route.h 16.4 KB
Newer Older
1 2 3
/*
 *	BIRD Internet Routing Daemon -- Routing Table
 *
4
 *	(c) 1998--2000 Martin Mares <mj@ucw.cz>
5 6 7 8 9 10 11
 *
 *	Can be freely distributed and used under the terms of the GNU GPL.
 */

#ifndef _BIRD_ROUTE_H_
#define _BIRD_ROUTE_H_

12
#include "lib/lists.h"
13
#include "lib/resource.h"
14
#include "lib/timer.h"
15

16
struct protocol;
17
struct proto;
18 19 20
struct symbol;
struct filter;
struct cli;
21

22 23
/*
 *	Generic data structure for storing network prefixes. Also used
24
 *	for the master routing table. Currently implemented as a hash
25
 *	table.
26 27 28 29
 *
 *	Available operations:
 *		- insertion of new entry
 *		- deletion of entry
30
 *		- searching for entry by network prefix
31
 *		- asynchronous retrieval of fib contents
32 33 34
 */

struct fib_node {
35 36
  struct fib_node *next;		/* Next in hash chain */
  struct fib_iterator *readers;		/* List of readers of this node */
37
  byte pxlen;
38
  byte flags;				/* User-defined */
39
  byte x0, x1;				/* User-defined */
40
  u32 uid;				/* Unique ID based on hash */
41 42 43 44 45 46 47
  ip_addr prefix;			/* In host order */
};

struct fib_iterator {			/* See lib/slists.h for an explanation */
  struct fib_iterator *prev, *next;	/* Must be synced with struct fib_node! */
  byte efef;				/* 0xff to distinguish between iterator and node */
  byte pad[3];
48
  struct fib_node *node;		/* Or NULL if freshly merged */
49
  unsigned int hash;
50 51
};

Martin Mareš's avatar
Martin Mareš committed
52 53
typedef void (*fib_init_func)(struct fib_node *);

54
struct fib {
55 56 57
  pool *fib_pool;			/* Pool holding all our data */
  slab *fib_slab;			/* Slab holding all fib nodes */
  struct fib_node **hash_table;		/* Node hash table */
58
  unsigned int hash_size;		/* Number of hash table entries (a power of two) */
59 60
  unsigned int hash_order;		/* Binary logarithm of hash_size */
  unsigned int hash_shift;		/* 16 - hash_log */
61 62
  unsigned int entries;			/* Number of entries */
  unsigned int entries_min, entries_max;/* Entry count limits (else start rehashing) */
Martin Mareš's avatar
Martin Mareš committed
63
  fib_init_func init;			/* Constructor */
64 65
};

Martin Mareš's avatar
Martin Mareš committed
66
void fib_init(struct fib *, pool *, unsigned node_size, unsigned hash_order, fib_init_func init);
67 68
void *fib_find(struct fib *, ip_addr *, int);	/* Find or return NULL if doesn't exist */
void *fib_get(struct fib *, ip_addr *, int); 	/* Find or create new if nonexistent */
69
void *fib_route(struct fib *, ip_addr, int);	/* Longest-match routing lookup */
70
void fib_delete(struct fib *, void *);	/* Remove fib entry */
71
void fib_free(struct fib *);		/* Destroy the fib */
72 73 74 75 76
void fib_check(struct fib *);		/* Consistency check for debugging */

void fit_init(struct fib_iterator *, struct fib *); /* Internal functions, don't call */
struct fib_node *fit_get(struct fib *, struct fib_iterator *);
void fit_put(struct fib_iterator *, struct fib_node *);
77

78
#define FIB_WALK(fib, z) do {					\
79
	struct fib_node *z, **ff = (fib)->hash_table;		\
80 81
	unsigned int count = (fib)->hash_size;			\
	while (count--)						\
82 83 84
	  for(z = *ff++; z; z=z->next)

#define FIB_WALK_END } while (0)
85

86 87 88 89 90 91 92
#define FIB_ITERATE_INIT(it, fib) fit_init(it, fib)

#define FIB_ITERATE_START(fib, it, z) do {			\
	struct fib_node *z = fit_get(fib, it);			\
	unsigned int count = (fib)->hash_size;			\
	unsigned int hpos = (it)->hash;				\
	for(;;) {						\
93 94 95 96 97 98 99
	  if (!z)						\
            {							\
	       if (++hpos >= count)				\
		 break;						\
	       z = (fib)->hash_table[hpos];			\
	       continue;					\
	    }
100

101
#define FIB_ITERATE_END(z) z = z->next; } } while(0)
102 103 104

#define FIB_ITERATE_PUT(it, z) fit_put(it, z)

105
/*
106 107 108 109
 *	Master Routing Tables. Generally speaking, each of them contains a FIB
 *	with each entry pointing to a list of route entries representing routes
 *	to given network (with the selected one at the head).
 *
110
 *	Each of the RTE's contains variable data (the preference and protocol-dependent
111
 *	metrics) and a pointer to a route attribute block common for many routes).
112 113
 *
 *	It's guaranteed that there is at most one RTE for every (prefix,proto) pair.
114 115
 */

116 117 118 119
struct rtable_config {
  node n;
  char *name;
  struct rtable *table;
120
  struct proto_config *krt_attached;	/* Kernel syncer attached to this table */
121 122
  int gc_max_ops;			/* Maximum number of operations before GC is run */
  int gc_min_time;			/* Minimum time between two consecutive GC runs */
123 124
};

125
typedef struct rtable {
126
  node n;				/* Node in list of all tables */
127 128
  struct fib fib;
  char *name;				/* Name of this table */
129
  list hooks;				/* List of announcement hooks */
130
  int pipe_busy;			/* Pipe loop detection */
131
  int use_count;			/* Number of protocols using this table */
132
  struct hostcache *hostcache;
133
  struct rtable_config *config;		/* Configuration of this table */
134 135 136 137
  struct config *deleted;		/* Table doesn't exist in current configuration,
					 * delete as soon as use_count becomes 0 and remove
					 * obstacle from this routing table.
					 */
138
  struct event *rt_event;		/* Routing table event */
139 140
  int gc_counter;			/* Number of operations since last GC */
  bird_clock_t gc_time;			/* Time of last GC */
141 142 143 144
  byte gc_scheduled;			/* GC is scheduled */
  byte hcu_scheduled;			/* Hostcache update is scheduled */
  byte nhu_state;			/* Next Hop Update state */
  struct fib_iterator nhu_fit;		/* Next Hop Update FIB iterator */
145 146
} rtable;

147
typedef struct network {
148
  struct fib_node n;			/* FIB flags reserved for kernel syncer */
149 150 151
  struct rte *routes;			/* Available routes for this network */
} net;

152
struct hostcache {
153 154 155 156 157
  slab *slab;				/* Slab holding all hostentries */
  struct hostentry **hash_table;	/* Hash table for hostentries */
  unsigned hash_order, hash_shift;
  unsigned hash_max, hash_min;
  unsigned hash_items;
158 159 160
  linpool *lp;				/* Linpool for trie */
  struct f_trie *trie;			/* Trie of prefixes that might affect hostentries */
  list hostentries;			/* List of all hostentries */
161 162 163 164 165
  byte update_hostcache;
};

struct hostentry {
  node ln;
166 167 168
  ip_addr addr;				/* IP address of host, part of key */
  ip_addr link;				/* (link-local) IP address of host, used as gw
					   if host is directly attached */
169 170 171
  struct rtable *tab;			/* Dependent table, part of key*/
  struct hostentry *next;		/* Next in hash chain */
  unsigned hash_key;			/* Hash key */
172
  unsigned uc;				/* Use count */
173
  struct rta *src;			/* Source rta entry */
174 175
  ip_addr gw;				/* Chosen next hop */
  byte dest;				/* Chosen route destination type (RTD_...) */
176
  u32 igp_metric;			/* Chosen route IGP metric */
177 178
};

179 180
typedef struct rte {
  struct rte *next;
181
  net *net;				/* Network this RTE belongs to */
182
  struct proto *sender;			/* Protocol instance that sent the route to the routing table */
183
  struct rta *attrs;			/* Attributes of this route */
184
  byte flags;				/* Flags (REF_...) */
185
  byte pflags;				/* Protocol-specific flags */
186
  word pref;				/* Route preference */
187
  bird_clock_t lastmod;			/* Last modified */
188 189 190
  union {				/* Protocol-dependent data (metrics etc.) */
#ifdef CONFIG_RIP
    struct {
191
      node garbage;			/* List for garbage collection */
192
      byte metric;			/* RIP metric */
193
      u16 tag;				/* External route tag */
194
      struct rip_entry *entry;
195 196 197 198 199
    } rip;
#endif
#ifdef CONFIG_OSPF
    struct {
      u32 metric1, metric2;		/* OSPF Type 1 and Type 2 metrics */
200
      u32 tag;				/* External route tag */
201
      u32 router_id;			/* Router that originated this route */
202 203
    } ospf;
#endif
204 205 206 207 208 209 210
    struct {				/* Routes generated by krt sync (both temporary and inherited ones) */
      s8 src;				/* Alleged route source (see krt.h) */
      u8 proto;				/* Kernel source protocol ID */
      u8 type;				/* Kernel route type */
      u8 seen;				/* Seen during last scan */
      u32 metric;			/* Kernel metric */
    } krt;
211 212 213
  } u;
} rte;

214 215
#define REF_COW 1			/* Copy this rte on write */

216 217 218 219
/* Types of route announcement, also used as flags */
#define RA_OPTIMAL 1			/* Announcement of optimal route change */
#define RA_ANY 2			/* Announcement of any route change */

220
struct config;
221 222

void rt_init(void);
223
void rt_preconfig(struct config *);
224 225 226
void rt_commit(struct config *new, struct config *old);
void rt_lock_table(rtable *);
void rt_unlock_table(rtable *);
227
void rt_setup(pool *, rtable *, char *, struct rtable_config *);
228 229
static inline net *net_find(rtable *tab, ip_addr addr, unsigned len) { return (net *) fib_find(&tab->fib, &addr, len); }
static inline net *net_get(rtable *tab, ip_addr addr, unsigned len) { return (net *) fib_get(&tab->fib, &addr, len); }
230
rte *rte_find(net *net, struct proto *p);
231
rte *rte_get_temp(struct rta *);
Ondřej Zajíček's avatar
Ondřej Zajíček committed
232
void rte_update(rtable *tab, net *net, struct proto *p, struct proto *src, rte *new);
233
void rte_discard(rtable *tab, rte *old);
234
void rte_dump(rte *);
235
void rte_free(rte *);
236 237
rte *rte_do_cow(rte *);
static inline rte * rte_cow(rte *r) { return (r->flags & REF_COW) ? rte_do_cow(r) : r; }
238
void rt_dump(rtable *);
239
void rt_dump_all(void);
240 241
int rt_feed_baby(struct proto *p);
void rt_feed_baby_abort(struct proto *p);
242
void rt_prune_all(void);
243
struct rtable_config *rt_new_table(struct symbol *s);
244

245 246 247 248 249 250 251
struct rt_show_data {
  ip_addr prefix;
  unsigned pxlen;
  rtable *table;
  struct filter *filter;
  int verbose;
  struct fib_iterator fit;
252
  struct proto *show_protocol;
253 254
  struct proto *export_protocol;
  int export_mode, primary_only;
255
  struct config *running_on_config;
256
  int net_counter, rt_counter, show_counter;
257
  int stats, show_for;
258 259 260
};
void rt_show(struct rt_show_data *);

261 262 263 264 265 266 267 268
/*
 *	Route Attributes
 *
 *	Beware: All standard BGP attributes must be represented here instead
 *	of making them local to the route. This is needed to ensure proper
 *	construction of BGP route attribute lists.
 */

269 270 271 272 273 274 275 276
/* Multipath next-hop */
struct mpnh {
  ip_addr gw;				/* Next hop */
  struct iface *iface;			/* Outgoing interface */
  struct mpnh *next;
  unsigned char weight;
};

277
typedef struct rta {
278
  struct rta *next, **pprev;		/* Hash chain */
279
  struct proto *proto;			/* Protocol instance that originally created the route */
280 281
  unsigned uc;				/* Use count */
  byte source;				/* Route source (RTS_...) */
282
  byte scope;				/* Route scope (SCOPE_... -- see ip.h) */
283 284
  byte cast;				/* Casting type (RTC_...) */
  byte dest;				/* Route destination type (RTD_...) */
285
  byte flags;				/* Route flags (RTF_...), now unused */
286
  byte aflags;				/* Attribute cache flags (RTAF_...) */
287
  u16 hash_key;				/* Hash over important fields */
288
  u32 igp_metric;			/* IGP metric to next hop (for iBGP routes) */
289
  ip_addr gw;				/* Next hop */
290
  ip_addr from;				/* Advertising router */
291
  struct hostentry *hostentry;		/* Hostentry for recursive next-hops */
292
  struct iface *iface;			/* Outgoing interface */
293
  struct mpnh *nexthops;		/* Next-hops for multipath routes */
294
  struct ea_list *eattrs;		/* Extended Attribute chain */
295 296
} rta;

297
#define RTS_DUMMY 0			/* Dummy route to be removed soon */
298 299 300 301 302 303
#define RTS_STATIC 1			/* Normal static route */
#define RTS_INHERIT 2			/* Route inherited from kernel */
#define RTS_DEVICE 3			/* Device route */
#define RTS_STATIC_DEVICE 4		/* Static device route */
#define RTS_REDIRECT 5			/* Learned via redirect */
#define RTS_RIP 6			/* RIP route */
Martin Mareš's avatar
Martin Mareš committed
304
#define RTS_OSPF 7			/* OSPF route */
305
#define RTS_OSPF_IA 8			/* OSPF inter-area route */
Ondřej Filip's avatar
Ondřej Filip committed
306 307 308 309
#define RTS_OSPF_EXT1 9			/* OSPF external route type 1 */
#define RTS_OSPF_EXT2 10		/* OSPF external route type 2 */
#define RTS_BGP 11			/* BGP route */
#define RTS_PIPE 12			/* Inter-table wormhole */
310 311 312 313 314 315 316 317 318 319 320

#define RTC_UNICAST 0
#define RTC_BROADCAST 1
#define RTC_MULTICAST 2
#define RTC_ANYCAST 3			/* IPv6 Anycast */

#define RTD_ROUTER 0			/* Next hop is neighbor router */
#define RTD_DEVICE 1			/* Points to device */
#define RTD_BLACKHOLE 2			/* Silently drop packets */
#define RTD_UNREACHABLE 3		/* Reject as unreachable */
#define RTD_PROHIBIT 4			/* Administratively prohibited */
321 322
#define RTD_MULTIPATH 5			/* Multipath route (nexthops != NULL) */
#define RTD_NONE 6			/* Invalid RTD */
323

324 325
#define RTAF_CACHED 1			/* This is a cached rta */

326 327 328
#define IGP_METRIC_UNKNOWN 0x80000000	/* Default igp_metric used when no other
					   protocol-specific metric is availabe */

329 330 331 332 333
/*
 *	Extended Route Attributes
 */

typedef struct eattr {
334 335 336
  word id;				/* EA_CODE(EAP_..., protocol-dependent ID) */
  byte flags;				/* Protocol-dependent flags */
  byte type;				/* Attribute type and several flags (EAF_...) */
337 338 339 340 341 342 343 344
  union {
    u32 data;
    struct adata *ptr;			/* Attribute data elsewhere */
  } u;
} eattr;

#define EAP_GENERIC 0			/* Generic attributes */
#define EAP_BGP 1			/* BGP attributes */
345
#define EAP_RIP 2			/* RIP */
346
#define EAP_OSPF 3			/* OSPF */
347 348
#define EAP_KRT 4			/* Kernel route attributes */
#define EAP_MAX 5
349

350 351 352 353
#define EA_CODE(proto,id) (((proto) << 8) | (id))
#define EA_PROTO(ea) ((ea) >> 8)
#define EA_ID(ea) ((ea) & 0xff)

354 355
#define EA_GEN_IGP_METRIC EA_CODE(EAP_GENERIC, 0)

356 357 358
#define EA_CODE_MASK 0xffff
#define EA_ALLOW_UNDEF 0x10000		/* ea_find: allow EAF_TYPE_UNDEF */

359 360 361
#define EAF_TYPE_MASK 0x0f		/* Mask with this to get type */
#define EAF_TYPE_INT 0x01		/* 32-bit signed integer number */
#define EAF_TYPE_OPAQUE 0x02		/* Opaque byte string (not filterable) */
362 363
#define EAF_TYPE_IP_ADDRESS 0x04	/* IP address */
#define EAF_TYPE_ROUTER_ID 0x05		/* Router ID (IPv4 address) */
Martin Mareš's avatar
Martin Mareš committed
364 365
#define EAF_TYPE_AS_PATH 0x06		/* BGP AS path (encoding per RFC 1771:4.3) */
#define EAF_TYPE_INT_SET 0x0a		/* Set of u32's (e.g., a community list) */
366
#define EAF_TYPE_EC_SET 0x0e		/* Set of pairs of u32's - ext. community list */
367
#define EAF_TYPE_UNDEF 0x0f		/* `force undefined' entry */
368
#define EAF_EMBEDDED 0x01		/* Data stored in eattr.u.data (part of type spec) */
369
#define EAF_VAR_LENGTH 0x02		/* Attribute length is variable (part of type spec) */
370
#define EAF_ORIGINATED 0x40		/* The attribute has originated locally */
371
#define EAF_TEMP 0x80			/* A temporary attribute (the one stored in the tmp attr list) */
372 373

struct adata {
Martin Mareš's avatar
Martin Mareš committed
374
  unsigned int length;			/* Length of data */
375 376 377 378 379
  byte data[0];
};

typedef struct ea_list {
  struct ea_list *next;			/* In case we have an override list */
380
  byte flags;				/* Flags: EALF_... */
381
  byte rfu;
382
  word count;				/* Number of attributes */
383 384 385
  eattr attrs[0];			/* Attribute definitions themselves */
} ea_list;

386 387 388
#define EALF_SORTED 1			/* Attributes are sorted by code */
#define EALF_BISECT 2			/* Use interval bisection for searching */
#define EALF_CACHED 4			/* Attributes belonging to cached rta */
389

390
eattr *ea_find(ea_list *, unsigned ea);
391
int ea_get_int(ea_list *, unsigned ea, int def);
392 393
void ea_dump(ea_list *);
void ea_sort(ea_list *);		/* Sort entries in all sub-lists */
394
unsigned ea_scan(ea_list *);		/* How many bytes do we need for merged ea_list */
395
void ea_merge(ea_list *from, ea_list *to); /* Merge sub-lists to allocated buffer */
396 397
int ea_same(ea_list *x, ea_list *y);	/* Test whether two ea_lists are identical */
unsigned int ea_hash(ea_list *e);	/* Calculate 16-bit hash value */
398
ea_list *ea_append(ea_list *to, ea_list *what);
399

400 401 402 403
int mpnh__same(struct mpnh *x, struct mpnh *y); /* Compare multipath nexthops */
static inline int mpnh_same(struct mpnh *x, struct mpnh *y)
{ return (x == y) || mpnh__same(x, y); }

404 405 406
void rta_init(void);
rta *rta_lookup(rta *);			/* Get rta equivalent to this one, uc++ */
static inline rta *rta_clone(rta *r) { r->uc++; return r; }
407 408
void rta__free(rta *r);
static inline void rta_free(rta *r) { if (r && !--r->uc) rta__free(r); }
409 410
void rta_dump(rta *);
void rta_dump_all(void);
411
void rta_show(struct cli *, rta *, ea_list *);
412
void rta_set_recursive_next_hop(rtable *dep, rta *a, rtable *tab, ip_addr *gw, ip_addr *ll);
413 414 415 416 417 418 419

/*
 * rta_set_recursive_next_hop() acquires hostentry from hostcache and
 * fills rta->hostentry field.  New hostentry has zero use
 * count. Cached rta locks its hostentry (increases its use count),
 * uncached rta does not lock it. Hostentry with zero use count is
 * removed asynchronously during host cache update, therefore it is
420 421 422 423 424 425 426 427
 * safe to hold such hostentry temorarily. Hostentry holds a lock for
 * a 'source' rta, mainly to share multipath nexthops. There is no
 * need to hold a lock for hostentry->dep table, because that table
 * contains routes responsible for that hostentry, and therefore is
 * non-empty if given hostentry has non-zero use count. The protocol
 * responsible for routes with recursive next hops should also hold a
 * lock for a table governing that routes (argument tab to
 * rta_set_recursive_next_hop()).
428 429 430 431 432
 */

static inline void rt_lock_hostentry(struct hostentry *he) { if (he) he->uc++; }
static inline void rt_unlock_hostentry(struct hostentry *he) { if (he) he->uc--; }

433

434 435
extern struct protocol *attr_class_to_protocol[EAP_MAX];

436 437 438 439 440 441
/*
 *	Default protocol preferences
 */

#define DEF_PREF_DIRECT	    	240	/* Directly connected */
#define DEF_PREF_STATIC		200	/* Static route */
442
#define DEF_PREF_OSPF		150	/* OSPF intra-area, inter-area and type 1 external routes */
443 444
#define DEF_PREF_RIP		120	/* RIP */
#define DEF_PREF_BGP		100	/* BGP */
Martin Mareš's avatar
Martin Mareš committed
445
#define DEF_PREF_PIPE		70	/* Routes piped from other tables */
446
#define DEF_PREF_INHERITED	10	/* Routes inherited from other routing daemons */
447

448
#endif