hostentry, and therefore is non-empty if
 703  * given hostentry has non-zero use count. If the hostentry has zero use count,
 704  * the entry is removed before dep is referenced.
 705  *
 706  * The protocol responsible for routes with recursive next hops should hold a
 707  * lock for a 'source' table governing that routes (argument tab to
 708  * rta_set_recursive_next_hop()), because its routes reference hostentries
 709  * (through rta) related to the governing table. When all such routes are
 710  * removed, rtas are immediately removed achieving zero uc. Then the 'source'
 711  * table lock could be immediately released, although hostentries may still
 712  * exist - they will be freed together with the 'source' table.
 713  */
 714
 715 static inline void rt_lock_hostentry(struct hostentry *he) { if (he) he->uc++; }
 716 static inline void rt_unlock_hostentry(struct hostentry *he) { if (he) he->uc--; }
 717
 718 /*
 719  *      Default protocol preferences
 720  */
 721
 722 #define DEF_PREF_DIRECT         240     /* Directly connected */
 723 #define DEF_PREF_STATIC         200     /* Static route */
 724 #define DEF_PREF_OSPF           150     /* OSPF intra-area, inter-area and type 1 external routes */
 725 #define DEF_PREF_BABEL          130     /* Babel */
 726 #define DEF_PREF_RIP            120     /* RIP */
 727 #define DEF_PREF_BGP            100     /* BGP */
 728 #define DEF_PREF_RPKI           100     /* RPKI */
 729 #define DEF_PREF_INHERITED      10      /* Routes inherited from other routing daemons */
 730
 731 /*
 732  *      Route Origin Authorization
 733  */
 734
 735 #define ROA_UNKNOWN     0
 736 #define ROA_VALID       1
 737 #define ROA_INVALID     2
 738
 739 #endif