]>
Commit | Line | Data |
---|---|---|
65efcfce LB |
1 | /* |
2 | * | |
3 | * Copyright 2009-2016, LabN Consulting, L.L.C. | |
4 | * | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License | |
8 | * as published by the Free Software Foundation; either version 2 | |
9 | * of the License, or (at your option) any later version. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * along with this program; if not, write to the Free Software | |
18 | * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. | |
19 | * | |
20 | */ | |
21 | ||
22 | /* | |
23 | * File: rfapi_import.c | |
24 | * Purpose: Handle import of routes from BGP to RFAPI | |
25 | */ | |
26 | ||
27 | #include <errno.h> | |
28 | ||
f8b6f499 LB |
29 | #include "lib/zebra.h" |
30 | #include "lib/prefix.h" | |
31 | #include "lib/table.h" | |
32 | #include "lib/vty.h" | |
33 | #include "lib/memory.h" | |
34 | #include "lib/log.h" | |
35 | #include "lib/skiplist.h" | |
36 | #include "lib/thread.h" | |
65efcfce | 37 | |
f8b6f499 LB |
38 | #include "bgpd/bgpd.h" |
39 | #include "bgpd/bgp_ecommunity.h" | |
40 | #include "bgpd/bgp_attr.h" | |
41 | #include "bgpd/bgp_route.h" | |
42 | #include "bgpd/bgp_mplsvpn.h" /* prefix_rd2str() */ | |
43 | #include "bgpd/bgp_vnc_types.h" | |
65efcfce | 44 | |
f8b6f499 LB |
45 | #include "bgpd/rfapi/rfapi.h" |
46 | #include "bgpd/rfapi/bgp_rfapi_cfg.h" | |
47 | #include "bgpd/rfapi/rfapi_backend.h" | |
48 | #include "bgpd/rfapi/rfapi_import.h" | |
49 | #include "bgpd/rfapi/rfapi_private.h" | |
50 | #include "bgpd/rfapi/rfapi_monitor.h" | |
51 | #include "bgpd/rfapi/rfapi_nve_addr.h" | |
52 | #include "bgpd/rfapi/rfapi_vty.h" | |
53 | #include "bgpd/rfapi/vnc_export_bgp.h" | |
54 | #include "bgpd/rfapi/vnc_export_bgp_p.h" | |
55 | #include "bgpd/rfapi/vnc_zebra.h" | |
56 | #include "bgpd/rfapi/vnc_import_bgp.h" | |
57 | #include "bgpd/rfapi/vnc_import_bgp_p.h" | |
58 | #include "bgpd/rfapi/rfapi_rib.h" | |
59 | #include "bgpd/rfapi/rfapi_encap_tlv.h" | |
60 | #include "bgpd/rfapi/vnc_debug.h" | |
65efcfce LB |
61 | |
62 | #ifdef HAVE_GLIBC_BACKTRACE | |
63 | /* for backtrace and friends */ | |
64 | #include <execinfo.h> | |
65 | #endif /* HAVE_GLIBC_BACKTRACE */ | |
66 | ||
67 | #undef DEBUG_MONITOR_MOVE_SHORTER | |
68 | #undef DEBUG_RETURNED_NHL | |
69 | #undef DEBUG_ROUTE_COUNTERS | |
70 | #undef DEBUG_ENCAP_MONITOR | |
71 | #undef DEBUG_L2_EXTRA | |
72 | #undef DEBUG_IT_NODES | |
73 | #undef DEBUG_BI_SEARCH | |
74 | ||
75 | /* | |
76 | * Allocated for each withdraw timer instance; freed when the timer | |
77 | * expires or is canceled | |
78 | */ | |
79 | struct rfapi_withdraw | |
80 | { | |
81 | struct rfapi_import_table *import_table; | |
82 | struct route_node *node; | |
83 | struct bgp_info *info; | |
84 | safi_t safi; /* used only for bulk operations */ | |
85 | /* | |
86 | * For import table node reference count checking (i.e., debugging). | |
87 | * Normally when a timer expires, lockoffset should be 0. However, if | |
88 | * the timer expiration function is called directly (e.g., | |
89 | * rfapiExpireVpnNow), the node could be locked by a preceding | |
90 | * route_top() or route_next() in a loop, so we need to pass this | |
91 | * value in. | |
92 | */ | |
93 | int lockoffset; | |
94 | }; | |
95 | ||
96 | /* | |
97 | * DEBUG FUNCTION | |
98 | * It's evil and fiendish. It's compiler-dependent. | |
99 | * ? Might need LDFLAGS -rdynamic to produce all function names | |
100 | */ | |
101 | void | |
102 | rfapiDebugBacktrace (void) | |
103 | { | |
104 | #ifdef HAVE_GLIBC_BACKTRACE | |
105 | #define RFAPI_DEBUG_BACKTRACE_NENTRIES 200 | |
106 | void *buf[RFAPI_DEBUG_BACKTRACE_NENTRIES]; | |
107 | char **syms; | |
108 | size_t i; | |
109 | size_t size; | |
110 | ||
111 | size = backtrace (buf, RFAPI_DEBUG_BACKTRACE_NENTRIES); | |
112 | syms = backtrace_symbols (buf, size); | |
113 | ||
114 | for (i = 0; i < size && i < RFAPI_DEBUG_BACKTRACE_NENTRIES; ++i) | |
115 | { | |
a3b55c25 | 116 | vnc_zlog_debug_verbose ("backtrace[%2lu]: %s", i, syms[i]); |
65efcfce LB |
117 | } |
118 | ||
119 | free (syms); | |
120 | #else | |
121 | #endif | |
122 | } | |
123 | ||
124 | /* | |
125 | * DEBUG FUNCTION | |
126 | * Count remote routes and compare with actively-maintained values. | |
127 | * Abort if they disagree. | |
128 | */ | |
129 | void | |
130 | rfapiCheckRouteCount () | |
131 | { | |
132 | struct bgp *bgp = bgp_get_default (); | |
133 | struct rfapi *h; | |
134 | struct rfapi_import_table *it; | |
135 | afi_t afi; | |
136 | ||
137 | assert (bgp); | |
138 | ||
139 | h = bgp->rfapi; | |
140 | assert (h); | |
141 | ||
142 | for (it = h->imports; it; it = it->next) | |
143 | { | |
144 | for (afi = AFI_IP; afi < AFI_MAX; ++afi) | |
145 | { | |
146 | ||
147 | struct route_table *rt; | |
148 | struct route_node *rn; | |
149 | ||
150 | int holddown_count = 0; | |
151 | int local_count = 0; | |
152 | int imported_count = 0; | |
153 | int remote_count = 0; | |
154 | ||
155 | rt = it->imported_vpn[afi]; | |
156 | ||
157 | for (rn = route_top (rt); rn; rn = route_next (rn)) | |
158 | { | |
159 | struct bgp_info *bi; | |
160 | struct bgp_info *next; | |
161 | ||
162 | for (bi = rn->info; bi; bi = next) | |
163 | { | |
164 | next = bi->next; | |
165 | ||
166 | if (CHECK_FLAG (bi->flags, BGP_INFO_REMOVED)) | |
167 | { | |
168 | ++holddown_count; | |
169 | ||
170 | } | |
171 | else | |
172 | { | |
173 | if (RFAPI_LOCAL_BI (bi)) | |
174 | { | |
175 | ++local_count; | |
176 | } | |
177 | else | |
178 | { | |
179 | if (RFAPI_DIRECT_IMPORT_BI (bi)) | |
180 | { | |
181 | ++imported_count; | |
182 | } | |
183 | else | |
184 | { | |
185 | ++remote_count; | |
186 | } | |
187 | } | |
188 | } | |
189 | } | |
190 | } | |
191 | ||
192 | if (it->holddown_count[afi] != holddown_count) | |
193 | { | |
a3b55c25 | 194 | vnc_zlog_debug_verbose ("%s: it->holddown_count %d != holddown_count %d", |
65efcfce LB |
195 | __func__, it->holddown_count[afi], holddown_count); |
196 | assert (0); | |
197 | } | |
198 | if (it->remote_count[afi] != remote_count) | |
199 | { | |
a3b55c25 | 200 | vnc_zlog_debug_verbose ("%s: it->remote_count %d != remote_count %d", |
65efcfce LB |
201 | __func__, it->remote_count[afi], remote_count); |
202 | assert (0); | |
203 | } | |
204 | if (it->imported_count[afi] != imported_count) | |
205 | { | |
a3b55c25 | 206 | vnc_zlog_debug_verbose ("%s: it->imported_count %d != imported_count %d", |
65efcfce LB |
207 | __func__, it->imported_count[afi], imported_count); |
208 | assert (0); | |
209 | } | |
210 | } | |
211 | } | |
212 | } | |
213 | ||
214 | #if DEBUG_ROUTE_COUNTERS | |
215 | #define VNC_ITRCCK do {rfapiCheckRouteCount();} while (0) | |
216 | #else | |
217 | #define VNC_ITRCCK | |
218 | #endif | |
219 | ||
220 | /* | |
221 | * Validate reference count for a node in an import table | |
222 | * | |
223 | * Normally lockoffset is 0 for nodes in quiescent state. However, | |
224 | * route_unlock_node will delete the node if it is called when | |
225 | * node->lock == 1, and we have to validate the refcount before | |
226 | * the node is deleted. In this case, we specify lockoffset 1. | |
227 | */ | |
228 | void | |
229 | rfapiCheckRefcount (struct route_node *rn, safi_t safi, int lockoffset) | |
230 | { | |
231 | unsigned int count_bi = 0; | |
232 | unsigned int count_monitor = 0; | |
233 | struct bgp_info *bi; | |
234 | struct rfapi_monitor_encap *hme; | |
235 | struct rfapi_monitor_vpn *hmv; | |
236 | ||
237 | for (bi = rn->info; bi; bi = bi->next) | |
238 | ++count_bi; | |
239 | ||
240 | ||
241 | if (rn->aggregate) | |
242 | { | |
243 | ++count_monitor; /* rfapi_it_extra */ | |
244 | ||
245 | switch (safi) | |
246 | { | |
247 | void *cursor; | |
248 | int rc; | |
249 | ||
250 | case SAFI_ENCAP: | |
251 | for (hme = RFAPI_MONITOR_ENCAP (rn); hme; hme = hme->next) | |
252 | ++count_monitor; | |
253 | break; | |
254 | ||
255 | case SAFI_MPLS_VPN: | |
256 | ||
257 | for (hmv = RFAPI_MONITOR_VPN (rn); hmv; hmv = hmv->next) | |
258 | ++count_monitor; | |
259 | ||
260 | if (RFAPI_MONITOR_EXTERIOR (rn)->source) | |
261 | { | |
262 | ++count_monitor; /* sl */ | |
263 | cursor = NULL; | |
264 | for (rc = skiplist_next (RFAPI_MONITOR_EXTERIOR (rn)->source, | |
265 | NULL, NULL, &cursor); | |
266 | !rc; | |
267 | rc = skiplist_next (RFAPI_MONITOR_EXTERIOR (rn)->source, | |
268 | NULL, NULL, &cursor)) | |
269 | { | |
270 | ||
271 | ++count_monitor; /* sl entry */ | |
272 | } | |
273 | } | |
274 | break; | |
275 | ||
276 | default: | |
277 | assert (0); | |
278 | } | |
279 | } | |
280 | ||
281 | if (count_bi + count_monitor + lockoffset != rn->lock) | |
282 | { | |
a3b55c25 | 283 | vnc_zlog_debug_verbose |
65efcfce LB |
284 | ("%s: count_bi=%d, count_monitor=%d, lockoffset=%d, rn->lock=%d", |
285 | __func__, count_bi, count_monitor, lockoffset, rn->lock); | |
286 | assert (0); | |
287 | } | |
288 | } | |
289 | ||
290 | /* | |
291 | * Perform deferred rfapi_close operations that were queued | |
292 | * during callbacks. | |
293 | */ | |
294 | static wq_item_status | |
295 | rfapi_deferred_close_workfunc (struct work_queue *q, void *data) | |
296 | { | |
297 | struct rfapi_descriptor *rfd = data; | |
298 | struct rfapi *h = q->spec.data; | |
299 | ||
300 | assert (!(h->flags & RFAPI_INCALLBACK)); | |
301 | rfapi_close (rfd); | |
a3b55c25 | 302 | vnc_zlog_debug_verbose ("%s: completed deferred close on handle %p", __func__, rfd); |
65efcfce LB |
303 | return WQ_SUCCESS; |
304 | } | |
305 | ||
306 | /* | |
307 | * Extract layer 2 option from Encap TLVS in BGP attrs | |
308 | */ | |
309 | int | |
310 | rfapiGetL2o (struct attr *attr, struct rfapi_l2address_option *l2o) | |
311 | { | |
312 | if (attr && attr->extra) | |
313 | { | |
314 | ||
315 | struct bgp_attr_encap_subtlv *pEncap; | |
316 | ||
317 | for (pEncap = attr->extra->vnc_subtlvs; pEncap; pEncap = pEncap->next) | |
318 | { | |
319 | ||
320 | if (pEncap->type == BGP_VNC_SUBTLV_TYPE_RFPOPTION) | |
321 | { | |
322 | if (pEncap->value[0] == RFAPI_VN_OPTION_TYPE_L2ADDR) | |
323 | { | |
324 | ||
325 | if (pEncap->value[1] == 14) | |
326 | { | |
327 | memcpy (l2o->macaddr.octet, pEncap->value + 2, | |
328 | ETHER_ADDR_LEN); | |
329 | l2o->label = | |
330 | ((pEncap->value[10] >> 4) & 0x0f) + | |
331 | ((pEncap->value[9] << 4) & 0xff0) + | |
332 | ((pEncap->value[8] << 12) & 0xff000); | |
333 | ||
334 | l2o->local_nve_id = pEncap->value[12]; | |
335 | ||
336 | l2o->logical_net_id = | |
337 | (pEncap->value[15] & 0xff) + | |
338 | ((pEncap->value[14] << 8) & 0xff00) + | |
339 | ((pEncap->value[13] << 16) & 0xff0000); | |
340 | } | |
341 | ||
342 | return 0; | |
343 | } | |
344 | } | |
345 | } | |
346 | } | |
347 | ||
348 | return ENOENT; | |
349 | } | |
350 | ||
351 | /* | |
352 | * Extract the lifetime from the Tunnel Encap attribute of a route in | |
353 | * an import table | |
354 | */ | |
355 | int | |
356 | rfapiGetVncLifetime (struct attr *attr, uint32_t * lifetime) | |
357 | { | |
358 | struct bgp_attr_encap_subtlv *pEncap; | |
359 | ||
360 | *lifetime = RFAPI_INFINITE_LIFETIME; /* default to infinite */ | |
361 | ||
362 | if (attr && attr->extra) | |
363 | { | |
364 | ||
365 | for (pEncap = attr->extra->vnc_subtlvs; pEncap; pEncap = pEncap->next) | |
366 | { | |
367 | ||
368 | if (pEncap->type == BGP_VNC_SUBTLV_TYPE_LIFETIME) | |
369 | { /* lifetime */ | |
370 | if (pEncap->length == 4) | |
371 | { | |
372 | memcpy (lifetime, pEncap->value, 4); | |
373 | *lifetime = ntohl (*lifetime); | |
374 | return 0; | |
375 | } | |
376 | } | |
377 | } | |
378 | } | |
379 | ||
380 | return ENOENT; | |
381 | } | |
382 | ||
383 | /* | |
384 | * Extract the tunnel type from the extended community | |
385 | */ | |
386 | int | |
387 | rfapiGetTunnelType (struct attr *attr, | |
388 | bgp_encap_types *type) | |
389 | { | |
390 | *type = BGP_ENCAP_TYPE_MPLS; /* default to MPLS */ | |
391 | if (attr && attr->extra && attr->extra->ecommunity) | |
392 | { | |
393 | struct ecommunity *ecom = attr->extra->ecommunity; | |
394 | int i; | |
395 | ||
396 | for (i = 0; i < (ecom->size * ECOMMUNITY_SIZE); i += ECOMMUNITY_SIZE) | |
397 | { | |
398 | uint8_t *ep; | |
399 | ||
400 | ep = ecom->val + i; | |
401 | if (ep[0] == ECOMMUNITY_ENCODE_OPAQUE && | |
402 | ep[1] == ECOMMUNITY_OPAQUE_SUBTYPE_ENCAP) | |
403 | { | |
404 | *type = (ep[6]<<8) + ep[7]; | |
405 | return 0; | |
406 | } | |
407 | } | |
408 | } | |
409 | ||
410 | return ENOENT; | |
411 | } | |
412 | ||
413 | ||
414 | /* | |
415 | * Look for UN address in Encap attribute | |
416 | */ | |
417 | int | |
418 | rfapiGetVncTunnelUnAddr (struct attr *attr, struct prefix *p) | |
419 | { | |
420 | struct bgp_attr_encap_subtlv *pEncap; | |
421 | bgp_encap_types tun_type; | |
422 | ||
423 | rfapiGetTunnelType (attr, &tun_type); | |
424 | if (p && tun_type == BGP_ENCAP_TYPE_MPLS) | |
425 | { | |
3eebadeb | 426 | return ENOENT; /* no UN for MPLS */ |
65efcfce LB |
427 | } |
428 | if (attr && attr->extra) | |
429 | { | |
430 | for (pEncap = attr->extra->encap_subtlvs; pEncap; pEncap = pEncap->next) | |
431 | { | |
432 | ||
433 | if (pEncap->type == BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT) | |
434 | { /* un addr */ | |
435 | switch (pEncap->length) | |
436 | { | |
437 | case 8: | |
438 | if (p) | |
439 | { | |
440 | p->family = AF_INET; | |
441 | p->prefixlen = 32; | |
442 | memcpy (p->u.val, pEncap->value, 4); | |
443 | } | |
444 | return 0; | |
445 | ||
446 | case 20: | |
447 | if (p) | |
448 | { | |
449 | p->family = AF_INET6; | |
450 | p->prefixlen = 128; | |
451 | memcpy (p->u.val, pEncap->value, 16); | |
452 | } | |
453 | return 0; | |
454 | } | |
455 | } | |
456 | } | |
457 | } | |
458 | ||
459 | return ENOENT; | |
460 | } | |
461 | ||
462 | /* | |
463 | * Get UN address wherever it might be | |
464 | */ | |
465 | int | |
466 | rfapiGetUnAddrOfVpnBi (struct bgp_info *bi, struct prefix *p) | |
467 | { | |
468 | /* If it's in this route's VNC attribute, we're done */ | |
469 | if (!rfapiGetVncTunnelUnAddr (bi->attr, p)) | |
470 | return 0; | |
471 | /* | |
472 | * Otherwise, see if it's cached from a corresponding ENCAP SAFI | |
473 | * advertisement | |
474 | */ | |
475 | if (bi->extra) | |
476 | { | |
477 | switch (bi->extra->vnc.import.un_family) | |
478 | { | |
479 | case AF_INET: | |
480 | if (p) | |
481 | { | |
482 | p->family = bi->extra->vnc.import.un_family; | |
483 | p->u.prefix4 = bi->extra->vnc.import.un.addr4; | |
484 | p->prefixlen = 32; | |
485 | } | |
486 | return 0; | |
487 | case AF_INET6: | |
488 | if (p) | |
489 | { | |
490 | p->family = bi->extra->vnc.import.un_family; | |
491 | p->u.prefix6 = bi->extra->vnc.import.un.addr6; | |
492 | p->prefixlen = 128; | |
493 | } | |
494 | return 0; | |
495 | default: | |
496 | if (p) | |
497 | p->family = 0; | |
498 | #if DEBUG_ENCAP_MONITOR | |
a3b55c25 | 499 | vnc_zlog_debug_verbose ("%s: bi->extra->vnc.import.un_family is 0, no UN addr", |
65efcfce LB |
500 | __func__); |
501 | #endif | |
502 | break; | |
503 | } | |
504 | } | |
505 | ||
506 | return ENOENT; | |
507 | } | |
508 | ||
509 | ||
510 | /* | |
511 | * Make a new bgp_info from gathered parameters | |
512 | */ | |
513 | static struct bgp_info * | |
514 | rfapiBgpInfoCreate ( | |
515 | struct attr *attr, | |
516 | struct peer *peer, | |
517 | void *rfd, | |
518 | struct prefix_rd *prd, | |
519 | u_char type, | |
520 | u_char sub_type, | |
521 | uint32_t *label) | |
522 | { | |
523 | struct bgp_info *new; | |
524 | ||
525 | new = bgp_info_new (); | |
526 | assert (new); | |
527 | ||
528 | if (attr) | |
529 | { | |
530 | if (!new->attr) | |
531 | new->attr = bgp_attr_intern (attr); | |
532 | } | |
533 | bgp_info_extra_get (new); | |
534 | if (prd) | |
535 | { | |
536 | new->extra->vnc.import.rd = *prd; | |
537 | rfapi_time (&new->extra->vnc.import.create_time); | |
538 | } | |
539 | if (label) | |
540 | encode_label (*label, new->extra->tag); | |
541 | new->type = type; | |
542 | new->sub_type = sub_type; | |
543 | new->peer = peer; | |
544 | peer_lock (peer); | |
545 | ||
546 | return new; | |
547 | } | |
548 | ||
549 | /* | |
550 | * Frees bgp_info as used in import tables (parts are not | |
551 | * allocated exactly the way they are in the main RIBs) | |
552 | */ | |
553 | static void | |
554 | rfapiBgpInfoFree (struct bgp_info *goner) | |
555 | { | |
556 | if (!goner) | |
557 | return; | |
558 | ||
559 | if (goner->peer) | |
560 | { | |
a3b55c25 | 561 | vnc_zlog_debug_verbose ("%s: calling peer_unlock(%p), #%d", |
65efcfce LB |
562 | __func__, goner->peer, goner->peer->lock); |
563 | peer_unlock (goner->peer); | |
564 | } | |
565 | ||
566 | if (goner->attr) | |
567 | { | |
568 | bgp_attr_unintern (&goner->attr); | |
569 | } | |
570 | if (goner->extra) | |
571 | { | |
572 | assert (!goner->extra->damp_info); /* Not used in import tbls */ | |
573 | XFREE (MTYPE_BGP_ROUTE_EXTRA, goner->extra); | |
574 | goner->extra = NULL; | |
575 | } | |
576 | XFREE (MTYPE_BGP_ROUTE, goner); | |
577 | } | |
578 | ||
579 | struct rfapi_import_table * | |
580 | rfapiMacImportTableGetNoAlloc (struct bgp *bgp, uint32_t lni) | |
581 | { | |
582 | struct rfapi *h; | |
583 | struct rfapi_import_table *it = NULL; | |
584 | uintptr_t lni_as_ptr = lni; | |
585 | ||
586 | h = bgp->rfapi; | |
587 | if (!h) | |
588 | return NULL; | |
589 | ||
590 | if (!h->import_mac) | |
591 | return NULL; | |
592 | ||
593 | if (skiplist_search (h->import_mac, (void *) lni_as_ptr, (void **) &it)) | |
594 | return NULL; | |
595 | ||
596 | return it; | |
597 | } | |
598 | ||
599 | struct rfapi_import_table * | |
600 | rfapiMacImportTableGet (struct bgp *bgp, uint32_t lni) | |
601 | { | |
602 | struct rfapi *h; | |
603 | struct rfapi_import_table *it = NULL; | |
604 | uintptr_t lni_as_ptr = lni; | |
605 | ||
606 | h = bgp->rfapi; | |
607 | assert (h); | |
608 | ||
609 | if (!h->import_mac) | |
610 | { | |
611 | /* default cmp is good enough for LNI */ | |
612 | h->import_mac = skiplist_new (0, NULL, NULL); | |
613 | } | |
614 | ||
615 | if (skiplist_search (h->import_mac, (void *) lni_as_ptr, (void **) &it)) | |
616 | { | |
617 | ||
618 | struct ecommunity *enew; | |
619 | struct ecommunity_val eval; | |
620 | afi_t afi; | |
621 | ||
622 | it = | |
623 | XCALLOC (MTYPE_RFAPI_IMPORTTABLE, sizeof (struct rfapi_import_table)); | |
624 | /* set RT list of new import table based on LNI */ | |
625 | memset ((char *) &eval, 0, sizeof (eval)); | |
626 | eval.val[0] = 0; /* VNC L2VPN */ | |
627 | eval.val[1] = 2; /* VNC L2VPN */ | |
628 | eval.val[5] = (lni >> 16) & 0xff; | |
629 | eval.val[6] = (lni >> 8) & 0xff; | |
630 | eval.val[7] = (lni >> 0) & 0xff; | |
631 | ||
632 | enew = ecommunity_new (); | |
633 | ecommunity_add_val (enew, &eval); | |
634 | it->rt_import_list = enew; | |
635 | ||
636 | for (afi = AFI_IP; afi < AFI_MAX; ++afi) | |
637 | { | |
638 | it->imported_vpn[afi] = route_table_init (); | |
639 | it->imported_encap[afi] = route_table_init (); | |
640 | } | |
641 | ||
642 | it->l2_logical_net_id = lni; | |
643 | ||
644 | skiplist_insert (h->import_mac, (void *) lni_as_ptr, it); | |
645 | } | |
646 | ||
647 | assert (it); | |
648 | return it; | |
649 | } | |
650 | ||
651 | /* | |
652 | * Implement MONITOR_MOVE_SHORTER(original_node) from | |
653 | * RFAPI-Import-Event-Handling.txt | |
654 | * | |
655 | * Returns pointer to the list of moved monitors | |
656 | */ | |
657 | static struct rfapi_monitor_vpn * | |
658 | rfapiMonitorMoveShorter (struct route_node *original_vpn_node, int lockoffset) | |
659 | { | |
660 | struct bgp_info *bi; | |
661 | struct route_node *par; | |
662 | struct rfapi_monitor_vpn *m; | |
663 | struct rfapi_monitor_vpn *mlast; | |
664 | struct rfapi_monitor_vpn *moved; | |
665 | int movecount = 0; | |
666 | int parent_already_refcounted = 0; | |
667 | ||
668 | RFAPI_CHECK_REFCOUNT (original_vpn_node, SAFI_MPLS_VPN, lockoffset); | |
669 | ||
670 | #if DEBUG_MONITOR_MOVE_SHORTER | |
671 | { | |
672 | char buf[BUFSIZ]; | |
673 | ||
674 | prefix2str (&original_vpn_node->p, buf, BUFSIZ); | |
675 | buf[BUFSIZ - 1] = 0; | |
a3b55c25 | 676 | vnc_zlog_debug_verbose ("%s: called with node pfx=%s", __func__, buf); |
65efcfce LB |
677 | } |
678 | #endif | |
679 | ||
680 | /* | |
681 | * 1. If there is at least one bi (either regular route or | |
682 | * route marked as withdrawn, with a pending timer) at | |
683 | * original_node with a valid UN address, we're done. Return. | |
684 | */ | |
685 | for (bi = original_vpn_node->info; bi; bi = bi->next) | |
686 | { | |
687 | struct prefix pfx; | |
688 | ||
689 | if (!rfapiGetUnAddrOfVpnBi (bi, &pfx)) | |
690 | { | |
691 | #if DEBUG_MONITOR_MOVE_SHORTER | |
a3b55c25 | 692 | vnc_zlog_debug_verbose ("%s: have valid UN at original node, no change", |
65efcfce LB |
693 | __func__); |
694 | #endif | |
695 | return NULL; | |
696 | } | |
697 | } | |
698 | ||
699 | /* | |
700 | * 2. Travel up the tree (toward less-specific prefixes) from | |
701 | * original_node to find the first node that has at least | |
702 | * one route (even if it is only a withdrawn route) with a | |
703 | * valid UN address. Call this node "Node P." | |
704 | */ | |
705 | for (par = original_vpn_node->parent; par; par = par->parent) | |
706 | { | |
707 | for (bi = par->info; bi; bi = bi->next) | |
708 | { | |
709 | struct prefix pfx; | |
710 | if (!rfapiGetUnAddrOfVpnBi (bi, &pfx)) | |
711 | { | |
712 | break; | |
713 | } | |
714 | } | |
715 | if (bi) | |
716 | break; | |
717 | } | |
718 | ||
719 | if (par) | |
720 | { | |
721 | RFAPI_CHECK_REFCOUNT (par, SAFI_MPLS_VPN, 0); | |
722 | } | |
723 | ||
724 | /* | |
725 | * If no less-specific routes, try to use the 0/0 node | |
726 | */ | |
727 | if (!par) | |
728 | { | |
729 | /* this isn't necessarily 0/0 */ | |
730 | par = route_top (original_vpn_node->table); | |
731 | ||
732 | /* | |
733 | * If we got the top node but it wasn't 0/0, | |
734 | * ignore it | |
735 | */ | |
736 | if (par && par->p.prefixlen) | |
737 | { | |
738 | route_unlock_node (par); /* maybe free */ | |
739 | par = NULL; | |
740 | } | |
741 | ||
742 | if (par) | |
743 | { | |
744 | ++parent_already_refcounted; | |
745 | } | |
746 | } | |
747 | ||
748 | /* | |
749 | * Create 0/0 node if it isn't there | |
750 | */ | |
751 | if (!par) | |
752 | { | |
753 | struct prefix pfx_default; | |
754 | ||
755 | memset (&pfx_default, 0, sizeof (pfx_default)); | |
756 | pfx_default.family = original_vpn_node->p.family; | |
757 | ||
758 | /* creates default node if none exists */ | |
759 | par = route_node_get (original_vpn_node->table, &pfx_default); | |
760 | ++parent_already_refcounted; | |
761 | } | |
762 | ||
763 | /* | |
764 | * 3. Move each of the monitors found at original_node to Node P. | |
765 | * These are "Moved Monitors." | |
766 | * | |
767 | */ | |
768 | ||
769 | /* | |
770 | * Attach at end so that the list pointer we return points | |
771 | * only to the moved routes | |
772 | */ | |
773 | for (m = RFAPI_MONITOR_VPN (par), mlast = NULL; m; mlast = m, m = m->next); | |
774 | ||
775 | if (mlast) | |
776 | { | |
777 | moved = mlast->next = RFAPI_MONITOR_VPN (original_vpn_node); | |
778 | } | |
779 | else | |
780 | { | |
781 | moved = RFAPI_MONITOR_VPN_W_ALLOC (par) = | |
782 | RFAPI_MONITOR_VPN (original_vpn_node); | |
783 | } | |
784 | if (RFAPI_MONITOR_VPN (original_vpn_node)) /* check agg, so not allocated */ | |
785 | RFAPI_MONITOR_VPN_W_ALLOC (original_vpn_node) = NULL; | |
786 | ||
787 | /* | |
788 | * update the node pointers on the monitors | |
789 | */ | |
790 | for (m = moved; m; m = m->next) | |
791 | { | |
792 | ++movecount; | |
793 | m->node = par; | |
794 | } | |
795 | ||
796 | RFAPI_CHECK_REFCOUNT (par, SAFI_MPLS_VPN, | |
797 | parent_already_refcounted - movecount); | |
798 | while (movecount > parent_already_refcounted) | |
799 | { | |
800 | route_lock_node (par); | |
801 | ++parent_already_refcounted; | |
802 | } | |
803 | while (movecount < parent_already_refcounted) | |
804 | { | |
805 | /* unlikely, but code defensively */ | |
806 | route_unlock_node (par); | |
807 | --parent_already_refcounted; | |
808 | } | |
809 | RFAPI_CHECK_REFCOUNT (original_vpn_node, SAFI_MPLS_VPN, | |
810 | movecount + lockoffset); | |
811 | while (movecount--) | |
812 | { | |
813 | route_unlock_node (original_vpn_node); | |
814 | } | |
815 | ||
816 | #if DEBUG_MONITOR_MOVE_SHORTER | |
817 | { | |
818 | char buf[BUFSIZ]; | |
819 | ||
820 | prefix2str (&par->p, buf, BUFSIZ); | |
821 | buf[BUFSIZ - 1] = 0; | |
a3b55c25 | 822 | vnc_zlog_debug_verbose ("%s: moved to node pfx=%s", __func__, buf); |
65efcfce LB |
823 | } |
824 | #endif | |
825 | ||
826 | ||
827 | return moved; | |
828 | } | |
829 | ||
830 | /* | |
831 | * Implement MONITOR_MOVE_LONGER(new_node) from | |
832 | * RFAPI-Import-Event-Handling.txt | |
833 | */ | |
834 | static void | |
835 | rfapiMonitorMoveLonger (struct route_node *new_vpn_node) | |
836 | { | |
837 | struct rfapi_monitor_vpn *monitor; | |
838 | struct rfapi_monitor_vpn *mlast; | |
839 | struct bgp_info *bi; | |
840 | struct route_node *par; | |
841 | ||
842 | RFAPI_CHECK_REFCOUNT (new_vpn_node, SAFI_MPLS_VPN, 0); | |
843 | ||
844 | /* | |
845 | * Make sure we have at least one valid route at the new node | |
846 | */ | |
847 | for (bi = new_vpn_node->info; bi; bi = bi->next) | |
848 | { | |
849 | struct prefix pfx; | |
850 | if (!rfapiGetUnAddrOfVpnBi (bi, &pfx)) | |
851 | break; | |
852 | } | |
853 | ||
854 | if (!bi) | |
855 | { | |
a3b55c25 | 856 | vnc_zlog_debug_verbose ("%s: no valid routes at node %p, so not attempting moves", |
65efcfce LB |
857 | __func__, new_vpn_node); |
858 | return; | |
859 | } | |
860 | ||
861 | /* | |
862 | * Find first parent node that has monitors | |
863 | */ | |
864 | for (par = new_vpn_node->parent; par; par = par->parent) | |
865 | { | |
866 | if (RFAPI_MONITOR_VPN (par)) | |
867 | break; | |
868 | } | |
869 | ||
870 | if (!par) | |
871 | { | |
a3b55c25 | 872 | vnc_zlog_debug_verbose ("%s: no parent nodes with monitors, done", __func__); |
65efcfce LB |
873 | return; |
874 | } | |
875 | ||
876 | /* | |
877 | * Check each of these monitors to see of their longest-match | |
878 | * is now the updated node. Move any such monitors to the more- | |
879 | * specific updated node | |
880 | */ | |
881 | for (mlast = NULL, monitor = RFAPI_MONITOR_VPN (par); monitor;) | |
882 | { | |
883 | ||
884 | /* | |
885 | * If new longest match for monitor prefix is the new | |
886 | * route's prefix, move monitor to new route's prefix | |
887 | */ | |
888 | if (prefix_match (&new_vpn_node->p, &monitor->p)) | |
889 | { | |
890 | /* detach */ | |
891 | if (mlast) | |
892 | { | |
893 | mlast->next = monitor->next; | |
894 | } | |
895 | else | |
896 | { | |
897 | RFAPI_MONITOR_VPN_W_ALLOC (par) = monitor->next; | |
898 | } | |
899 | ||
900 | ||
901 | /* attach */ | |
902 | monitor->next = RFAPI_MONITOR_VPN (new_vpn_node); | |
903 | RFAPI_MONITOR_VPN_W_ALLOC (new_vpn_node) = monitor; | |
904 | monitor->node = new_vpn_node; | |
905 | ||
906 | route_lock_node (new_vpn_node); /* incr refcount */ | |
907 | ||
908 | monitor = mlast ? mlast->next : RFAPI_MONITOR_VPN (par); | |
909 | ||
910 | RFAPI_CHECK_REFCOUNT (par, SAFI_MPLS_VPN, 1); | |
911 | /* decr refcount after we're done with par as this might free it */ | |
912 | route_unlock_node (par); | |
913 | ||
914 | continue; | |
915 | } | |
916 | mlast = monitor; | |
917 | monitor = monitor->next; | |
918 | } | |
919 | ||
920 | RFAPI_CHECK_REFCOUNT (new_vpn_node, SAFI_MPLS_VPN, 0); | |
921 | } | |
922 | ||
923 | ||
924 | static void | |
925 | rfapiBgpInfoChainFree (struct bgp_info *bi) | |
926 | { | |
927 | struct bgp_info *next; | |
928 | ||
929 | while (bi) | |
930 | { | |
931 | ||
932 | /* | |
933 | * If there is a timer waiting to delete this bi, cancel | |
934 | * the timer and delete immediately | |
935 | */ | |
936 | if (CHECK_FLAG (bi->flags, BGP_INFO_REMOVED) && | |
937 | bi->extra->vnc.import.timer) | |
938 | { | |
939 | ||
940 | struct thread *t = (struct thread *) bi->extra->vnc.import.timer; | |
941 | struct rfapi_withdraw *wcb = t->arg; | |
942 | ||
943 | XFREE (MTYPE_RFAPI_WITHDRAW, wcb); | |
944 | thread_cancel (t); | |
945 | } | |
946 | ||
947 | next = bi->next; | |
948 | bi->next = NULL; | |
949 | rfapiBgpInfoFree (bi); | |
950 | bi = next; | |
951 | } | |
952 | } | |
953 | ||
954 | static void | |
955 | rfapiImportTableFlush (struct rfapi_import_table *it) | |
956 | { | |
957 | afi_t afi; | |
958 | ||
959 | /* | |
960 | * Free ecommunity | |
961 | */ | |
962 | ecommunity_free (&it->rt_import_list); | |
963 | it->rt_import_list = NULL; | |
964 | ||
965 | for (afi = AFI_IP; afi < AFI_MAX; ++afi) | |
966 | { | |
967 | ||
968 | struct route_node *rn; | |
969 | ||
970 | for (rn = route_top (it->imported_vpn[afi]); rn; rn = route_next (rn)) | |
971 | { | |
972 | /* | |
973 | * Each route_node has: | |
974 | * aggregate: points to rfapi_it_extra with monitor chain(s) | |
975 | * info: points to chain of bgp_info | |
976 | */ | |
977 | /* free bgp_info and its children */ | |
978 | rfapiBgpInfoChainFree (rn->info); | |
979 | rn->info = NULL; | |
980 | ||
981 | rfapiMonitorExtraFlush (SAFI_MPLS_VPN, rn); | |
982 | } | |
983 | ||
984 | for (rn = route_top (it->imported_encap[afi]); rn; rn = route_next (rn)) | |
985 | { | |
986 | /* free bgp_info and its children */ | |
987 | rfapiBgpInfoChainFree (rn->info); | |
988 | rn->info = NULL; | |
989 | ||
990 | rfapiMonitorExtraFlush (SAFI_ENCAP, rn); | |
991 | } | |
992 | ||
993 | route_table_finish (it->imported_vpn[afi]); | |
994 | route_table_finish (it->imported_encap[afi]); | |
995 | } | |
996 | if (it->monitor_exterior_orphans) | |
997 | { | |
998 | skiplist_free (it->monitor_exterior_orphans); | |
999 | } | |
1000 | } | |
1001 | ||
1002 | void | |
1003 | rfapiImportTableRefDelByIt ( | |
1004 | struct bgp *bgp, | |
1005 | struct rfapi_import_table *it_target) | |
1006 | { | |
1007 | struct rfapi *h; | |
1008 | struct rfapi_import_table *it; | |
1009 | struct rfapi_import_table *prev = NULL; | |
1010 | ||
1011 | assert (it_target); | |
1012 | ||
1013 | h = bgp->rfapi; | |
1014 | assert (h); | |
1015 | ||
1016 | for (it = h->imports; it; prev = it, it = it->next) | |
1017 | { | |
1018 | if (it == it_target) | |
1019 | break; | |
1020 | } | |
1021 | ||
1022 | assert (it); | |
1023 | assert (it->refcount); | |
1024 | ||
1025 | it->refcount -= 1; | |
1026 | ||
1027 | if (!it->refcount) | |
1028 | { | |
1029 | if (prev) | |
1030 | { | |
1031 | prev->next = it->next; | |
1032 | } | |
1033 | else | |
1034 | { | |
1035 | h->imports = it->next; | |
1036 | } | |
1037 | rfapiImportTableFlush (it); | |
1038 | XFREE (MTYPE_RFAPI_IMPORTTABLE, it); | |
1039 | } | |
1040 | } | |
1041 | ||
1042 | #if RFAPI_REQUIRE_ENCAP_BEEC | |
1043 | /* | |
1044 | * Look for magic BGP Encapsulation Extended Community value | |
1045 | * Format in RFC 5512 Sect. 4.5 | |
1046 | */ | |
1047 | static int | |
1048 | rfapiEcommunitiesMatchBeec (struct ecommunity *ecom, | |
1049 | bgp_encap_types type) | |
1050 | { | |
1051 | int i; | |
1052 | ||
1053 | if (!ecom) | |
1054 | return 0; | |
1055 | ||
1056 | for (i = 0; i < (ecom->size * ECOMMUNITY_SIZE); i += ECOMMUNITY_SIZE) | |
1057 | { | |
1058 | ||
1059 | uint8_t *ep; | |
1060 | ||
1061 | ep = ecom->val + i; | |
1062 | ||
1063 | if (ep[0] == ECOMMUNITY_ENCODE_OPAQUE && | |
1064 | ep[1] == ECOMMUNITY_OPAQUE_SUBTYPE_ENCAP && | |
1065 | ep[6] == ((type && 0xff00)>>8) && | |
1066 | ep[7] == (type&0xff)) | |
1067 | { | |
1068 | ||
1069 | return 1; | |
1070 | } | |
1071 | } | |
1072 | return 0; | |
1073 | ||
1074 | } | |
1075 | #endif | |
1076 | ||
1077 | int | |
1078 | rfapiEcommunitiesIntersect (struct ecommunity *e1, struct ecommunity *e2) | |
1079 | { | |
1080 | int i, j; | |
1081 | ||
1082 | if (!e1 || !e2) | |
1083 | return 0; | |
1084 | ||
1085 | { | |
1086 | char *s1, *s2; | |
1087 | s1 = ecommunity_ecom2str (e1, ECOMMUNITY_FORMAT_DISPLAY); | |
1088 | s2 = ecommunity_ecom2str (e2, ECOMMUNITY_FORMAT_DISPLAY); | |
a3b55c25 | 1089 | vnc_zlog_debug_verbose ("%s: e1[%s], e2[%s]", __func__, s1, s2); |
65efcfce LB |
1090 | XFREE (MTYPE_ECOMMUNITY_STR, s1); |
1091 | XFREE (MTYPE_ECOMMUNITY_STR, s2); | |
1092 | } | |
1093 | ||
1094 | for (i = 0; i < e1->size; ++i) | |
1095 | { | |
1096 | for (j = 0; j < e2->size; ++j) | |
1097 | { | |
1098 | if (!memcmp (e1->val + (i * ECOMMUNITY_SIZE), | |
1099 | e2->val + (j * ECOMMUNITY_SIZE), ECOMMUNITY_SIZE)) | |
1100 | { | |
1101 | ||
1102 | return 1; | |
1103 | } | |
1104 | } | |
1105 | } | |
1106 | return 0; | |
1107 | } | |
1108 | ||
1109 | int | |
1110 | rfapiEcommunityGetLNI (struct ecommunity *ecom, uint32_t * lni) | |
1111 | { | |
1112 | if (ecom) | |
1113 | { | |
1114 | int i; | |
1115 | for (i = 0; i < ecom->size; ++i) | |
1116 | { | |
1117 | uint8_t *p = ecom->val + (i * ECOMMUNITY_SIZE); | |
1118 | ||
1119 | if ((*(p + 0) == 0x00) && (*(p + 1) == 0x02)) | |
1120 | { | |
1121 | ||
1122 | *lni = (*(p + 5) << 16) | (*(p + 6) << 8) | (*(p + 7)); | |
1123 | return 0; | |
1124 | } | |
1125 | } | |
1126 | } | |
1127 | return ENOENT; | |
1128 | } | |
1129 | ||
79799987 LB |
1130 | int |
1131 | rfapiEcommunityGetEthernetTag (struct ecommunity *ecom, uint16_t * tag_id) | |
1132 | { | |
1133 | struct bgp *bgp = bgp_get_default (); | |
1134 | *tag_id = 0; /* default to untagged */ | |
1135 | if (ecom) | |
1136 | { | |
1137 | int i; | |
1138 | for (i = 0; i < ecom->size; ++i) | |
1139 | { | |
1140 | as_t as = 0; | |
1141 | int encode = 0; | |
1142 | uint8_t *p = ecom->val + (i * ECOMMUNITY_SIZE); | |
1143 | ||
1144 | /* High-order octet of type. */ | |
1145 | encode = *p++; | |
1146 | ||
1147 | if (*p++ == ECOMMUNITY_ROUTE_TARGET) { | |
1148 | if (encode == ECOMMUNITY_ENCODE_AS4) | |
1149 | { | |
1150 | as = (*p++ << 24); | |
1151 | as |= (*p++ << 16); | |
1152 | as |= (*p++ << 8); | |
1153 | as |= (*p++); | |
1154 | } | |
1155 | else if (encode == ECOMMUNITY_ENCODE_AS) | |
1156 | { | |
1157 | as = (*p++ << 8); | |
1158 | as |= (*p++); | |
1159 | p += 2; /* skip next two, tag/vid always in lowest bytes */ | |
1160 | } | |
1161 | if (as == bgp->as) | |
1162 | { | |
1163 | *tag_id = *p++ << 8; | |
1164 | *tag_id |= (*p++); | |
1165 | return 0; | |
1166 | } | |
1167 | } | |
1168 | } | |
1169 | } | |
1170 | return ENOENT; | |
1171 | } | |
1172 | ||
65efcfce LB |
1173 | static int |
1174 | rfapiVpnBiNhEqualsPt (struct bgp_info *bi, struct rfapi_ip_addr *hpt) | |
1175 | { | |
1176 | uint8_t family; | |
1177 | ||
1178 | if (!hpt || !bi) | |
1179 | return 0; | |
1180 | ||
1181 | family = BGP_MP_NEXTHOP_FAMILY (bi->attr->extra->mp_nexthop_len); | |
1182 | ||
1183 | if (hpt->addr_family != family) | |
1184 | return 0; | |
1185 | ||
1186 | switch (family) | |
1187 | { | |
1188 | case AF_INET: | |
1189 | if (bi->attr->extra->mp_nexthop_global_in.s_addr != hpt->addr.v4.s_addr) | |
1190 | return 0; | |
1191 | break; | |
1192 | ||
1193 | case AF_INET6: | |
1194 | if (IPV6_ADDR_CMP (&bi->attr->extra->mp_nexthop_global, &hpt->addr.v6)) | |
1195 | return 0; | |
1196 | break; | |
1197 | ||
1198 | default: | |
1199 | return 0; | |
1200 | break; | |
1201 | } | |
1202 | ||
1203 | return 1; | |
1204 | } | |
1205 | ||
1206 | ||
1207 | /* | |
1208 | * Compare 2 VPN BIs. Return true if they have the same VN and UN addresses | |
1209 | */ | |
1210 | static int | |
1211 | rfapiVpnBiSamePtUn (struct bgp_info *bi1, struct bgp_info *bi2) | |
1212 | { | |
1213 | struct prefix pfx_un1; | |
1214 | struct prefix pfx_un2; | |
1215 | ||
1216 | if (!bi1 || !bi2) | |
1217 | return 0; | |
1218 | ||
1219 | if (!bi1->attr || !bi2->attr) | |
1220 | return 0; | |
1221 | ||
1222 | if (!bi1->attr->extra || !bi2->attr->extra) | |
1223 | return 0; | |
1224 | ||
1225 | /* | |
1226 | * VN address comparisons | |
1227 | */ | |
1228 | ||
1229 | if (BGP_MP_NEXTHOP_FAMILY (bi1->attr->extra->mp_nexthop_len) != | |
1230 | BGP_MP_NEXTHOP_FAMILY (bi2->attr->extra->mp_nexthop_len)) | |
1231 | { | |
1232 | return 0; | |
1233 | } | |
1234 | ||
1235 | switch (BGP_MP_NEXTHOP_FAMILY (bi1->attr->extra->mp_nexthop_len)) | |
1236 | { | |
1237 | ||
1238 | case AF_INET: | |
1239 | if (bi1->attr->extra->mp_nexthop_global_in.s_addr != | |
1240 | bi2->attr->extra->mp_nexthop_global_in.s_addr) | |
1241 | return 0; | |
1242 | break; | |
1243 | ||
1244 | case AF_INET6: | |
1245 | if (IPV6_ADDR_CMP (&bi1->attr->extra->mp_nexthop_global, | |
1246 | &bi2->attr->extra->mp_nexthop_global)) | |
1247 | return 0; | |
1248 | break; | |
1249 | ||
1250 | default: | |
1251 | return 0; | |
1252 | break; | |
1253 | } | |
1254 | ||
1255 | /* | |
1256 | * UN address comparisons | |
1257 | */ | |
1258 | if (rfapiGetVncTunnelUnAddr (bi1->attr, &pfx_un1)) | |
1259 | { | |
1260 | if (bi1->extra) | |
1261 | { | |
1262 | pfx_un1.family = bi1->extra->vnc.import.un_family; | |
1263 | switch (bi1->extra->vnc.import.un_family) | |
1264 | { | |
1265 | case AF_INET: | |
1266 | pfx_un1.u.prefix4 = bi1->extra->vnc.import.un.addr4; | |
1267 | break; | |
1268 | case AF_INET6: | |
1269 | pfx_un1.u.prefix6 = bi1->extra->vnc.import.un.addr6; | |
1270 | break; | |
1271 | default: | |
1272 | pfx_un1.family = 0; | |
1273 | break; | |
1274 | } | |
1275 | } | |
1276 | } | |
1277 | ||
1278 | if (rfapiGetVncTunnelUnAddr (bi2->attr, &pfx_un2)) | |
1279 | { | |
1280 | if (bi2->extra) | |
1281 | { | |
1282 | pfx_un2.family = bi2->extra->vnc.import.un_family; | |
1283 | switch (bi2->extra->vnc.import.un_family) | |
1284 | { | |
1285 | case AF_INET: | |
1286 | pfx_un2.u.prefix4 = bi2->extra->vnc.import.un.addr4; | |
1287 | break; | |
1288 | case AF_INET6: | |
1289 | pfx_un2.u.prefix6 = bi2->extra->vnc.import.un.addr6; | |
1290 | break; | |
1291 | default: | |
1292 | pfx_un2.family = 0; | |
1293 | break; | |
1294 | } | |
1295 | } | |
1296 | } | |
1297 | ||
1298 | if (!pfx_un1.family || !pfx_un2.family) | |
1299 | return 0; | |
1300 | ||
1301 | if (pfx_un1.family != pfx_un2.family) | |
1302 | return 0; | |
1303 | ||
1304 | switch (pfx_un1.family) | |
1305 | { | |
1306 | case AF_INET: | |
1307 | if (!IPV4_ADDR_SAME | |
1308 | (&pfx_un1.u.prefix4.s_addr, &pfx_un2.u.prefix4.s_addr)) | |
1309 | return 0; | |
1310 | break; | |
1311 | case AF_INET6: | |
1312 | if (!IPV6_ADDR_SAME (&pfx_un1.u.prefix6, &pfx_un2.u.prefix6)) | |
1313 | return 0; | |
1314 | break; | |
1315 | } | |
1316 | ||
1317 | ||
1318 | ||
1319 | return 1; | |
1320 | } | |
1321 | ||
1322 | uint8_t | |
1323 | rfapiRfpCost (struct attr * attr) | |
1324 | { | |
1325 | if (attr->flag & ATTR_FLAG_BIT (BGP_ATTR_LOCAL_PREF)) | |
1326 | { | |
1327 | if (attr->local_pref > 255) | |
1328 | { | |
1329 | return 0; | |
1330 | } | |
1331 | return 255 - attr->local_pref; | |
1332 | } | |
1333 | ||
1334 | return 255; | |
1335 | } | |
1336 | ||
1337 | /*------------------------------------------ | |
1338 | * rfapi_extract_l2o | |
1339 | * | |
1340 | * Find Layer 2 options in an option chain | |
1341 | * | |
1342 | * input: | |
1343 | * pHop option chain | |
1344 | * | |
1345 | * output: | |
1346 | * l2o layer 2 options extracted | |
1347 | * | |
1348 | * return value: | |
1349 | * 0 OK | |
1350 | * 1 no options found | |
1351 | * | |
1352 | --------------------------------------------*/ | |
1353 | int | |
1354 | rfapi_extract_l2o (struct bgp_tea_options *pHop, /* chain of options */ | |
1355 | struct rfapi_l2address_option *l2o) /* return extracted value */ | |
1356 | { | |
1357 | struct bgp_tea_options *p; | |
1358 | ||
1359 | for (p = pHop; p; p = p->next) | |
1360 | { | |
1361 | if ((p->type == RFAPI_VN_OPTION_TYPE_L2ADDR) && (p->length >= 8)) | |
1362 | { | |
1363 | ||
1364 | char *v = p->value; | |
1365 | ||
1366 | memcpy (&l2o->macaddr, v, 6); | |
1367 | ||
1368 | l2o->label = | |
1369 | ((v[6] << 12) & 0xff000) + | |
1370 | ((v[7] << 4) & 0xff0) + ((v[8] >> 4) & 0xf); | |
1371 | ||
1372 | l2o->local_nve_id = (uint8_t) v[10]; | |
1373 | ||
1374 | l2o->logical_net_id = (v[11] << 16) + (v[12] << 8) + (v[13] << 0); | |
1375 | ||
1376 | return 0; | |
1377 | } | |
1378 | } | |
1379 | return 1; | |
1380 | } | |
1381 | ||
1382 | static struct rfapi_next_hop_entry * | |
1383 | rfapiRouteInfo2NextHopEntry ( | |
1384 | struct rfapi_ip_prefix *rprefix, | |
1385 | struct bgp_info *bi, /* route to encode */ | |
1386 | uint32_t lifetime, /* use this in nhe */ | |
1387 | struct route_node *rn) /* req for L2 eth addr */ | |
1388 | { | |
1389 | struct rfapi_next_hop_entry *new; | |
1390 | int have_vnc_tunnel_un = 0; | |
1391 | ||
1392 | #if DEBUG_ENCAP_MONITOR | |
a3b55c25 | 1393 | vnc_zlog_debug_verbose ("%s: entry, bi %p, rn %p", __func__, bi, rn); |
65efcfce LB |
1394 | #endif |
1395 | ||
1396 | new = XCALLOC (MTYPE_RFAPI_NEXTHOP, sizeof (struct rfapi_next_hop_entry)); | |
1397 | assert (new); | |
1398 | ||
1399 | new->prefix = *rprefix; | |
1400 | ||
1401 | if (bi->extra && | |
1402 | decode_rd_type(bi->extra->vnc.import.rd.val) == RD_TYPE_VNC_ETH) | |
1403 | { | |
1404 | /* ethernet */ | |
1405 | ||
1406 | struct rfapi_vn_option *vo; | |
1407 | ||
1408 | vo = XCALLOC (MTYPE_RFAPI_VN_OPTION, sizeof (struct rfapi_vn_option)); | |
1409 | assert (vo); | |
1410 | ||
1411 | vo->type = RFAPI_VN_OPTION_TYPE_L2ADDR; | |
1412 | ||
1413 | memcpy (&vo->v.l2addr.macaddr, &rn->p.u.prefix_eth.octet, | |
1414 | ETHER_ADDR_LEN); | |
1415 | /* only low 3 bytes of this are significant */ | |
1416 | if (bi->attr && bi->attr->extra) | |
1417 | { | |
1418 | (void) rfapiEcommunityGetLNI (bi->attr->extra->ecommunity, | |
1419 | &vo->v.l2addr.logical_net_id); | |
79799987 LB |
1420 | (void) rfapiEcommunityGetEthernetTag (bi->attr->extra->ecommunity, |
1421 | &vo->v.l2addr.tag_id); | |
65efcfce LB |
1422 | } |
1423 | ||
1424 | /* local_nve_id comes from lower byte of RD type */ | |
1425 | vo->v.l2addr.local_nve_id = bi->extra->vnc.import.rd.val[1]; | |
1426 | ||
1427 | /* label comes from MP_REACH_NLRI label */ | |
1428 | vo->v.l2addr.label = decode_label (bi->extra->tag); | |
1429 | ||
1430 | new->vn_options = vo; | |
1431 | ||
1432 | /* | |
1433 | * If there is an auxiliary prefix (i.e., host IP address), | |
1434 | * use it as the nexthop prefix instead of the query prefix | |
1435 | */ | |
1436 | if (bi->extra->vnc.import.aux_prefix.family) | |
1437 | { | |
1438 | rfapiQprefix2Rprefix (&bi->extra->vnc.import.aux_prefix, | |
1439 | &new->prefix); | |
1440 | } | |
1441 | } | |
1442 | ||
1443 | if (bi->attr) | |
1444 | { | |
1445 | bgp_encap_types tun_type; | |
1446 | new->prefix.cost = rfapiRfpCost (bi->attr); | |
1447 | ||
1448 | if (bi->attr->extra) | |
1449 | { | |
1450 | ||
1451 | struct bgp_attr_encap_subtlv *pEncap; | |
1452 | ||
1453 | switch (BGP_MP_NEXTHOP_FAMILY (bi->attr->extra->mp_nexthop_len)) | |
1454 | { | |
1455 | case AF_INET: | |
1456 | new->vn_address.addr_family = AF_INET; | |
1457 | new->vn_address.addr.v4 = bi->attr->extra->mp_nexthop_global_in; | |
1458 | break; | |
1459 | ||
1460 | case AF_INET6: | |
1461 | new->vn_address.addr_family = AF_INET6; | |
1462 | new->vn_address.addr.v6 = bi->attr->extra->mp_nexthop_global; | |
1463 | break; | |
1464 | ||
1465 | default: | |
1466 | zlog_warn ("%s: invalid vpn nexthop length: %d", | |
1467 | __func__, bi->attr->extra->mp_nexthop_len); | |
1468 | rfapi_free_next_hop_list (new); | |
1469 | return NULL; | |
1470 | } | |
1471 | ||
1472 | for (pEncap = bi->attr->extra->vnc_subtlvs; pEncap; | |
1473 | pEncap = pEncap->next) | |
1474 | { | |
1475 | switch (pEncap->type) | |
1476 | { | |
1477 | case BGP_VNC_SUBTLV_TYPE_LIFETIME: | |
1478 | /* use configured lifetime, not attr lifetime */ | |
1479 | break; | |
1480 | ||
1481 | default: | |
1482 | zlog_warn ("%s: unknown VNC option type %d", | |
1483 | __func__, pEncap->type); | |
1484 | ||
1485 | ||
1486 | break; | |
1487 | } | |
1488 | } | |
1489 | ||
1490 | rfapiGetTunnelType (bi->attr, &tun_type); | |
1491 | if (tun_type == BGP_ENCAP_TYPE_MPLS) | |
1492 | { | |
1493 | struct prefix p; | |
1494 | /* MPLS carries UN address in next hop */ | |
1495 | rfapiNexthop2Prefix (bi->attr, &p); | |
1496 | if (p.family != 0) | |
1497 | { | |
1498 | rfapiQprefix2Raddr(&p, &new->un_address); | |
1499 | have_vnc_tunnel_un = 1; | |
1500 | } | |
1501 | } | |
1502 | ||
1503 | for (pEncap = bi->attr->extra->encap_subtlvs; pEncap; | |
1504 | pEncap = pEncap->next) | |
1505 | { | |
1506 | switch (pEncap->type) | |
1507 | { | |
1508 | case BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT: | |
1509 | /* | |
1510 | * Overrides ENCAP UN address, if any | |
1511 | */ | |
1512 | switch (pEncap->length) | |
1513 | { | |
1514 | ||
1515 | case 8: | |
1516 | new->un_address.addr_family = AF_INET; | |
1517 | memcpy (&new->un_address.addr.v4, pEncap->value, 4); | |
1518 | have_vnc_tunnel_un = 1; | |
1519 | break; | |
1520 | ||
1521 | case 20: | |
1522 | new->un_address.addr_family = AF_INET6; | |
1523 | memcpy (&new->un_address.addr.v6, pEncap->value, 16); | |
1524 | have_vnc_tunnel_un = 1; | |
1525 | break; | |
1526 | ||
1527 | default: | |
1528 | zlog_warn | |
1529 | ("%s: invalid tunnel subtlv UN addr length (%d) for bi %p", | |
1530 | __func__, pEncap->length, bi); | |
1531 | } | |
1532 | break; | |
1533 | ||
1534 | default: | |
1535 | zlog_warn ("%s: unknown Encap Attribute option type %d", | |
1536 | __func__, pEncap->type); | |
1537 | ||
1538 | ||
1539 | break; | |
1540 | } | |
1541 | } | |
1542 | ||
1543 | new->un_options = rfapi_encap_tlv_to_un_option (bi->attr); | |
1544 | ||
1545 | #if DEBUG_ENCAP_MONITOR | |
a3b55c25 | 1546 | vnc_zlog_debug_verbose ("%s: line %d: have_vnc_tunnel_un=%d", |
65efcfce LB |
1547 | __func__, __LINE__, have_vnc_tunnel_un); |
1548 | #endif | |
1549 | ||
1550 | if (!have_vnc_tunnel_un && bi && bi->extra) | |
1551 | { | |
1552 | /* | |
1553 | * use cached UN address from ENCAP route | |
1554 | */ | |
1555 | new->un_address.addr_family = bi->extra->vnc.import.un_family; | |
1556 | switch (new->un_address.addr_family) | |
1557 | { | |
1558 | case AF_INET: | |
1559 | new->un_address.addr.v4 = bi->extra->vnc.import.un.addr4; | |
1560 | break; | |
1561 | case AF_INET6: | |
1562 | new->un_address.addr.v6 = bi->extra->vnc.import.un.addr6; | |
1563 | break; | |
1564 | default: | |
1565 | zlog_warn ("%s: invalid UN addr family (%d) for bi %p", | |
1566 | __func__, new->un_address.addr_family, bi); | |
1567 | rfapi_free_next_hop_list (new); | |
1568 | return NULL; | |
1569 | break; | |
1570 | } | |
1571 | } | |
1572 | } | |
1573 | } | |
1574 | new->lifetime = lifetime; | |
1575 | return new; | |
1576 | } | |
1577 | ||
1578 | int | |
1579 | rfapiHasNonRemovedRoutes (struct route_node *rn) | |
1580 | { | |
1581 | struct bgp_info *bi; | |
1582 | ||
1583 | for (bi = rn->info; bi; bi = bi->next) | |
1584 | { | |
1585 | struct prefix pfx; | |
1586 | ||
1587 | if (!CHECK_FLAG (bi->flags, BGP_INFO_REMOVED) && | |
1588 | (bi->extra && !rfapiGetUnAddrOfVpnBi (bi, &pfx))) | |
1589 | { | |
1590 | ||
1591 | return 1; | |
1592 | } | |
1593 | } | |
1594 | return 0; | |
1595 | } | |
1596 | ||
1597 | #if DEBUG_IT_NODES | |
1598 | /* | |
1599 | * DEBUG FUNCTION | |
1600 | */ | |
1601 | void | |
1602 | rfapiDumpNode (struct route_node *rn) | |
1603 | { | |
1604 | struct bgp_info *bi; | |
1605 | ||
a3b55c25 | 1606 | vnc_zlog_debug_verbose ("%s: rn=%p", __func__, rn); |
65efcfce LB |
1607 | for (bi = rn->info; bi; bi = bi->next) |
1608 | { | |
1609 | struct prefix pfx; | |
1610 | int ctrc = rfapiGetUnAddrOfVpnBi (bi, &pfx); | |
1611 | int nr; | |
1612 | ||
1613 | if (!CHECK_FLAG (bi->flags, BGP_INFO_REMOVED) && (bi->extra && !ctrc)) | |
1614 | { | |
1615 | ||
1616 | nr = 1; | |
1617 | } | |
1618 | else | |
1619 | { | |
1620 | nr = 0; | |
1621 | } | |
1622 | ||
a3b55c25 | 1623 | vnc_zlog_debug_verbose (" bi=%p, nr=%d, flags=0x%x, extra=%p, ctrc=%d", |
65efcfce LB |
1624 | bi, nr, bi->flags, bi->extra, ctrc); |
1625 | } | |
1626 | } | |
1627 | #endif | |
1628 | ||
1629 | static int | |
1630 | rfapiNhlAddNodeRoutes ( | |
1631 | struct route_node *rn, /* in */ | |
1632 | struct rfapi_ip_prefix *rprefix, /* in */ | |
1633 | uint32_t lifetime, /* in */ | |
1634 | int removed, /* in */ | |
1635 | struct rfapi_next_hop_entry **head, /* in/out */ | |
1636 | struct rfapi_next_hop_entry **tail, /* in/out */ | |
1637 | struct rfapi_ip_addr *exclude_vnaddr, /* omit routes to same NVE */ | |
1638 | struct route_node *rfd_rib_node,/* preload this NVE rib node */ | |
1639 | struct prefix *pfx_target_original) /* query target */ | |
1640 | { | |
1641 | struct bgp_info *bi; | |
1642 | struct rfapi_next_hop_entry *new; | |
1643 | struct prefix pfx_un; | |
1644 | struct skiplist *seen_nexthops; | |
1645 | int count = 0; | |
1646 | int is_l2 = (rn->p.family == AF_ETHERNET); | |
1647 | ||
1648 | if (rfapiRibFTDFilterRecentPrefix( | |
1649 | (struct rfapi_descriptor *)(rfd_rib_node->table->info), rn, | |
1650 | pfx_target_original)) | |
1651 | { | |
1652 | return 0; | |
1653 | } | |
1654 | ||
1655 | seen_nexthops = | |
1656 | skiplist_new (0, vnc_prefix_cmp, (void (*)(void *)) prefix_free); | |
1657 | ||
1658 | for (bi = rn->info; bi; bi = bi->next) | |
1659 | { | |
1660 | ||
1661 | struct prefix pfx_vn; | |
1662 | struct prefix *newpfx; | |
1663 | ||
1664 | if (removed && !CHECK_FLAG (bi->flags, BGP_INFO_REMOVED)) | |
1665 | { | |
1666 | #if DEBUG_RETURNED_NHL | |
a3b55c25 | 1667 | vnc_zlog_debug_verbose ("%s: want holddown, this route not holddown, skip", |
65efcfce LB |
1668 | __func__); |
1669 | #endif | |
1670 | continue; | |
1671 | } | |
1672 | if (!removed && CHECK_FLAG (bi->flags, BGP_INFO_REMOVED)) | |
1673 | { | |
1674 | continue; | |
1675 | } | |
1676 | ||
1677 | if (!bi->extra) | |
1678 | { | |
1679 | continue; | |
1680 | } | |
1681 | ||
1682 | /* | |
1683 | * Check for excluded VN address | |
1684 | */ | |
1685 | if (rfapiVpnBiNhEqualsPt (bi, exclude_vnaddr)) | |
1686 | continue; | |
1687 | ||
1688 | /* | |
1689 | * Check for VN address (nexthop) copied already | |
1690 | */ | |
1691 | if (is_l2) | |
1692 | { | |
1693 | /* L2 routes: semantic nexthop in aux_prefix; VN addr ain't it */ | |
1694 | pfx_vn = bi->extra->vnc.import.aux_prefix; | |
1695 | } | |
1696 | else | |
1697 | { | |
1698 | rfapiNexthop2Prefix (bi->attr, &pfx_vn); | |
1699 | } | |
1700 | if (!skiplist_search (seen_nexthops, &pfx_vn, NULL)) | |
1701 | { | |
1702 | #if DEBUG_RETURNED_NHL | |
1703 | char buf[BUFSIZ]; | |
1704 | ||
1705 | prefix2str (&pfx_vn, buf, BUFSIZ); | |
1706 | buf[BUFSIZ - 1] = 0; /* guarantee NUL-terminated */ | |
a3b55c25 | 1707 | vnc_zlog_debug_verbose ("%s: already put VN/nexthop %s, skip", __func__, buf); |
65efcfce LB |
1708 | #endif |
1709 | continue; | |
1710 | } | |
1711 | ||
1712 | if (rfapiGetUnAddrOfVpnBi (bi, &pfx_un)) | |
1713 | { | |
1714 | #if DEBUG_ENCAP_MONITOR | |
a3b55c25 | 1715 | vnc_zlog_debug_verbose ("%s: failed to get UN address of this VPN bi", |
65efcfce LB |
1716 | __func__); |
1717 | #endif | |
1718 | continue; | |
1719 | } | |
1720 | ||
1721 | newpfx = prefix_new (); | |
1722 | *newpfx = pfx_vn; | |
1723 | skiplist_insert (seen_nexthops, newpfx, newpfx); | |
1724 | ||
1725 | new = rfapiRouteInfo2NextHopEntry(rprefix, bi, lifetime, rn); | |
1726 | if (new) | |
1727 | { | |
1728 | if (rfapiRibPreloadBi(rfd_rib_node, &pfx_vn, &pfx_un, lifetime, bi)) | |
1729 | { | |
1730 | /* duplicate filtered by RIB */ | |
1731 | rfapi_free_next_hop_list (new); | |
1732 | new = NULL; | |
1733 | } | |
1734 | } | |
1735 | ||
1736 | if (new) | |
1737 | { | |
1738 | if (*tail) | |
1739 | { | |
1740 | (*tail)->next = new; | |
1741 | } | |
1742 | else | |
1743 | { | |
1744 | *head = new; | |
1745 | } | |
1746 | *tail = new; | |
1747 | ++count; | |
1748 | } | |
1749 | } | |
1750 | ||
1751 | skiplist_free (seen_nexthops); | |
1752 | ||
1753 | return count; | |
1754 | } | |
1755 | ||
1756 | ||
1757 | /* | |
1758 | * Breadth-first | |
1759 | * | |
1760 | * omit_node is meant for the situation where we are adding a subtree | |
1761 | * of a parent of some original requested node. The response already | |
1762 | * contains the original requested node, and we don't want to duplicate | |
1763 | * its routes in the list, so we skip it if the right or left node | |
1764 | * matches (of course, we still travel down its child subtrees). | |
1765 | */ | |
1766 | static int | |
1767 | rfapiNhlAddSubtree ( | |
1768 | struct route_node *rn, /* in */ | |
1769 | uint32_t lifetime, /* in */ | |
1770 | struct rfapi_next_hop_entry **head, /* in/out */ | |
1771 | struct rfapi_next_hop_entry **tail, /* in/out */ | |
1772 | struct route_node *omit_node, /* in */ | |
1773 | struct rfapi_ip_addr *exclude_vnaddr,/* omit routes to same NVE */ | |
1774 | struct route_table *rfd_rib_table,/* preload here */ | |
1775 | struct prefix *pfx_target_original) /* query target */ | |
1776 | { | |
1777 | struct rfapi_ip_prefix rprefix; | |
1778 | int rcount = 0; | |
1779 | ||
1780 | if (rn->l_left && rn->l_left != omit_node) | |
1781 | { | |
1782 | if (rn->l_left->info) | |
1783 | { | |
1784 | int count = 0; | |
1785 | struct route_node *rib_rn = NULL; | |
1786 | ||
1787 | rfapiQprefix2Rprefix (&rn->l_left->p, &rprefix); | |
1788 | if (rfd_rib_table) | |
1789 | { | |
1790 | rib_rn = route_node_get(rfd_rib_table, &rn->l_left->p); | |
1791 | } | |
1792 | ||
1793 | count = rfapiNhlAddNodeRoutes (rn->l_left, &rprefix, lifetime, 0, | |
1794 | head, tail, exclude_vnaddr, rib_rn, pfx_target_original); | |
1795 | if (!count) | |
1796 | { | |
1797 | count = rfapiNhlAddNodeRoutes (rn->l_left, &rprefix, lifetime, 1, | |
1798 | head, tail, exclude_vnaddr, rib_rn, pfx_target_original); | |
1799 | } | |
1800 | rcount += count; | |
1801 | if (rib_rn) | |
1802 | route_unlock_node(rib_rn); | |
1803 | } | |
1804 | } | |
1805 | ||
1806 | if (rn->l_right && rn->l_right != omit_node) | |
1807 | { | |
1808 | if (rn->l_right->info) | |
1809 | { | |
1810 | int count = 0; | |
1811 | struct route_node *rib_rn = NULL; | |
1812 | ||
1813 | rfapiQprefix2Rprefix (&rn->l_right->p, &rprefix); | |
1814 | if (rfd_rib_table) | |
1815 | { | |
1816 | rib_rn = route_node_get(rfd_rib_table, &rn->l_right->p); | |
1817 | } | |
1818 | count = rfapiNhlAddNodeRoutes (rn->l_right, &rprefix, lifetime, 0, | |
1819 | head, tail, exclude_vnaddr, rib_rn, pfx_target_original); | |
1820 | if (!count) | |
1821 | { | |
1822 | count = rfapiNhlAddNodeRoutes (rn->l_right, &rprefix, lifetime, 1, | |
1823 | head, tail, exclude_vnaddr, rib_rn, pfx_target_original); | |
1824 | } | |
1825 | rcount += count; | |
1826 | if (rib_rn) | |
1827 | route_unlock_node(rib_rn); | |
1828 | } | |
1829 | } | |
1830 | ||
1831 | if (rn->l_left) | |
1832 | { | |
1833 | rcount += rfapiNhlAddSubtree (rn->l_left, lifetime, head, tail, omit_node, | |
1834 | exclude_vnaddr, rfd_rib_table, pfx_target_original); | |
1835 | } | |
1836 | if (rn->l_right) | |
1837 | { | |
1838 | rcount += rfapiNhlAddSubtree (rn->l_right, lifetime, head, tail, | |
1839 | omit_node, exclude_vnaddr, rfd_rib_table, pfx_target_original); | |
1840 | } | |
1841 | ||
1842 | return rcount; | |
1843 | } | |
1844 | ||
1845 | /* | |
1846 | * Implementation of ROUTE_LIST(node) from RFAPI-Import-Event-Handling.txt | |
1847 | * | |
1848 | * Construct an rfapi nexthop list based on the routes attached to | |
1849 | * the specified node. | |
1850 | * | |
1851 | * If there are any routes that do NOT have BGP_INFO_REMOVED set, | |
1852 | * return those only. If there are ONLY routes with BGP_INFO_REMOVED, | |
1853 | * then return those, and also include all the non-removed routes from the | |
1854 | * next less-specific node (i.e., this node's parent) at the end. | |
1855 | */ | |
1856 | struct rfapi_next_hop_entry * | |
1857 | rfapiRouteNode2NextHopList ( | |
1858 | struct route_node *rn, | |
1859 | uint32_t lifetime, /* put into nexthop entries */ | |
1860 | struct rfapi_ip_addr *exclude_vnaddr,/* omit routes to same NVE */ | |
1861 | struct route_table *rfd_rib_table,/* preload here */ | |
1862 | struct prefix *pfx_target_original) /* query target */ | |
1863 | { | |
1864 | struct rfapi_ip_prefix rprefix; | |
1865 | struct rfapi_next_hop_entry *answer = NULL; | |
1866 | struct rfapi_next_hop_entry *last = NULL; | |
1867 | struct route_node *parent; | |
1868 | int count = 0; | |
1869 | struct route_node *rib_rn; | |
1870 | ||
1871 | #if DEBUG_RETURNED_NHL | |
1872 | { | |
1873 | char buf[BUFSIZ]; | |
1874 | ||
1875 | prefix2str (&rn->p, buf, BUFSIZ); | |
1876 | buf[BUFSIZ - 1] = 0; | |
a3b55c25 | 1877 | vnc_zlog_debug_verbose ("%s: called with node pfx=%s", __func__, buf); |
65efcfce LB |
1878 | } |
1879 | rfapiDebugBacktrace (); | |
1880 | #endif | |
1881 | ||
1882 | rfapiQprefix2Rprefix (&rn->p, &rprefix); | |
1883 | ||
1884 | rib_rn = rfd_rib_table? route_node_get(rfd_rib_table, &rn->p): NULL; | |
1885 | ||
1886 | /* | |
1887 | * Add non-withdrawn routes at this node | |
1888 | */ | |
1889 | count = rfapiNhlAddNodeRoutes (rn, &rprefix, lifetime, 0, &answer, &last, | |
1890 | exclude_vnaddr, rib_rn, pfx_target_original); | |
1891 | ||
1892 | /* | |
1893 | * If the list has at least one entry, it's finished | |
1894 | */ | |
1895 | if (count) | |
1896 | { | |
1897 | count += rfapiNhlAddSubtree (rn, lifetime, &answer, &last, NULL, | |
1898 | exclude_vnaddr, rfd_rib_table, pfx_target_original); | |
a3b55c25 | 1899 | vnc_zlog_debug_verbose ("%s: %d nexthops, answer=%p", __func__, count, answer); |
65efcfce LB |
1900 | #if DEBUG_RETURNED_NHL |
1901 | rfapiPrintNhl (NULL, answer); | |
1902 | #endif | |
1903 | if (rib_rn) | |
1904 | route_unlock_node(rib_rn); | |
1905 | return answer; | |
1906 | } | |
1907 | ||
1908 | /* | |
1909 | * Add withdrawn routes at this node | |
1910 | */ | |
1911 | count = rfapiNhlAddNodeRoutes (rn, &rprefix, lifetime, 1, &answer, &last, | |
1912 | exclude_vnaddr, rib_rn, pfx_target_original); | |
1913 | if (rib_rn) | |
1914 | route_unlock_node(rib_rn); | |
1915 | ||
1916 | // rfapiPrintNhl(NULL, answer); | |
1917 | ||
1918 | /* | |
1919 | * walk up the tree until we find a node with non-deleted | |
1920 | * routes, then add them | |
1921 | */ | |
1922 | for (parent = rn->parent; parent; parent = parent->parent) | |
1923 | { | |
1924 | if (rfapiHasNonRemovedRoutes (parent)) | |
1925 | { | |
1926 | break; | |
1927 | } | |
1928 | } | |
1929 | ||
1930 | /* | |
1931 | * Add non-withdrawn routes from less-specific prefix | |
1932 | */ | |
1933 | if (parent) | |
1934 | { | |
1935 | rib_rn = rfd_rib_table? route_node_get(rfd_rib_table, &parent->p): NULL; | |
1936 | rfapiQprefix2Rprefix (&parent->p, &rprefix); | |
1937 | count += rfapiNhlAddNodeRoutes (parent, &rprefix, lifetime, 0, | |
1938 | &answer, &last, exclude_vnaddr, rib_rn, pfx_target_original); | |
1939 | count += rfapiNhlAddSubtree (parent, lifetime, &answer, &last, rn, | |
1940 | exclude_vnaddr, rfd_rib_table, pfx_target_original); | |
1941 | if (rib_rn) | |
1942 | route_unlock_node(rib_rn); | |
1943 | } | |
1944 | else | |
1945 | { | |
1946 | /* | |
1947 | * There is no parent with non-removed routes. Still need to | |
1948 | * add subtree of original node if it contributed routes to the | |
1949 | * answer. | |
1950 | */ | |
1951 | if (count) | |
1952 | count += rfapiNhlAddSubtree (rn, lifetime, &answer, &last, rn, | |
1953 | exclude_vnaddr, rfd_rib_table, pfx_target_original); | |
1954 | } | |
1955 | ||
a3b55c25 | 1956 | vnc_zlog_debug_verbose ("%s: %d nexthops, answer=%p", __func__, count, answer); |
65efcfce LB |
1957 | #if DEBUG_RETURNED_NHL |
1958 | rfapiPrintNhl (NULL, answer); | |
1959 | #endif | |
1960 | return answer; | |
1961 | } | |
1962 | ||
1963 | /* | |
1964 | * Construct nexthop list of all routes in table | |
1965 | */ | |
1966 | struct rfapi_next_hop_entry * | |
1967 | rfapiRouteTable2NextHopList ( | |
1968 | struct route_table *rt, | |
1969 | uint32_t lifetime, /* put into nexthop entries */ | |
1970 | struct rfapi_ip_addr *exclude_vnaddr,/* omit routes to same NVE */ | |
1971 | struct route_table *rfd_rib_table, /* preload this NVE rib table */ | |
1972 | struct prefix *pfx_target_original) /* query target */ | |
1973 | { | |
1974 | struct route_node *rn; | |
1975 | struct rfapi_next_hop_entry *biglist = NULL; | |
1976 | struct rfapi_next_hop_entry *nhl; | |
1977 | struct rfapi_next_hop_entry *tail = NULL; | |
1978 | int count = 0; | |
1979 | ||
1980 | for (rn = route_top (rt); rn; rn = route_next (rn)) | |
1981 | { | |
1982 | ||
1983 | nhl = rfapiRouteNode2NextHopList (rn, lifetime, exclude_vnaddr, | |
1984 | rfd_rib_table, pfx_target_original); | |
1985 | if (!tail) | |
1986 | { | |
1987 | tail = biglist = nhl; | |
1988 | if (tail) | |
1989 | count = 1; | |
1990 | } | |
1991 | else | |
1992 | { | |
1993 | tail->next = nhl; | |
1994 | } | |
1995 | if (tail) | |
1996 | { | |
1997 | while (tail->next) | |
1998 | { | |
1999 | ++count; | |
2000 | tail = tail->next; | |
2001 | } | |
2002 | } | |
2003 | } | |
2004 | ||
a3b55c25 | 2005 | vnc_zlog_debug_verbose ("%s: returning %d routes", __func__, count); |
65efcfce LB |
2006 | return biglist; |
2007 | } | |
2008 | ||
2009 | struct rfapi_next_hop_entry * | |
2010 | rfapiEthRouteNode2NextHopList ( | |
2011 | struct route_node *rn, | |
2012 | struct rfapi_ip_prefix *rprefix, | |
2013 | uint32_t lifetime, /* put into nexthop entries */ | |
2014 | struct rfapi_ip_addr *exclude_vnaddr,/* omit routes to same NVE */ | |
2015 | struct route_table *rfd_rib_table,/* preload NVE rib table */ | |
2016 | struct prefix *pfx_target_original) /* query target */ | |
2017 | { | |
2018 | int count = 0; | |
2019 | struct rfapi_next_hop_entry *answer = NULL; | |
2020 | struct rfapi_next_hop_entry *last = NULL; | |
2021 | struct route_node *rib_rn; | |
2022 | ||
2023 | rib_rn = rfd_rib_table? route_node_get(rfd_rib_table, &rn->p): NULL; | |
2024 | ||
2025 | count = rfapiNhlAddNodeRoutes (rn, rprefix, lifetime, 0, &answer, &last, | |
2026 | NULL, rib_rn, pfx_target_original); | |
2027 | ||
2028 | #if DEBUG_ENCAP_MONITOR | |
a3b55c25 | 2029 | vnc_zlog_debug_verbose ("%s: node %p: %d non-holddown routes", __func__, rn, count); |
65efcfce LB |
2030 | #endif |
2031 | ||
2032 | if (!count) | |
2033 | { | |
2034 | count = rfapiNhlAddNodeRoutes (rn, rprefix, lifetime, 1, &answer, &last, | |
2035 | exclude_vnaddr, rib_rn, pfx_target_original); | |
a3b55c25 | 2036 | vnc_zlog_debug_verbose ("%s: node %p: %d holddown routes", __func__, rn, count); |
65efcfce LB |
2037 | } |
2038 | ||
2039 | if (rib_rn) | |
2040 | route_unlock_node(rib_rn); | |
2041 | ||
2042 | #if DEBUG_RETURNED_NHL | |
2043 | rfapiPrintNhl (NULL, answer); | |
2044 | #endif | |
2045 | ||
2046 | return answer; | |
2047 | } | |
2048 | ||
2049 | ||
2050 | /* | |
2051 | * Construct nexthop list of all routes in table | |
2052 | */ | |
2053 | struct rfapi_next_hop_entry * | |
2054 | rfapiEthRouteTable2NextHopList ( | |
2055 | uint32_t logical_net_id, | |
2056 | struct rfapi_ip_prefix *rprefix, | |
2057 | uint32_t lifetime, /* put into nexthop entries */ | |
2058 | struct rfapi_ip_addr *exclude_vnaddr,/* omit routes to same NVE */ | |
2059 | struct route_table *rfd_rib_table, /* preload NVE rib node */ | |
2060 | struct prefix *pfx_target_original) /* query target */ | |
2061 | { | |
2062 | struct rfapi_import_table *it; | |
2063 | struct bgp *bgp = bgp_get_default (); | |
2064 | struct route_table *rt; | |
2065 | struct route_node *rn; | |
2066 | struct rfapi_next_hop_entry *biglist = NULL; | |
2067 | struct rfapi_next_hop_entry *nhl; | |
2068 | struct rfapi_next_hop_entry *tail = NULL; | |
2069 | int count = 0; | |
2070 | ||
2071 | ||
2072 | it = rfapiMacImportTableGet (bgp, logical_net_id); | |
2073 | rt = it->imported_vpn[AFI_ETHER]; | |
2074 | ||
2075 | for (rn = route_top (rt); rn; rn = route_next (rn)) | |
2076 | { | |
2077 | ||
2078 | nhl = rfapiEthRouteNode2NextHopList(rn, rprefix, lifetime, | |
2079 | exclude_vnaddr, rfd_rib_table, pfx_target_original); | |
2080 | if (!tail) | |
2081 | { | |
2082 | tail = biglist = nhl; | |
2083 | if (tail) | |
2084 | count = 1; | |
2085 | } | |
2086 | else | |
2087 | { | |
2088 | tail->next = nhl; | |
2089 | } | |
2090 | if (tail) | |
2091 | { | |
2092 | while (tail->next) | |
2093 | { | |
2094 | ++count; | |
2095 | tail = tail->next; | |
2096 | } | |
2097 | } | |
2098 | } | |
2099 | ||
a3b55c25 | 2100 | vnc_zlog_debug_verbose ("%s: returning %d routes", __func__, count); |
65efcfce LB |
2101 | return biglist; |
2102 | } | |
2103 | ||
2104 | /* | |
2105 | * Insert a new bi to the imported route table node, | |
2106 | * keeping the list of BIs sorted best route first | |
2107 | */ | |
2108 | static void | |
2109 | rfapiBgpInfoAttachSorted ( | |
2110 | struct route_node *rn, | |
2111 | struct bgp_info *info_new, | |
2112 | afi_t afi, | |
2113 | safi_t safi) | |
2114 | { | |
2115 | struct bgp *bgp; | |
2116 | struct bgp_info *prev; | |
2117 | struct bgp_info *next; | |
2118 | ||
2119 | bgp = bgp_get_default (); /* assume 1 instance for now */ | |
2120 | ||
2121 | if (VNC_DEBUG(IMPORT_BI_ATTACH)) | |
2122 | { | |
a3b55c25 LB |
2123 | vnc_zlog_debug_verbose ("%s: info_new->peer=%p", __func__, info_new->peer); |
2124 | vnc_zlog_debug_verbose ("%s: info_new->peer->su_remote=%p", __func__, | |
65efcfce LB |
2125 | info_new->peer->su_remote); |
2126 | } | |
2127 | ||
2128 | for (prev = NULL, next = rn->info; next; prev = next, next = next->next) | |
2129 | { | |
2130 | if (!bgp || | |
2131 | (!CHECK_FLAG (info_new->flags, BGP_INFO_REMOVED) && | |
2132 | CHECK_FLAG (next->flags, BGP_INFO_REMOVED)) || | |
2133 | bgp_info_cmp_compatible (bgp, info_new, next, afi, safi) == -1) | |
2134 | { /* -1 if 1st is better */ | |
2135 | break; | |
2136 | } | |
2137 | } | |
a3b55c25 | 2138 | vnc_zlog_debug_verbose ("%s: prev=%p, next=%p", __func__, prev, next); |
65efcfce LB |
2139 | if (prev) |
2140 | { | |
2141 | prev->next = info_new; | |
2142 | } | |
2143 | else | |
2144 | { | |
2145 | rn->info = info_new; | |
2146 | } | |
2147 | info_new->prev = prev; | |
2148 | info_new->next = next; | |
2149 | if (next) | |
2150 | next->prev = info_new; | |
bede7744 | 2151 | bgp_attr_intern (info_new->attr); |
65efcfce LB |
2152 | } |
2153 | ||
2154 | static void | |
2155 | rfapiBgpInfoDetach (struct route_node *rn, struct bgp_info *bi) | |
2156 | { | |
2157 | /* | |
2158 | * Remove the route (doubly-linked) | |
2159 | */ | |
bede7744 | 2160 | // bgp_attr_unintern (&bi->attr); |
65efcfce LB |
2161 | if (bi->next) |
2162 | bi->next->prev = bi->prev; | |
2163 | if (bi->prev) | |
2164 | bi->prev->next = bi->next; | |
2165 | else | |
2166 | rn->info = bi->next; | |
2167 | } | |
2168 | ||
2169 | /* | |
2170 | * For L3-indexed import tables | |
2171 | */ | |
2172 | static int | |
2173 | rfapi_bi_peer_rd_cmp (void *b1, void *b2) | |
2174 | { | |
2175 | struct bgp_info *bi1 = b1; | |
2176 | struct bgp_info *bi2 = b2; | |
2177 | ||
2178 | /* | |
2179 | * Compare peers | |
2180 | */ | |
2181 | if (bi1->peer < bi2->peer) | |
2182 | return -1; | |
2183 | if (bi1->peer > bi2->peer) | |
2184 | return 1; | |
2185 | ||
2186 | /* | |
2187 | * compare RDs | |
2188 | */ | |
2189 | return vnc_prefix_cmp ((struct prefix *) &bi1->extra->vnc.import.rd, | |
2190 | (struct prefix *) &bi2->extra->vnc.import.rd); | |
2191 | } | |
2192 | ||
2193 | /* | |
2194 | * For L2-indexed import tables | |
2195 | * The BIs in these tables should ALWAYS have an aux_prefix set because | |
2196 | * they arrive via IPv4 or IPv6 advertisements. | |
2197 | */ | |
2198 | static int | |
2199 | rfapi_bi_peer_rd_aux_cmp (void *b1, void *b2) | |
2200 | { | |
2201 | struct bgp_info *bi1 = b1; | |
2202 | struct bgp_info *bi2 = b2; | |
2203 | int rc; | |
2204 | ||
2205 | /* | |
2206 | * Compare peers | |
2207 | */ | |
2208 | if (bi1->peer < bi2->peer) | |
2209 | return -1; | |
2210 | if (bi1->peer > bi2->peer) | |
2211 | return 1; | |
2212 | ||
2213 | /* | |
2214 | * compare RDs | |
2215 | */ | |
2216 | rc = vnc_prefix_cmp ((struct prefix *) &bi1->extra->vnc.import.rd, | |
2217 | (struct prefix *) &bi2->extra->vnc.import.rd); | |
2218 | if (rc) | |
2219 | { | |
2220 | return rc; | |
2221 | } | |
2222 | ||
2223 | /* | |
2224 | * L2 import tables can have multiple entries with the | |
2225 | * same MAC address, same RD, but different L3 addresses. | |
2226 | * | |
2227 | * Use presence of aux_prefix with AF=ethernet and prefixlen=1 | |
2228 | * as magic value to signify explicit wildcarding of the aux_prefix. | |
2229 | * This magic value will not appear in bona fide bi entries in | |
2230 | * the import table, but is allowed in the "fake" bi used to | |
2231 | * probe the table when searching. (We have to test both b1 and b2 | |
2232 | * because there is no guarantee of the order the test key and | |
2233 | * the real key will be passed) | |
2234 | */ | |
2235 | if ((bi1->extra->vnc.import.aux_prefix.family == AF_ETHERNET && | |
2236 | (bi1->extra->vnc.import.aux_prefix.prefixlen == 1)) || | |
2237 | (bi2->extra->vnc.import.aux_prefix.family == AF_ETHERNET && | |
2238 | (bi2->extra->vnc.import.aux_prefix.prefixlen == 1))) | |
2239 | { | |
2240 | ||
2241 | /* | |
2242 | * wildcard aux address specified | |
2243 | */ | |
2244 | return 0; | |
2245 | } | |
2246 | ||
2247 | return vnc_prefix_cmp (&bi1->extra->vnc.import.aux_prefix, | |
2248 | &bi2->extra->vnc.import.aux_prefix); | |
2249 | } | |
2250 | ||
2251 | ||
2252 | /* | |
2253 | * Index on RD and Peer | |
2254 | */ | |
2255 | static void | |
2256 | rfapiItBiIndexAdd ( | |
2257 | struct route_node *rn, /* Import table VPN node */ | |
2258 | struct bgp_info *bi) /* new BI */ | |
2259 | { | |
2260 | struct skiplist *sl; | |
2261 | ||
2262 | assert (rn); | |
2263 | assert (bi); | |
2264 | assert (bi->extra); | |
2265 | ||
2266 | { | |
2267 | char buf[BUFSIZ]; | |
2268 | prefix_rd2str (&bi->extra->vnc.import.rd, buf, BUFSIZ); | |
a3b55c25 | 2269 | vnc_zlog_debug_verbose ("%s: bi %p, peer %p, rd %s", __func__, bi, bi->peer, buf); |
65efcfce LB |
2270 | } |
2271 | ||
2272 | sl = RFAPI_RDINDEX_W_ALLOC (rn); | |
2273 | if (!sl) | |
2274 | { | |
2275 | if (AF_ETHERNET == rn->p.family) | |
2276 | { | |
2277 | sl = skiplist_new (0, rfapi_bi_peer_rd_aux_cmp, NULL); | |
2278 | } | |
2279 | else | |
2280 | { | |
2281 | sl = skiplist_new (0, rfapi_bi_peer_rd_cmp, NULL); | |
2282 | } | |
2283 | RFAPI_IT_EXTRA_GET (rn)->u.vpn.idx_rd = sl; | |
2284 | route_lock_node (rn); /* for skiplist */ | |
2285 | } | |
2286 | assert (!skiplist_insert (sl, (void *) bi, (void *) bi)); | |
2287 | route_lock_node (rn); /* for skiplist entry */ | |
2288 | ||
2289 | /* NB: BIs in import tables are not refcounted */ | |
2290 | } | |
2291 | ||
2292 | static void | |
2293 | rfapiItBiIndexDump (struct route_node *rn) | |
2294 | { | |
2295 | struct skiplist *sl; | |
2296 | void *cursor = NULL; | |
2297 | struct bgp_info *k; | |
2298 | struct bgp_info *v; | |
2299 | int rc; | |
2300 | ||
2301 | sl = RFAPI_RDINDEX (rn); | |
2302 | if (!sl) | |
2303 | return; | |
2304 | ||
2305 | for (rc = skiplist_next (sl, (void **) &k, (void **) &v, &cursor); | |
2306 | !rc; rc = skiplist_next (sl, (void **) &k, (void **) &v, &cursor)) | |
2307 | { | |
2308 | ||
2309 | char buf[BUFSIZ]; | |
2310 | char buf_aux_pfx[BUFSIZ]; | |
2311 | ||
2312 | prefix_rd2str (&k->extra->vnc.import.rd, buf, BUFSIZ); | |
2313 | buf_aux_pfx[0] = 0; | |
2314 | if (k->extra->vnc.import.aux_prefix.family) | |
2315 | { | |
2316 | prefix2str (&k->extra->vnc.import.aux_prefix, buf_aux_pfx, BUFSIZ); | |
2317 | } | |
2318 | else | |
2319 | { | |
2320 | strncpy (buf_aux_pfx, "(none)", BUFSIZ); | |
2321 | buf_aux_pfx[BUFSIZ - 1] = 0; | |
2322 | } | |
2323 | ||
a3b55c25 | 2324 | vnc_zlog_debug_verbose ("bi %p, peer %p, rd %s, aux_prefix %s", k, k->peer, buf, |
65efcfce LB |
2325 | buf_aux_pfx); |
2326 | } | |
2327 | } | |
2328 | ||
2329 | static struct bgp_info * | |
2330 | rfapiItBiIndexSearch ( | |
2331 | struct route_node *rn, /* Import table VPN node */ | |
2332 | struct prefix_rd *prd, | |
2333 | struct peer *peer, | |
2334 | struct prefix *aux_prefix) /* optional L3 addr for L2 ITs */ | |
2335 | { | |
2336 | struct skiplist *sl; | |
2337 | int rc; | |
2338 | struct bgp_info bi_fake; | |
2339 | struct bgp_info_extra bi_extra; | |
2340 | struct bgp_info *bi_result; | |
2341 | ||
2342 | sl = RFAPI_RDINDEX (rn); | |
2343 | if (!sl) | |
2344 | return NULL; | |
2345 | ||
2346 | #if DEBUG_BI_SEARCH | |
2347 | { | |
2348 | char buf[BUFSIZ]; | |
2349 | char buf_aux_pfx[BUFSIZ]; | |
2350 | ||
2351 | prefix_rd2str (prd, buf, BUFSIZ); | |
2352 | if (aux_prefix) | |
2353 | { | |
2354 | prefix2str (aux_prefix, buf_aux_pfx, BUFSIZ); | |
2355 | } | |
2356 | else | |
2357 | { | |
2358 | strncpy (buf_aux_pfx, "(nil)", BUFSIZ - 1); | |
2359 | buf_aux_pfx[BUFSIZ - 1] = 0; | |
2360 | } | |
2361 | ||
a3b55c25 | 2362 | vnc_zlog_debug_verbose ("%s want prd=%s, peer=%p, aux_prefix=%s", |
65efcfce LB |
2363 | __func__, buf, peer, buf_aux_pfx); |
2364 | rfapiItBiIndexDump (rn); | |
2365 | } | |
2366 | #endif | |
2367 | ||
2368 | /* threshold is a WAG */ | |
2369 | if (sl->count < 3) | |
2370 | { | |
2371 | #if DEBUG_BI_SEARCH | |
a3b55c25 | 2372 | vnc_zlog_debug_verbose ("%s: short list algorithm", __func__); |
65efcfce LB |
2373 | #endif |
2374 | /* if short list, linear search might be faster */ | |
2375 | for (bi_result = rn->info; bi_result; bi_result = bi_result->next) | |
2376 | { | |
2377 | #if DEBUG_BI_SEARCH | |
2378 | { | |
2379 | char buf[BUFSIZ]; | |
2380 | prefix_rd2str (&bi_result->extra->vnc.import.rd, buf, BUFSIZ); | |
a3b55c25 | 2381 | vnc_zlog_debug_verbose ("%s: bi has prd=%s, peer=%p", __func__, |
65efcfce LB |
2382 | buf, bi_result->peer); |
2383 | } | |
2384 | #endif | |
2385 | if (peer == bi_result->peer && | |
2386 | !prefix_cmp ((struct prefix *) &bi_result->extra->vnc.import.rd, | |
2387 | (struct prefix *) prd)) | |
2388 | { | |
2389 | ||
2390 | #if DEBUG_BI_SEARCH | |
a3b55c25 | 2391 | vnc_zlog_debug_verbose ("%s: peer and RD same, doing aux_prefix check", |
65efcfce LB |
2392 | __func__); |
2393 | #endif | |
2394 | if (!aux_prefix || | |
2395 | !prefix_cmp (aux_prefix, | |
2396 | &bi_result->extra->vnc.import.aux_prefix)) | |
2397 | { | |
2398 | ||
2399 | #if DEBUG_BI_SEARCH | |
a3b55c25 | 2400 | vnc_zlog_debug_verbose ("%s: match", __func__); |
65efcfce LB |
2401 | #endif |
2402 | break; | |
2403 | } | |
2404 | ||
2405 | } | |
2406 | } | |
2407 | return bi_result; | |
2408 | } | |
2409 | ||
2410 | bi_fake.peer = peer; | |
2411 | bi_fake.extra = &bi_extra; | |
2412 | bi_fake.extra->vnc.import.rd = *(struct prefix_rd *) prd; | |
2413 | if (aux_prefix) | |
2414 | { | |
2415 | bi_fake.extra->vnc.import.aux_prefix = *aux_prefix; | |
2416 | } | |
2417 | else | |
2418 | { | |
2419 | /* wildcard */ | |
2420 | bi_fake.extra->vnc.import.aux_prefix.family = AF_ETHERNET; | |
2421 | bi_fake.extra->vnc.import.aux_prefix.prefixlen = 1; | |
2422 | } | |
2423 | ||
2424 | rc = skiplist_search (sl, (void *) &bi_fake, (void *) &bi_result); | |
2425 | ||
2426 | if (rc) | |
2427 | { | |
2428 | #if DEBUG_BI_SEARCH | |
a3b55c25 | 2429 | vnc_zlog_debug_verbose ("%s: no match", __func__); |
65efcfce LB |
2430 | #endif |
2431 | return NULL; | |
2432 | } | |
2433 | ||
2434 | #if DEBUG_BI_SEARCH | |
a3b55c25 | 2435 | vnc_zlog_debug_verbose ("%s: matched bi=%p", __func__, bi_result); |
65efcfce LB |
2436 | #endif |
2437 | ||
2438 | return bi_result; | |
2439 | } | |
2440 | ||
2441 | static void | |
2442 | rfapiItBiIndexDel ( | |
2443 | struct route_node *rn, /* Import table VPN node */ | |
2444 | struct bgp_info *bi) /* old BI */ | |
2445 | { | |
2446 | struct skiplist *sl; | |
2447 | int rc; | |
2448 | ||
2449 | { | |
2450 | char buf[BUFSIZ]; | |
2451 | prefix_rd2str (&bi->extra->vnc.import.rd, buf, BUFSIZ); | |
a3b55c25 | 2452 | vnc_zlog_debug_verbose ("%s: bi %p, peer %p, rd %s", __func__, bi, bi->peer, buf); |
65efcfce LB |
2453 | } |
2454 | ||
2455 | sl = RFAPI_RDINDEX (rn); | |
2456 | assert (sl); | |
2457 | ||
2458 | rc = skiplist_delete (sl, (void *) (bi), (void *) bi); | |
2459 | if (rc) | |
2460 | { | |
2461 | rfapiItBiIndexDump (rn); | |
2462 | } | |
2463 | assert (!rc); | |
2464 | ||
2465 | route_unlock_node (rn); /* for skiplist entry */ | |
2466 | ||
2467 | /* NB: BIs in import tables are not refcounted */ | |
2468 | } | |
2469 | ||
2470 | /* | |
2471 | * Add a backreference at the ENCAP node to the VPN route that | |
2472 | * refers to it | |
2473 | */ | |
2474 | static void | |
2475 | rfapiMonitorEncapAdd ( | |
2476 | struct rfapi_import_table *import_table, | |
2477 | struct prefix *p, /* VN address */ | |
2478 | struct route_node *vpn_rn, /* VPN node */ | |
2479 | struct bgp_info *vpn_bi) /* VPN bi/route */ | |
2480 | { | |
2481 | afi_t afi = family2afi (p->family); | |
2482 | struct route_node *rn; | |
2483 | struct rfapi_monitor_encap *m; | |
2484 | ||
2485 | assert (afi); | |
2486 | rn = route_node_get (import_table->imported_encap[afi], p); /* locks rn */ | |
2487 | assert (rn); | |
2488 | ||
2489 | m = | |
2490 | XCALLOC (MTYPE_RFAPI_MONITOR_ENCAP, sizeof (struct rfapi_monitor_encap)); | |
2491 | assert (m); | |
2492 | ||
2493 | m->node = vpn_rn; | |
2494 | m->bi = vpn_bi; | |
2495 | m->rn = rn; | |
2496 | ||
2497 | /* insert to encap node's list */ | |
2498 | m->next = RFAPI_MONITOR_ENCAP (rn); | |
2499 | if (m->next) | |
2500 | m->next->prev = m; | |
2501 | RFAPI_MONITOR_ENCAP_W_ALLOC (rn) = m; | |
2502 | ||
2503 | /* for easy lookup when deleting vpn route */ | |
2504 | vpn_bi->extra->vnc.import.hme = m; | |
2505 | ||
a3b55c25 | 2506 | vnc_zlog_debug_verbose |
65efcfce LB |
2507 | ("%s: it=%p, vpn_bi=%p, afi=%d, encap rn=%p, setting vpn_bi->extra->vnc.import.hme=%p", |
2508 | __func__, import_table, vpn_bi, afi, rn, m); | |
2509 | ||
2510 | RFAPI_CHECK_REFCOUNT (rn, SAFI_ENCAP, 0); | |
bede7744 | 2511 | bgp_attr_intern (vpn_bi->attr); |
65efcfce LB |
2512 | } |
2513 | ||
2514 | static void | |
2515 | rfapiMonitorEncapDelete (struct bgp_info *vpn_bi) | |
2516 | { | |
2517 | /* | |
2518 | * Remove encap monitor | |
2519 | */ | |
a3b55c25 | 2520 | vnc_zlog_debug_verbose ("%s: vpn_bi=%p", __func__, vpn_bi); |
65efcfce LB |
2521 | if (vpn_bi->extra) |
2522 | { | |
2523 | struct rfapi_monitor_encap *hme = vpn_bi->extra->vnc.import.hme; | |
2524 | ||
2525 | if (hme) | |
2526 | { | |
2527 | ||
a3b55c25 | 2528 | vnc_zlog_debug_verbose ("%s: hme=%p", __func__, hme); |
65efcfce LB |
2529 | |
2530 | /* Refcount checking takes too long here */ | |
2531 | //RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 0); | |
2532 | if (hme->next) | |
2533 | hme->next->prev = hme->prev; | |
2534 | if (hme->prev) | |
2535 | hme->prev->next = hme->next; | |
2536 | else | |
2537 | RFAPI_MONITOR_ENCAP_W_ALLOC (hme->rn) = hme->next; | |
2538 | /* Refcount checking takes too long here */ | |
2539 | //RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 1); | |
2540 | ||
2541 | /* see if the struct rfapi_it_extra is empty and can be freed */ | |
2542 | rfapiMonitorExtraPrune (SAFI_ENCAP, hme->rn); | |
2543 | ||
2544 | route_unlock_node (hme->rn); /* decr ref count */ | |
2545 | XFREE (MTYPE_RFAPI_MONITOR_ENCAP, hme); | |
2546 | vpn_bi->extra->vnc.import.hme = NULL; | |
2547 | } | |
2548 | } | |
2549 | } | |
2550 | ||
2551 | /* | |
2552 | * quagga lib/thread.h says this must return int even though | |
2553 | * it doesn't do anything with the return value | |
2554 | */ | |
2555 | static int | |
2556 | rfapiWithdrawTimerVPN (struct thread *t) | |
2557 | { | |
2558 | struct rfapi_withdraw *wcb = t->arg; | |
2559 | struct bgp_info *bi = wcb->info; | |
2560 | struct bgp *bgp = bgp_get_default (); | |
2561 | ||
2562 | struct rfapi_monitor_vpn *moved; | |
2563 | afi_t afi; | |
2564 | ||
2565 | assert (wcb->node); | |
2566 | assert (bi); | |
2567 | assert (wcb->import_table); | |
2568 | assert (bi->extra); | |
2569 | ||
2570 | RFAPI_CHECK_REFCOUNT (wcb->node, SAFI_MPLS_VPN, wcb->lockoffset); | |
2571 | ||
2572 | { | |
2573 | char buf[BUFSIZ]; | |
2574 | ||
a3b55c25 | 2575 | vnc_zlog_debug_verbose ("%s: removing bi %p at prefix %s/%d", |
65efcfce LB |
2576 | __func__, |
2577 | bi, | |
2578 | rfapi_ntop (wcb->node->p.family, &wcb->node->p.u.prefix, buf, | |
2579 | BUFSIZ), wcb->node->p.prefixlen); | |
2580 | } | |
2581 | ||
2582 | /* | |
2583 | * Remove the route (doubly-linked) | |
2584 | */ | |
2585 | if (CHECK_FLAG (bi->flags, BGP_INFO_VALID) | |
2586 | && VALID_INTERIOR_TYPE (bi->type)) | |
2587 | RFAPI_MONITOR_EXTERIOR (wcb->node)->valid_interior_count--; | |
2588 | ||
2589 | afi = family2afi (wcb->node->p.family); | |
2590 | wcb->import_table->holddown_count[afi] -= 1; /* keep count consistent */ | |
2591 | rfapiItBiIndexDel (wcb->node, bi); | |
2592 | rfapiBgpInfoDetach (wcb->node, bi); /* with removed bi */ | |
2593 | ||
2594 | vnc_import_bgp_exterior_del_route_interior (bgp, wcb->import_table, | |
2595 | wcb->node, bi); | |
2596 | ||
2597 | ||
2598 | /* | |
2599 | * If VNC is configured to send response remove messages, AND | |
2600 | * if the removed route had a UN address, do response removal | |
2601 | * processing. | |
2602 | */ | |
2603 | if (!(bgp->rfapi_cfg->flags & BGP_VNC_CONFIG_RESPONSE_REMOVAL_DISABLE)) | |
2604 | { | |
2605 | ||
2606 | int has_valid_duplicate = 0; | |
2607 | struct bgp_info *bii; | |
2608 | ||
2609 | /* | |
2610 | * First check if there are any OTHER routes at this node | |
2611 | * that have the same nexthop and a valid UN address. If | |
2612 | * there are (e.g., from other peers), then the route isn't | |
2613 | * really gone, so skip sending a response removal message. | |
2614 | */ | |
2615 | for (bii = wcb->node->info; bii; bii = bii->next) | |
2616 | { | |
2617 | if (rfapiVpnBiSamePtUn (bi, bii)) | |
2618 | { | |
2619 | has_valid_duplicate = 1; | |
2620 | break; | |
2621 | } | |
2622 | } | |
2623 | ||
a3b55c25 | 2624 | vnc_zlog_debug_verbose ("%s: has_valid_duplicate=%d", __func__, |
65efcfce LB |
2625 | has_valid_duplicate); |
2626 | ||
2627 | if (!has_valid_duplicate) | |
2628 | { | |
2629 | rfapiRibPendingDeleteRoute (bgp, wcb->import_table, afi, wcb->node); | |
2630 | } | |
2631 | } | |
2632 | ||
2633 | rfapiMonitorEncapDelete (bi); | |
2634 | ||
2635 | /* | |
2636 | * If there are no VPN monitors at this VPN Node A, | |
2637 | * we are done | |
2638 | */ | |
2639 | if (!RFAPI_MONITOR_VPN (wcb->node)) | |
2640 | { | |
a3b55c25 | 2641 | vnc_zlog_debug_verbose ("%s: no VPN monitors at this node", __func__); |
65efcfce LB |
2642 | goto done; |
2643 | } | |
2644 | ||
2645 | /* | |
2646 | * rfapiMonitorMoveShorter only moves monitors if there are | |
2647 | * no remaining valid routes at the current node | |
2648 | */ | |
2649 | moved = rfapiMonitorMoveShorter (wcb->node, 1); | |
2650 | ||
2651 | if (moved) | |
2652 | { | |
2653 | rfapiMonitorMovedUp (wcb->import_table, wcb->node, moved->node, moved); | |
2654 | } | |
2655 | ||
2656 | done: | |
2657 | /* | |
2658 | * Free VPN bi | |
2659 | */ | |
2660 | rfapiBgpInfoFree (bi); | |
2661 | wcb->info = NULL; | |
2662 | ||
2663 | /* | |
2664 | * If route count at this node has gone to 0, withdraw exported prefix | |
2665 | */ | |
2666 | if (!wcb->node->info) | |
2667 | { | |
2668 | /* see if the struct rfapi_it_extra is empty and can be freed */ | |
2669 | rfapiMonitorExtraPrune (SAFI_MPLS_VPN, wcb->node); | |
2670 | vnc_direct_bgp_del_prefix (bgp, wcb->import_table, wcb->node); | |
2671 | vnc_zebra_del_prefix (bgp, wcb->import_table, wcb->node); | |
2672 | } | |
2673 | else | |
2674 | { | |
2675 | /* | |
2676 | * nexthop change event | |
2677 | * vnc_direct_bgp_add_prefix() will recompute the VN addr ecommunity | |
2678 | */ | |
2679 | vnc_direct_bgp_add_prefix (bgp, wcb->import_table, wcb->node); | |
2680 | } | |
2681 | ||
2682 | RFAPI_CHECK_REFCOUNT (wcb->node, SAFI_MPLS_VPN, 1 + wcb->lockoffset); | |
2683 | route_unlock_node (wcb->node); /* decr ref count */ | |
2684 | XFREE (MTYPE_RFAPI_WITHDRAW, wcb); | |
2685 | return 0; | |
2686 | } | |
2687 | ||
2688 | /* | |
2689 | * This works for multiprotocol extension, but not for plain ol' | |
2690 | * unicast IPv4 because that nexthop is stored in attr->nexthop | |
2691 | */ | |
2692 | void | |
2693 | rfapiNexthop2Prefix (struct attr *attr, struct prefix *p) | |
2694 | { | |
2695 | assert (p); | |
2696 | assert (attr); | |
2697 | assert (attr->extra); | |
2698 | ||
2699 | memset (p, 0, sizeof (struct prefix)); | |
2700 | ||
2701 | switch (p->family = BGP_MP_NEXTHOP_FAMILY (attr->extra->mp_nexthop_len)) | |
2702 | { | |
2703 | case AF_INET: | |
2704 | p->u.prefix4 = attr->extra->mp_nexthop_global_in; | |
2705 | p->prefixlen = 32; | |
2706 | break; | |
2707 | ||
2708 | case AF_INET6: | |
2709 | p->u.prefix6 = attr->extra->mp_nexthop_global; | |
2710 | p->prefixlen = 128; | |
2711 | break; | |
2712 | ||
2713 | default: | |
a3b55c25 | 2714 | vnc_zlog_debug_verbose ("%s: Family is unknown = %d", |
65efcfce LB |
2715 | __func__, p->family); |
2716 | } | |
2717 | } | |
2718 | ||
2719 | void | |
2720 | rfapiUnicastNexthop2Prefix (afi_t afi, struct attr *attr, struct prefix *p) | |
2721 | { | |
2722 | if (afi == AFI_IP) | |
2723 | { | |
2724 | p->family = AF_INET; | |
2725 | p->prefixlen = 32; | |
2726 | p->u.prefix4 = attr->nexthop; | |
2727 | } | |
2728 | else | |
2729 | { | |
2730 | rfapiNexthop2Prefix (attr, p); | |
2731 | } | |
2732 | } | |
2733 | ||
2734 | static int | |
2735 | rfapiAttrNexthopAddrDifferent (struct prefix *p1, struct prefix *p2) | |
2736 | { | |
2737 | if (!p1 || !p2) | |
2738 | { | |
a3b55c25 | 2739 | vnc_zlog_debug_verbose ("%s: p1 or p2 is NULL", __func__); |
65efcfce LB |
2740 | return 1; |
2741 | } | |
2742 | ||
2743 | /* | |
2744 | * Are address families the same? | |
2745 | */ | |
2746 | if (p1->family != p2->family) | |
2747 | { | |
2748 | return 1; | |
2749 | } | |
2750 | ||
2751 | switch (p1->family) | |
2752 | { | |
2753 | case AF_INET: | |
2754 | if (IPV4_ADDR_SAME (&p1->u.prefix4, &p2->u.prefix4)) | |
2755 | return 0; | |
2756 | break; | |
2757 | ||
2758 | case AF_INET6: | |
2759 | if (IPV6_ADDR_SAME (&p1->u.prefix6, &p2->u.prefix6)) | |
2760 | return 0; | |
2761 | break; | |
2762 | ||
2763 | default: | |
2764 | assert (1); | |
2765 | ||
2766 | } | |
2767 | ||
2768 | return 1; | |
2769 | } | |
2770 | ||
2771 | static void | |
2772 | rfapiCopyUnEncap2VPN (struct bgp_info *encap_bi, struct bgp_info *vpn_bi) | |
2773 | { | |
2774 | struct attr_extra *attre; | |
2775 | ||
2776 | if (!encap_bi->attr || !encap_bi->attr->extra) | |
2777 | { | |
2778 | zlog_warn ("%s: no encap bi attr/extra, can't copy UN address", | |
2779 | __func__); | |
2780 | return; | |
2781 | } | |
2782 | ||
2783 | if (!vpn_bi || !vpn_bi->extra) | |
2784 | { | |
2785 | zlog_warn ("%s: no vpn bi attr/extra, can't copy UN address", | |
2786 | __func__); | |
2787 | return; | |
2788 | } | |
2789 | ||
2790 | attre = encap_bi->attr->extra; | |
2791 | ||
2792 | switch (BGP_MP_NEXTHOP_FAMILY (attre->mp_nexthop_len)) | |
2793 | { | |
2794 | case AF_INET: | |
2795 | ||
2796 | /* | |
2797 | * instrumentation to debug segfault of 091127 | |
2798 | */ | |
a3b55c25 | 2799 | vnc_zlog_debug_verbose ("%s: vpn_bi=%p", __func__, vpn_bi); |
65efcfce LB |
2800 | if (vpn_bi) |
2801 | { | |
a3b55c25 | 2802 | vnc_zlog_debug_verbose ("%s: vpn_bi->extra=%p", __func__, vpn_bi->extra); |
65efcfce LB |
2803 | } |
2804 | ||
2805 | vpn_bi->extra->vnc.import.un_family = AF_INET; | |
2806 | vpn_bi->extra->vnc.import.un.addr4 = attre->mp_nexthop_global_in; | |
2807 | break; | |
2808 | ||
2809 | case AF_INET6: | |
2810 | vpn_bi->extra->vnc.import.un_family = AF_INET6; | |
2811 | vpn_bi->extra->vnc.import.un.addr6 = attre->mp_nexthop_global; | |
2812 | break; | |
2813 | ||
2814 | default: | |
2815 | zlog_warn ("%s: invalid encap nexthop length: %d", | |
2816 | __func__, attre->mp_nexthop_len); | |
2817 | vpn_bi->extra->vnc.import.un_family = 0; | |
2818 | break; | |
2819 | } | |
2820 | } | |
2821 | ||
2822 | /* | |
2823 | * returns 0 on success, nonzero on error | |
2824 | */ | |
2825 | static int | |
2826 | rfapiWithdrawEncapUpdateCachedUn ( | |
2827 | struct rfapi_import_table *import_table, | |
2828 | struct bgp_info *encap_bi, | |
2829 | struct route_node *vpn_rn, | |
2830 | struct bgp_info *vpn_bi) | |
2831 | { | |
2832 | if (!encap_bi) | |
2833 | { | |
2834 | ||
2835 | /* | |
2836 | * clear cached UN address | |
2837 | */ | |
2838 | if (!vpn_bi || !vpn_bi->extra) | |
2839 | { | |
2840 | zlog_warn ("%s: missing VPN bi/extra, can't clear UN addr", | |
2841 | __func__); | |
2842 | return 1; | |
2843 | } | |
2844 | vpn_bi->extra->vnc.import.un_family = 0; | |
2845 | memset (&vpn_bi->extra->vnc.import.un, 0, | |
2846 | sizeof (vpn_bi->extra->vnc.import.un)); | |
2847 | if (CHECK_FLAG (vpn_bi->flags, BGP_INFO_VALID)) | |
2848 | { | |
2849 | if (rfapiGetVncTunnelUnAddr (vpn_bi->attr, NULL)) | |
2850 | { | |
2851 | UNSET_FLAG (vpn_bi->flags, BGP_INFO_VALID); | |
2852 | if (VALID_INTERIOR_TYPE (vpn_bi->type)) | |
2853 | RFAPI_MONITOR_EXTERIOR (vpn_rn)->valid_interior_count--; | |
2854 | /* signal interior route withdrawal to import-exterior */ | |
2855 | vnc_import_bgp_exterior_del_route_interior (bgp_get_default (), | |
2856 | import_table, | |
2857 | vpn_rn, vpn_bi); | |
2858 | } | |
2859 | } | |
2860 | ||
2861 | } | |
2862 | else | |
2863 | { | |
2864 | if (!vpn_bi) | |
2865 | { | |
2866 | zlog_warn ("%s: missing VPN bi, can't clear UN addr", __func__); | |
2867 | return 1; | |
2868 | } | |
2869 | rfapiCopyUnEncap2VPN (encap_bi, vpn_bi); | |
2870 | if (!CHECK_FLAG (vpn_bi->flags, BGP_INFO_VALID)) | |
2871 | { | |
2872 | SET_FLAG (vpn_bi->flags, BGP_INFO_VALID); | |
2873 | if (VALID_INTERIOR_TYPE (vpn_bi->type)) | |
2874 | RFAPI_MONITOR_EXTERIOR (vpn_rn)->valid_interior_count++; | |
2875 | /* signal interior route withdrawal to import-exterior */ | |
2876 | vnc_import_bgp_exterior_add_route_interior (bgp_get_default (), | |
2877 | import_table, | |
2878 | vpn_rn, vpn_bi); | |
2879 | } | |
2880 | } | |
2881 | return 0; | |
2882 | } | |
2883 | ||
2884 | static int | |
2885 | rfapiWithdrawTimerEncap (struct thread *t) | |
2886 | { | |
2887 | struct rfapi_withdraw *wcb = t->arg; | |
2888 | struct bgp_info *bi = wcb->info; | |
2889 | int was_first_route = 0; | |
2890 | struct rfapi_monitor_encap *em; | |
2891 | struct skiplist *vpn_node_sl = skiplist_new (0, NULL, NULL); | |
2892 | ||
2893 | assert (wcb->node); | |
2894 | assert (bi); | |
2895 | assert (wcb->import_table); | |
2896 | ||
2897 | RFAPI_CHECK_REFCOUNT (wcb->node, SAFI_ENCAP, 0); | |
2898 | ||
2899 | if (wcb->node->info == bi) | |
2900 | was_first_route = 1; | |
2901 | ||
2902 | /* | |
2903 | * Remove the route/bi and free it | |
2904 | */ | |
2905 | rfapiBgpInfoDetach (wcb->node, bi); | |
2906 | rfapiBgpInfoFree (bi); | |
2907 | ||
2908 | if (!was_first_route) | |
2909 | goto done; | |
2910 | ||
2911 | for (em = RFAPI_MONITOR_ENCAP (wcb->node); em; em = em->next) | |
2912 | { | |
2913 | ||
2914 | /* | |
2915 | * Update monitoring VPN BIs with new encap info at the | |
2916 | * head of the encap bi chain (which could be NULL after | |
2917 | * removing the expiring bi above) | |
2918 | */ | |
2919 | if (rfapiWithdrawEncapUpdateCachedUn | |
2920 | (wcb->import_table, wcb->node->info, em->node, em->bi)) | |
2921 | continue; | |
2922 | ||
2923 | /* | |
2924 | * Build a list of unique VPN nodes referenced by these monitors. | |
2925 | * Use a skiplist for speed. | |
2926 | */ | |
2927 | skiplist_insert (vpn_node_sl, em->node, em->node); | |
2928 | } | |
2929 | ||
2930 | ||
2931 | /* | |
2932 | * for each VPN node referenced in the ENCAP monitors: | |
2933 | */ | |
2934 | struct route_node *rn; | |
2935 | while (!skiplist_first (vpn_node_sl, (void **) &rn, NULL)) | |
2936 | { | |
2937 | if (!wcb->node->info) | |
2938 | { | |
2939 | struct rfapi_monitor_vpn *moved; | |
2940 | ||
2941 | moved = rfapiMonitorMoveShorter (rn, 0); | |
2942 | if (moved) | |
2943 | { | |
2944 | //rfapiDoRouteCallback(wcb->import_table, moved->node, moved); | |
2945 | rfapiMonitorMovedUp (wcb->import_table, rn, moved->node, moved); | |
2946 | } | |
2947 | } | |
2948 | else | |
2949 | { | |
2950 | //rfapiDoRouteCallback(wcb->import_table, rn, NULL); | |
2951 | rfapiMonitorItNodeChanged (wcb->import_table, rn, NULL); | |
2952 | } | |
2953 | skiplist_delete_first (vpn_node_sl); | |
2954 | } | |
2955 | ||
2956 | done: | |
2957 | RFAPI_CHECK_REFCOUNT (wcb->node, SAFI_ENCAP, 1); | |
2958 | route_unlock_node (wcb->node); /* decr ref count */ | |
2959 | XFREE (MTYPE_RFAPI_WITHDRAW, wcb); | |
2960 | skiplist_free (vpn_node_sl); | |
2961 | return 0; | |
2962 | } | |
2963 | ||
2964 | ||
2965 | /* | |
2966 | * Works for both VPN and ENCAP routes; timer_service_func is different | |
2967 | * in each case | |
2968 | */ | |
2969 | static void | |
2970 | rfapiBiStartWithdrawTimer ( | |
2971 | struct rfapi_import_table *import_table, | |
2972 | struct route_node *rn, | |
2973 | struct bgp_info *bi, | |
2974 | afi_t afi, | |
2975 | safi_t safi, | |
2976 | int (*timer_service_func) (struct thread *)) | |
2977 | { | |
2978 | uint32_t lifetime; | |
2979 | struct rfapi_withdraw *wcb; | |
2980 | ||
2981 | if CHECK_FLAG | |
2982 | (bi->flags, BGP_INFO_REMOVED) | |
2983 | { | |
2984 | /* | |
2985 | * Already on the path to being withdrawn, | |
2986 | * should already have a timer set up to | |
2987 | * delete it. | |
2988 | */ | |
a3b55c25 | 2989 | vnc_zlog_debug_verbose ("%s: already being withdrawn, do nothing", __func__); |
65efcfce LB |
2990 | return; |
2991 | } | |
2992 | ||
2993 | rfapiGetVncLifetime (bi->attr, &lifetime); | |
a3b55c25 | 2994 | vnc_zlog_debug_verbose ("%s: VNC lifetime is %u", __func__, lifetime); |
65efcfce LB |
2995 | |
2996 | /* | |
2997 | * withdrawn routes get to hang around for a while | |
2998 | */ | |
2999 | SET_FLAG (bi->flags, BGP_INFO_REMOVED); | |
3000 | ||
3001 | /* set timer to remove the route later */ | |
3002 | lifetime = rfapiGetHolddownFromLifetime (lifetime); | |
a3b55c25 | 3003 | vnc_zlog_debug_verbose ("%s: using timeout %u", __func__, lifetime); |
65efcfce LB |
3004 | |
3005 | /* | |
3006 | * Stash import_table, node, and info for use by timer | |
3007 | * service routine, which is supposed to free the wcb. | |
3008 | */ | |
3009 | wcb = XCALLOC (MTYPE_RFAPI_WITHDRAW, sizeof (struct rfapi_withdraw)); | |
3010 | assert (wcb); | |
3011 | wcb->node = rn; | |
3012 | wcb->info = bi; | |
3013 | wcb->import_table = import_table; | |
bede7744 | 3014 | bgp_attr_intern (bi->attr); |
65efcfce | 3015 | |
a3b55c25 LB |
3016 | if (VNC_DEBUG(VERBOSE)) |
3017 | { | |
3018 | vnc_zlog_debug_verbose | |
3019 | ("%s: wcb values: node=%p, info=%p, import_table=%p (bi follows)", | |
3020 | __func__, wcb->node, wcb->info, wcb->import_table); | |
3021 | rfapiPrintBi (NULL, bi); | |
3022 | } | |
65efcfce LB |
3023 | |
3024 | ||
3025 | assert (bi->extra); | |
3026 | if (lifetime > UINT32_MAX / 1001) | |
3027 | { | |
3028 | /* sub-optimal case, but will probably never happen */ | |
3029 | bi->extra->vnc.import.timer = thread_add_timer (bm->master, | |
3030 | timer_service_func, | |
3031 | wcb, lifetime); | |
3032 | } | |
3033 | else | |
3034 | { | |
3035 | static uint32_t jitter; | |
3036 | uint32_t lifetime_msec; | |
3037 | ||
3038 | /* | |
3039 | * the goal here is to spread out the timers so they are | |
3040 | * sortable in the skip list | |
3041 | */ | |
3042 | if (++jitter >= 1000) | |
3043 | jitter = 0; | |
3044 | ||
3045 | lifetime_msec = (lifetime * 1000) + jitter; | |
3046 | ||
3047 | bi->extra->vnc.import.timer = thread_add_background (bm->master, | |
3048 | timer_service_func, | |
3049 | wcb, | |
3050 | lifetime_msec); | |
3051 | } | |
3052 | ||
3053 | /* re-sort route list (BGP_INFO_REMOVED routes are last) */ | |
3054 | if (((struct bgp_info *) rn->info)->next) | |
3055 | { | |
3056 | rfapiBgpInfoDetach (rn, bi); | |
3057 | rfapiBgpInfoAttachSorted (rn, bi, afi, safi); | |
3058 | } | |
3059 | } | |
3060 | ||
3061 | ||
3062 | typedef void (rfapi_bi_filtered_import_f) (struct rfapi_import_table *, | |
3063 | int, | |
3064 | struct peer *, | |
3065 | void *, | |
3066 | struct prefix *, | |
3067 | struct prefix *, | |
3068 | afi_t, | |
3069 | struct prefix_rd *, | |
3070 | struct attr *, | |
3071 | u_char, u_char, uint32_t *); | |
3072 | ||
3073 | ||
3074 | static void | |
3075 | rfapiExpireEncapNow ( | |
3076 | struct rfapi_import_table *it, | |
3077 | struct route_node *rn, | |
3078 | struct bgp_info *bi) | |
3079 | { | |
3080 | struct rfapi_withdraw *wcb; | |
3081 | struct thread t; | |
3082 | ||
3083 | /* | |
3084 | * pretend we're an expiring timer | |
3085 | */ | |
3086 | wcb = XCALLOC (MTYPE_RFAPI_WITHDRAW, sizeof (struct rfapi_withdraw)); | |
3087 | wcb->info = bi; | |
3088 | wcb->node = rn; | |
3089 | wcb->import_table = it; | |
3090 | memset (&t, 0, sizeof (t)); | |
3091 | t.arg = wcb; | |
3092 | rfapiWithdrawTimerEncap (&t); /* frees wcb */ | |
3093 | } | |
3094 | ||
3095 | static int | |
3096 | rfapiGetNexthop (struct attr *attr, struct prefix *prefix) | |
3097 | { | |
3098 | switch (BGP_MP_NEXTHOP_FAMILY (attr->extra->mp_nexthop_len)) | |
3099 | { | |
3100 | case AF_INET: | |
3101 | prefix->family = AF_INET; | |
3102 | prefix->prefixlen = 32; | |
3103 | prefix->u.prefix4 = attr->extra->mp_nexthop_global_in; | |
3104 | break; | |
3105 | case AF_INET6: | |
3106 | prefix->family = AF_INET6; | |
3107 | prefix->prefixlen = 128; | |
3108 | prefix->u.prefix6 = attr->extra->mp_nexthop_global; | |
3109 | break; | |
3110 | default: | |
a3b55c25 | 3111 | vnc_zlog_debug_verbose ("%s: unknown attr->extra->mp_nexthop_len %d", __func__, |
65efcfce LB |
3112 | attr->extra->mp_nexthop_len); |
3113 | return EINVAL; | |
3114 | } | |
3115 | return 0; | |
3116 | } | |
3117 | ||
3118 | /* | |
3119 | * import a bgp_info if its route target list intersects with the | |
3120 | * import table's route target list | |
3121 | */ | |
3122 | static void | |
3123 | rfapiBgpInfoFilteredImportEncap ( | |
3124 | struct rfapi_import_table *import_table, | |
3125 | int action, | |
3126 | struct peer *peer, | |
3127 | void *rfd, /* set for looped back routes */ | |
3128 | struct prefix *p, | |
3129 | struct prefix *aux_prefix, /* Unused for encap routes */ | |
3130 | afi_t afi, | |
3131 | struct prefix_rd *prd, | |
3132 | struct attr *attr, /* part of bgp_info */ | |
3133 | u_char type, /* part of bgp_info */ | |
3134 | u_char sub_type, /* part of bgp_info */ | |
3135 | uint32_t *label) /* part of bgp_info */ | |
3136 | { | |
3137 | struct route_table *rt = NULL; | |
3138 | struct route_node *rn; | |
3139 | struct bgp_info *info_new; | |
3140 | struct bgp_info *bi; | |
3141 | struct bgp_info *next; | |
3142 | char buf[BUFSIZ]; | |
3143 | ||
3144 | struct prefix p_firstbi_old; | |
3145 | struct prefix p_firstbi_new; | |
3146 | int replacing = 0; | |
3147 | const char *action_str = NULL; | |
3148 | struct prefix un_prefix; | |
3149 | ||
3150 | struct bgp *bgp; | |
3151 | bgp = bgp_get_default (); /* assume 1 instance for now */ | |
3152 | ||
3153 | switch (action) | |
3154 | { | |
3155 | case FIF_ACTION_UPDATE: | |
3156 | action_str = "update"; | |
3157 | break; | |
3158 | case FIF_ACTION_WITHDRAW: | |
3159 | action_str = "withdraw"; | |
3160 | break; | |
3161 | case FIF_ACTION_KILL: | |
3162 | action_str = "kill"; | |
3163 | break; | |
3164 | default: | |
3165 | assert (0); | |
3166 | break; | |
3167 | } | |
3168 | ||
a3b55c25 | 3169 | vnc_zlog_debug_verbose ("%s: entry: %s: prefix %s/%d", __func__, |
65efcfce LB |
3170 | action_str, |
3171 | inet_ntop (p->family, &p->u.prefix, buf, BUFSIZ), p->prefixlen); | |
3172 | ||
3173 | memset (&p_firstbi_old, 0, sizeof (p_firstbi_old)); | |
3174 | memset (&p_firstbi_new, 0, sizeof (p_firstbi_new)); | |
3175 | ||
3176 | if (action == FIF_ACTION_UPDATE) | |
3177 | { | |
3178 | /* | |
3179 | * Compare rt lists. If no intersection, don't import this route | |
3180 | * On a withdraw, peer and RD are sufficient to determine if | |
3181 | * we should act. | |
3182 | */ | |
3183 | if (!attr || !attr->extra || !attr->extra->ecommunity) | |
3184 | { | |
3185 | ||
a3b55c25 | 3186 | vnc_zlog_debug_verbose ("%s: attr, extra, or ecommunity missing, not importing", |
65efcfce LB |
3187 | __func__); |
3188 | return; | |
3189 | } | |
3190 | #if RFAPI_REQUIRE_ENCAP_BEEC | |
3191 | if (!rfapiEcommunitiesMatchBeec (attr->extra->ecommunity)) | |
3192 | { | |
a3b55c25 | 3193 | vnc_zlog_debug_verbose ("%s: it=%p: no match for BGP Encapsulation ecommunity", |
65efcfce LB |
3194 | __func__, import_table); |
3195 | return; | |
3196 | } | |
3197 | #endif | |
3198 | if (!rfapiEcommunitiesIntersect (import_table->rt_import_list, | |
3199 | attr->extra->ecommunity)) | |
3200 | { | |
3201 | ||
a3b55c25 | 3202 | vnc_zlog_debug_verbose ("%s: it=%p: no ecommunity intersection", |
65efcfce LB |
3203 | __func__, import_table); |
3204 | return; | |
3205 | } | |
3206 | ||
3207 | /* | |
3208 | * Updates must also have a nexthop address | |
3209 | */ | |
3210 | memset (&un_prefix, 0, sizeof (un_prefix)); /* keep valgrind happy */ | |
3211 | if (rfapiGetNexthop (attr, &un_prefix)) | |
3212 | { | |
a3b55c25 | 3213 | vnc_zlog_debug_verbose ("%s: missing nexthop address", __func__); |
65efcfce LB |
3214 | return; |
3215 | } | |
3216 | } | |
3217 | ||
3218 | /* | |
3219 | * Figure out which radix tree the route would go into | |
3220 | */ | |
3221 | switch (afi) | |
3222 | { | |
3223 | case AFI_IP: | |
3224 | case AFI_IP6: | |
3225 | rt = import_table->imported_encap[afi]; | |
3226 | break; | |
3227 | ||
3228 | default: | |
3229 | zlog_err ("%s: bad afi %d", __func__, afi); | |
3230 | return; | |
3231 | } | |
3232 | ||
3233 | /* | |
3234 | * route_node_lookup returns a node only if there is at least | |
3235 | * one route attached. | |
3236 | */ | |
3237 | rn = route_node_lookup (rt, p); | |
3238 | ||
3239 | #if DEBUG_ENCAP_MONITOR | |
a3b55c25 | 3240 | vnc_zlog_debug_verbose ("%s: initial encap lookup (it=%p) rn=%p", |
65efcfce LB |
3241 | __func__, import_table, rn); |
3242 | #endif | |
3243 | ||
3244 | if (rn) | |
3245 | { | |
3246 | ||
3247 | RFAPI_CHECK_REFCOUNT (rn, SAFI_ENCAP, 1); | |
3248 | route_unlock_node (rn); /* undo lock in route_node_lookup */ | |
3249 | ||
3250 | ||
3251 | /* | |
3252 | * capture nexthop of first bi | |
3253 | */ | |
3254 | if (rn->info) | |
3255 | { | |
3256 | rfapiNexthop2Prefix (((struct bgp_info *) (rn->info))->attr, | |
3257 | &p_firstbi_old); | |
3258 | } | |
3259 | ||
3260 | for (bi = rn->info; bi; bi = bi->next) | |
3261 | { | |
3262 | ||
3263 | /* | |
3264 | * Does this bgp_info refer to the same route | |
3265 | * as we are trying to add? | |
3266 | */ | |
a3b55c25 | 3267 | vnc_zlog_debug_verbose ("%s: comparing BI %p", __func__, bi); |
65efcfce LB |
3268 | |
3269 | ||
3270 | /* | |
3271 | * Compare RDs | |
3272 | * | |
3273 | * RD of import table bi is in bi->extra->vnc.import.rd | |
3274 | * RD of info_orig is in prd | |
3275 | */ | |
3276 | if (!bi->extra) | |
3277 | { | |
a3b55c25 | 3278 | vnc_zlog_debug_verbose ("%s: no bi->extra", __func__); |
65efcfce LB |
3279 | continue; |
3280 | } | |
3281 | if (prefix_cmp ((struct prefix *) &bi->extra->vnc.import.rd, | |
3282 | (struct prefix *) prd)) | |
3283 | { | |
3284 | ||
a3b55c25 | 3285 | vnc_zlog_debug_verbose ("%s: prd does not match", __func__); |
65efcfce LB |
3286 | continue; |
3287 | } | |
3288 | ||
3289 | /* | |
3290 | * Compare peers | |
3291 | */ | |
3292 | if (bi->peer != peer) | |
3293 | { | |
a3b55c25 | 3294 | vnc_zlog_debug_verbose ("%s: peer does not match", __func__); |
65efcfce LB |
3295 | continue; |
3296 | } | |
3297 | ||
a3b55c25 | 3298 | vnc_zlog_debug_verbose ("%s: found matching bi", __func__); |
65efcfce LB |
3299 | |
3300 | /* Same route. Delete this bi, replace with new one */ | |
3301 | ||
3302 | if (action == FIF_ACTION_WITHDRAW) | |
3303 | { | |
3304 | ||
a3b55c25 | 3305 | vnc_zlog_debug_verbose ("%s: withdrawing at prefix %s/%d", |
65efcfce LB |
3306 | __func__, |
3307 | inet_ntop (rn->p.family, &rn->p.u.prefix, buf, | |
3308 | BUFSIZ), rn->p.prefixlen); | |
3309 | ||
3310 | rfapiBiStartWithdrawTimer (import_table, rn, bi, | |
3311 | afi, SAFI_ENCAP, | |
3312 | rfapiWithdrawTimerEncap); | |
3313 | ||
3314 | } | |
3315 | else | |
3316 | { | |
a3b55c25 | 3317 | vnc_zlog_debug_verbose ("%s: %s at prefix %s/%d", |
65efcfce LB |
3318 | __func__, |
3319 | ((action == | |
3320 | FIF_ACTION_KILL) ? "killing" : "replacing"), | |
3321 | inet_ntop (rn->p.family, &rn->p.u.prefix, buf, | |
3322 | BUFSIZ), rn->p.prefixlen); | |
3323 | ||
3324 | /* | |
3325 | * If this route is waiting to be deleted because of | |
3326 | * a previous withdraw, we must cancel its timer. | |
3327 | */ | |
3328 | if (CHECK_FLAG (bi->flags, BGP_INFO_REMOVED) | |
3329 | && bi->extra->vnc.import.timer) | |
3330 | { | |
3331 | ||
3332 | struct thread *t = | |
3333 | (struct thread *) bi->extra->vnc.import.timer; | |
3334 | struct rfapi_withdraw *wcb = t->arg; | |
3335 | ||
3336 | XFREE (MTYPE_RFAPI_WITHDRAW, wcb); | |
3337 | thread_cancel (t); | |
3338 | } | |
3339 | ||
3340 | if (action == FIF_ACTION_UPDATE) | |
3341 | { | |
3342 | rfapiBgpInfoDetach (rn, bi); | |
3343 | rfapiBgpInfoFree (bi); | |
3344 | replacing = 1; | |
3345 | } | |
3346 | else | |
3347 | { | |
3348 | /* | |
3349 | * Kill: do export stuff when removing bi | |
3350 | */ | |
3351 | struct rfapi_withdraw *wcb; | |
3352 | struct thread t; | |
3353 | ||
3354 | /* | |
3355 | * pretend we're an expiring timer | |
3356 | */ | |
3357 | wcb = | |
3358 | XCALLOC (MTYPE_RFAPI_WITHDRAW, | |
3359 | sizeof (struct rfapi_withdraw)); | |
3360 | wcb->info = bi; | |
3361 | wcb->node = rn; | |
3362 | wcb->import_table = import_table; | |
3363 | memset (&t, 0, sizeof (t)); | |
3364 | t.arg = wcb; | |
3365 | rfapiWithdrawTimerEncap (&t); /* frees wcb */ | |
3366 | } | |
3367 | } | |
3368 | ||
3369 | break; | |
3370 | } | |
3371 | } | |
3372 | ||
3373 | if (rn) | |
3374 | RFAPI_CHECK_REFCOUNT (rn, SAFI_ENCAP, replacing ? 1 : 0); | |
3375 | ||
3376 | if (action == FIF_ACTION_WITHDRAW || action == FIF_ACTION_KILL) | |
3377 | return; | |
3378 | ||
3379 | info_new = rfapiBgpInfoCreate (attr, peer, rfd, prd, type, sub_type, NULL); | |
3380 | ||
3381 | if (rn) | |
3382 | { | |
3383 | if (!replacing) | |
3384 | route_lock_node (rn); /* incr ref count for new BI */ | |
3385 | } | |
3386 | else | |
3387 | { | |
3388 | rn = route_node_get (rt, p); | |
3389 | } | |
3390 | ||
a3b55c25 | 3391 | vnc_zlog_debug_verbose ("%s: (afi=%d, rn=%p) inserting at prefix %s/%d", |
65efcfce LB |
3392 | __func__, |
3393 | afi, | |
3394 | rn, | |
3395 | inet_ntop (rn->p.family, &rn->p.u.prefix, buf, BUFSIZ), | |
3396 | rn->p.prefixlen); | |
3397 | ||
3398 | rfapiBgpInfoAttachSorted (rn, info_new, afi, SAFI_ENCAP); | |
3399 | ||
3400 | /* | |
3401 | * Delete holddown routes from same NVE. See details in | |
3402 | * rfapiBgpInfoFilteredImportVPN() | |
3403 | */ | |
3404 | for (bi = info_new->next; bi; bi = next) | |
3405 | { | |
3406 | ||
3407 | struct prefix pfx_un; | |
3408 | int un_match = 0; | |
3409 | ||
3410 | next = bi->next; | |
3411 | if (!CHECK_FLAG (bi->flags, BGP_INFO_REMOVED)) | |
3412 | continue; | |
3413 | ||
3414 | /* | |
3415 | * We already match the VN address (it is the prefix | |
3416 | * of the route node) | |
3417 | */ | |
3418 | ||
3419 | if (!rfapiGetNexthop (bi->attr, &pfx_un) && | |
3420 | prefix_same (&pfx_un, &un_prefix)) | |
3421 | { | |
3422 | ||
3423 | un_match = 1; | |
3424 | } | |
3425 | ||
3426 | if (!un_match) | |
3427 | continue; | |
3428 | ||
a3b55c25 | 3429 | vnc_zlog_debug_verbose ("%s: removing holddown bi matching NVE of new route", |
65efcfce LB |
3430 | __func__); |
3431 | if (bi->extra->vnc.import.timer) | |
3432 | { | |
3433 | struct thread *t = (struct thread *) bi->extra->vnc.import.timer; | |
3434 | struct rfapi_withdraw *wcb = t->arg; | |
3435 | ||
3436 | XFREE (MTYPE_RFAPI_WITHDRAW, wcb); | |
3437 | thread_cancel (t); | |
3438 | } | |
3439 | rfapiExpireEncapNow (import_table, rn, bi); | |
3440 | } | |
3441 | ||
3442 | rfapiNexthop2Prefix (((struct bgp_info *) (rn->info))->attr, | |
3443 | &p_firstbi_new); | |
3444 | ||
3445 | /* | |
3446 | * If the nexthop address of the selected Encap route (i.e., | |
3447 | * the UN address) has changed, then we must update the VPN | |
3448 | * routes that refer to this Encap route and possibly force | |
3449 | * rfapi callbacks. | |
3450 | */ | |
3451 | if (rfapiAttrNexthopAddrDifferent (&p_firstbi_old, &p_firstbi_new)) | |
3452 | { | |
3453 | ||
3454 | struct rfapi_monitor_encap *m; | |
3455 | struct rfapi_monitor_encap *mnext; | |
3456 | ||
3457 | struct route_node *referenced_vpn_prefix; | |
3458 | ||
3459 | /* | |
3460 | * Optimized approach: build radix tree on the fly to | |
3461 | * hold list of VPN nodes referenced by the ENCAP monitors | |
3462 | * | |
3463 | * The nodes in this table correspond to prefixes of VPN routes. | |
3464 | * The "info" pointer of the node points to a chain of | |
3465 | * struct rfapi_monitor_encap, each of which refers to a | |
3466 | * specific VPN node. | |
3467 | */ | |
3468 | struct route_table *referenced_vpn_table; | |
3469 | ||
3470 | referenced_vpn_table = route_table_init (); | |
3471 | assert (referenced_vpn_table); | |
3472 | ||
3473 | /* | |
3474 | * iterate over the set of monitors at this ENCAP node. | |
3475 | */ | |
3476 | #if DEBUG_ENCAP_MONITOR | |
a3b55c25 | 3477 | vnc_zlog_debug_verbose ("%s: examining monitors at rn=%p", __func__, rn); |
65efcfce LB |
3478 | #endif |
3479 | for (m = RFAPI_MONITOR_ENCAP (rn); m; m = m->next) | |
3480 | { | |
3481 | ||
3482 | /* | |
3483 | * For each referenced bi/route, copy the ENCAP route's | |
3484 | * nexthop to the VPN route's cached UN address field and set | |
3485 | * the address family of the cached UN address field. | |
3486 | */ | |
3487 | rfapiCopyUnEncap2VPN (info_new, m->bi); | |
3488 | if (!CHECK_FLAG (m->bi->flags, BGP_INFO_VALID)) | |
3489 | { | |
3490 | SET_FLAG (m->bi->flags, BGP_INFO_VALID); | |
3491 | if (VALID_INTERIOR_TYPE (m->bi->type)) | |
3492 | RFAPI_MONITOR_EXTERIOR (m->node)->valid_interior_count++; | |
3493 | vnc_import_bgp_exterior_add_route_interior (bgp, | |
3494 | import_table, | |
3495 | m->node, m->bi); | |
3496 | } | |
3497 | ||
3498 | /* | |
3499 | * Build a list of unique VPN nodes referenced by these monitors | |
3500 | * | |
3501 | * There could be more than one VPN node here with a given | |
3502 | * prefix. Those are currently in an unsorted linear list | |
3503 | * per prefix. | |
3504 | */ | |
3505 | ||
3506 | referenced_vpn_prefix = | |
3507 | route_node_get (referenced_vpn_table, &m->node->p); | |
3508 | assert (referenced_vpn_prefix); | |
3509 | for (mnext = referenced_vpn_prefix->info; mnext; | |
3510 | mnext = mnext->next) | |
3511 | { | |
3512 | ||
3513 | if (mnext->node == m->node) | |
3514 | break; | |
3515 | } | |
3516 | ||
3517 | if (mnext) | |
3518 | { | |
3519 | /* | |
3520 | * already have an entry for this VPN node | |
3521 | */ | |
3522 | route_unlock_node (referenced_vpn_prefix); | |
3523 | } | |
3524 | else | |
3525 | { | |
3526 | mnext = XCALLOC (MTYPE_RFAPI_MONITOR_ENCAP, | |
3527 | sizeof (struct rfapi_monitor_encap)); | |
3528 | assert (mnext); | |
3529 | mnext->node = m->node; | |
3530 | mnext->next = referenced_vpn_prefix->info; | |
3531 | referenced_vpn_prefix->info = mnext; | |
3532 | } | |
3533 | ||
3534 | } | |
3535 | ||
3536 | /* | |
3537 | * for each VPN node referenced in the ENCAP monitors: | |
3538 | */ | |
3539 | for (referenced_vpn_prefix = route_top (referenced_vpn_table); | |
3540 | referenced_vpn_prefix; | |
3541 | referenced_vpn_prefix = route_next (referenced_vpn_prefix)) | |
3542 | { | |
3543 | ||
3544 | while ((m = referenced_vpn_prefix->info)) | |
3545 | { | |
3546 | ||
3547 | struct route_node *n; | |
3548 | ||
3549 | rfapiMonitorMoveLonger (m->node); | |
3550 | for (n = m->node; n; n = n->parent) | |
3551 | { | |
3552 | //rfapiDoRouteCallback(import_table, n, NULL); | |
3553 | } | |
3554 | rfapiMonitorItNodeChanged (import_table, m->node, NULL); | |
3555 | ||
3556 | referenced_vpn_prefix->info = m->next; | |
3557 | route_unlock_node (referenced_vpn_prefix); | |
3558 | XFREE (MTYPE_RFAPI_MONITOR_ENCAP, m); | |
3559 | } | |
3560 | ||
3561 | } | |
3562 | route_table_finish (referenced_vpn_table); | |
3563 | } | |
3564 | ||
3565 | RFAPI_CHECK_REFCOUNT (rn, SAFI_ENCAP, 0); | |
3566 | } | |
3567 | ||
3568 | static void | |
3569 | rfapiExpireVpnNow ( | |
3570 | struct rfapi_import_table *it, | |
3571 | struct route_node *rn, | |
3572 | struct bgp_info *bi, | |
3573 | int lockoffset) | |
3574 | { | |
3575 | struct rfapi_withdraw *wcb; | |
3576 | struct thread t; | |
3577 | ||
3578 | /* | |
3579 | * pretend we're an expiring timer | |
3580 | */ | |
3581 | wcb = XCALLOC (MTYPE_RFAPI_WITHDRAW, sizeof (struct rfapi_withdraw)); | |
3582 | wcb->info = bi; | |
3583 | wcb->node = rn; | |
3584 | wcb->import_table = it; | |
3585 | wcb->lockoffset = lockoffset; | |
3586 | memset (&t, 0, sizeof (t)); | |
3587 | t.arg = wcb; | |
3588 | rfapiWithdrawTimerVPN (&t); /* frees wcb */ | |
3589 | } | |
3590 | ||
3591 | ||
3592 | /* | |
3593 | * import a bgp_info if its route target list intersects with the | |
3594 | * import table's route target list | |
3595 | */ | |
3596 | void | |
3597 | rfapiBgpInfoFilteredImportVPN ( | |
3598 | struct rfapi_import_table *import_table, | |
3599 | int action, | |
3600 | struct peer *peer, | |
3601 | void *rfd, /* set for looped back routes */ | |
3602 | struct prefix *p, | |
3603 | struct prefix *aux_prefix, /* AFI_ETHER: optional IP */ | |
3604 | afi_t afi, | |
3605 | struct prefix_rd *prd, | |
3606 | struct attr *attr, /* part of bgp_info */ | |
3607 | u_char type, /* part of bgp_info */ | |
3608 | u_char sub_type, /* part of bgp_info */ | |
3609 | uint32_t *label) /* part of bgp_info */ | |
3610 | { | |
3611 | struct route_table *rt = NULL; | |
3612 | struct route_node *rn; | |
3613 | struct route_node *n; | |
3614 | struct bgp_info *info_new; | |
3615 | struct bgp_info *bi; | |
3616 | struct bgp_info *next; | |
3617 | char buf[BUFSIZ]; | |
3618 | struct prefix vn_prefix; | |
3619 | struct prefix un_prefix; | |
3620 | int un_prefix_valid = 0; | |
3621 | struct route_node *ern; | |
3622 | int replacing = 0; | |
3623 | int original_had_routes = 0; | |
3624 | struct prefix original_nexthop; | |
3625 | const char *action_str = NULL; | |
3626 | int is_it_ce = 0; | |
3627 | ||
3628 | struct bgp *bgp; | |
3629 | bgp = bgp_get_default (); /* assume 1 instance for now */ | |
3630 | ||
3631 | switch (action) | |
3632 | { | |
3633 | case FIF_ACTION_UPDATE: | |
3634 | action_str = "update"; | |
3635 | break; | |
3636 | case FIF_ACTION_WITHDRAW: | |
3637 | action_str = "withdraw"; | |
3638 | break; | |
3639 | case FIF_ACTION_KILL: | |
3640 | action_str = "kill"; | |
3641 | break; | |
3642 | default: | |
3643 | assert (0); | |
3644 | break; | |
3645 | } | |
3646 | ||
3647 | if (import_table == bgp->rfapi->it_ce) | |
3648 | is_it_ce = 1; | |
3649 | ||
a3b55c25 | 3650 | vnc_zlog_debug_verbose ("%s: entry: %s%s: prefix %s/%d: it %p, afi %s", __func__, |
65efcfce LB |
3651 | (is_it_ce ? "CE-IT " : ""), |
3652 | action_str, | |
3653 | rfapi_ntop (p->family, &p->u.prefix, buf, BUFSIZ), | |
3654 | p->prefixlen, import_table, afi2str (afi)); | |
3655 | ||
3656 | VNC_ITRCCK; | |
3657 | ||
3658 | /* | |
3659 | * Compare rt lists. If no intersection, don't import this route | |
3660 | * On a withdraw, peer and RD are sufficient to determine if | |
3661 | * we should act. | |
3662 | */ | |
3663 | if (action == FIF_ACTION_UPDATE) | |
3664 | { | |
3665 | if (!attr || !attr->extra || !attr->extra->ecommunity) | |
3666 | { | |
3667 | ||
a3b55c25 | 3668 | vnc_zlog_debug_verbose ("%s: attr, extra, or ecommunity missing, not importing", |
65efcfce LB |
3669 | __func__); |
3670 | return; | |
3671 | } | |
3672 | if ((import_table != bgp->rfapi->it_ce) && | |
3673 | !rfapiEcommunitiesIntersect (import_table->rt_import_list, | |
3674 | attr->extra->ecommunity)) | |
3675 | { | |
3676 | ||
a3b55c25 | 3677 | vnc_zlog_debug_verbose ("%s: it=%p: no ecommunity intersection", |
65efcfce LB |
3678 | __func__, import_table); |
3679 | return; | |
3680 | } | |
3681 | ||
3682 | memset (&vn_prefix, 0, sizeof (vn_prefix)); /* keep valgrind happy */ | |
3683 | if (rfapiGetNexthop (attr, &vn_prefix)) | |
3684 | { | |
3685 | /* missing nexthop address would be a bad, bad thing */ | |
a3b55c25 | 3686 | vnc_zlog_debug_verbose ("%s: missing nexthop", __func__); |
65efcfce LB |
3687 | return; |
3688 | } | |
3689 | } | |
3690 | ||
3691 | /* | |
3692 | * Figure out which radix tree the route would go into | |
3693 | */ | |
3694 | switch (afi) | |
3695 | { | |
3696 | case AFI_IP: | |
3697 | case AFI_IP6: | |
3698 | case AFI_ETHER: | |
3699 | rt = import_table->imported_vpn[afi]; | |
3700 | break; | |
3701 | ||
3702 | default: | |
3703 | zlog_err ("%s: bad afi %d", __func__, afi); | |
3704 | return; | |
3705 | } | |
3706 | ||
3707 | /* clear it */ | |
3708 | memset (&original_nexthop, 0, sizeof (original_nexthop)); | |
3709 | ||
3710 | /* | |
3711 | * route_node_lookup returns a node only if there is at least | |
3712 | * one route attached. | |
3713 | */ | |
3714 | rn = route_node_lookup (rt, p); | |
3715 | ||
a3b55c25 | 3716 | vnc_zlog_debug_verbose ("%s: rn=%p", __func__, rn); |
65efcfce LB |
3717 | |
3718 | if (rn) | |
3719 | { | |
3720 | ||
3721 | RFAPI_CHECK_REFCOUNT (rn, SAFI_MPLS_VPN, 1); | |
3722 | route_unlock_node (rn); /* undo lock in route_node_lookup */ | |
3723 | ||
3724 | if (rn->info) | |
3725 | original_had_routes = 1; | |
3726 | ||
3727 | /* | |
3728 | * Look for same route (will have same RD and peer) | |
3729 | */ | |
3730 | bi = rfapiItBiIndexSearch (rn, prd, peer, aux_prefix); | |
3731 | ||
3732 | if (bi) | |
3733 | { | |
3734 | ||
3735 | /* | |
3736 | * This was an old test when we iterated over the | |
3737 | * BIs linearly. Since we're now looking up with | |
3738 | * RD and peer, comparing types should not be | |
3739 | * needed. Changed to assertion. | |
3740 | * | |
3741 | * Compare types. Doing so prevents a RFP-originated | |
3742 | * route from matching an imported route, for example. | |
3743 | */ | |
3744 | assert (bi->type == type); | |
3745 | ||
a3b55c25 | 3746 | vnc_zlog_debug_verbose ("%s: found matching bi", __func__); |
65efcfce LB |
3747 | |
3748 | /* | |
3749 | * In the special CE table, withdrawals occur without holddown | |
3750 | */ | |
3751 | if (import_table == bgp->rfapi->it_ce) | |
3752 | { | |
3753 | vnc_direct_bgp_del_route_ce (bgp, rn, bi); | |
3754 | if (action == FIF_ACTION_WITHDRAW) | |
3755 | action = FIF_ACTION_KILL; | |
3756 | } | |
3757 | ||
3758 | if (action == FIF_ACTION_WITHDRAW) | |
3759 | { | |
3760 | ||
3761 | int washolddown = CHECK_FLAG (bi->flags, BGP_INFO_REMOVED); | |
3762 | ||
a3b55c25 | 3763 | vnc_zlog_debug_verbose ("%s: withdrawing at prefix %s/%d%s", |
65efcfce LB |
3764 | __func__, |
3765 | rfapi_ntop (rn->p.family, &rn->p.u.prefix, buf, | |
3766 | BUFSIZ), rn->p.prefixlen, | |
3767 | (washolddown ? " (already being withdrawn)" : "")); | |
3768 | ||
3769 | VNC_ITRCCK; | |
3770 | if (!washolddown) | |
3771 | { | |
3772 | rfapiBiStartWithdrawTimer (import_table, rn, bi, | |
3773 | afi, SAFI_MPLS_VPN, | |
3774 | rfapiWithdrawTimerVPN); | |
3775 | ||
3776 | RFAPI_UPDATE_ITABLE_COUNT (bi, import_table, afi, -1); | |
3777 | import_table->holddown_count[afi] += 1; | |
3778 | } | |
3779 | VNC_ITRCCK; | |
3780 | } | |
3781 | else | |
3782 | { | |
a3b55c25 | 3783 | vnc_zlog_debug_verbose ("%s: %s at prefix %s/%d", |
65efcfce LB |
3784 | __func__, |
3785 | ((action == | |
3786 | FIF_ACTION_KILL) ? "killing" : "replacing"), | |
3787 | rfapi_ntop (rn->p.family, &rn->p.u.prefix, buf, | |
3788 | BUFSIZ), rn->p.prefixlen); | |
3789 | ||
3790 | /* | |
3791 | * If this route is waiting to be deleted because of | |
3792 | * a previous withdraw, we must cancel its timer. | |
3793 | */ | |
3794 | if (CHECK_FLAG (bi->flags, BGP_INFO_REMOVED) && | |
3795 | bi->extra->vnc.import.timer) | |
3796 | { | |
3797 | ||
3798 | struct thread *t = | |
3799 | (struct thread *) bi->extra->vnc.import.timer; | |
3800 | struct rfapi_withdraw *wcb = t->arg; | |
3801 | ||
3802 | XFREE (MTYPE_RFAPI_WITHDRAW, wcb); | |
3803 | thread_cancel (t); | |
3804 | ||
3805 | import_table->holddown_count[afi] -= 1; | |
3806 | RFAPI_UPDATE_ITABLE_COUNT (bi, import_table, afi, 1); | |
65efcfce LB |
3807 | } |
3808 | /* | |
3809 | * decrement remote count (if route is remote) because | |
3810 | * we are going to remove it below | |
3811 | */ | |
3812 | RFAPI_UPDATE_ITABLE_COUNT (bi, import_table, afi, -1); | |
3813 | if (action == FIF_ACTION_UPDATE) | |
3814 | { | |
3815 | replacing = 1; | |
3816 | ||
3817 | /* | |
3818 | * make copy of original nexthop so we can see if it changed | |
3819 | */ | |
3820 | rfapiGetNexthop (bi->attr, &original_nexthop); | |
3821 | ||
3822 | /* | |
3823 | * remove bi without doing any export processing | |
3824 | */ | |
3825 | if (CHECK_FLAG (bi->flags, BGP_INFO_VALID) | |
3826 | && VALID_INTERIOR_TYPE (bi->type)) | |
3827 | RFAPI_MONITOR_EXTERIOR (rn)->valid_interior_count--; | |
3828 | rfapiItBiIndexDel (rn, bi); | |
3829 | rfapiBgpInfoDetach (rn, bi); | |
3830 | rfapiMonitorEncapDelete (bi); | |
3831 | vnc_import_bgp_exterior_del_route_interior (bgp, | |
3832 | import_table, | |
3833 | rn, bi); | |
3834 | rfapiBgpInfoFree (bi); | |
3835 | } | |
3836 | else | |
3837 | { | |
3838 | /* Kill */ | |
3839 | /* | |
3840 | * remove bi and do export processing | |
3841 | */ | |
3842 | import_table->holddown_count[afi] += 1; | |
3843 | rfapiExpireVpnNow (import_table, rn, bi, 0); | |
3844 | } | |
3845 | ||
3846 | } | |
3847 | } | |
3848 | ||
3849 | } | |
3850 | ||
3851 | if (rn) | |
3852 | RFAPI_CHECK_REFCOUNT (rn, SAFI_MPLS_VPN, replacing ? 1 : 0); | |
3853 | ||
3854 | if (action == FIF_ACTION_WITHDRAW || action == FIF_ACTION_KILL) | |
3855 | { | |
3856 | VNC_ITRCCK; | |
3857 | return; | |
3858 | } | |
3859 | ||
3860 | info_new = rfapiBgpInfoCreate (attr, peer, rfd, prd, type, sub_type, label); | |
3861 | ||
3862 | /* | |
3863 | * lookup un address in encap table | |
3864 | */ | |
3865 | ern = route_node_match (import_table->imported_encap[afi], &vn_prefix); | |
3866 | if (ern) | |
3867 | { | |
3868 | rfapiCopyUnEncap2VPN (ern->info, info_new); | |
3869 | route_unlock_node (ern); /* undo lock in route_note_match */ | |
3870 | } | |
3871 | else | |
3872 | { | |
3873 | char buf[BUFSIZ]; | |
3874 | prefix2str (&vn_prefix, buf, sizeof (buf)); | |
3875 | buf[BUFSIZ - 1] = 0; | |
3876 | /* Not a big deal, just means VPN route got here first */ | |
a3b55c25 | 3877 | vnc_zlog_debug_verbose ("%s: no encap route for vn addr %s", __func__, buf); |
65efcfce LB |
3878 | info_new->extra->vnc.import.un_family = 0; |
3879 | } | |
3880 | ||
3881 | if (rn) | |
3882 | { | |
3883 | if (!replacing) | |
3884 | route_lock_node (rn); | |
3885 | } | |
3886 | else | |
3887 | { | |
3888 | /* | |
3889 | * No need to increment reference count, so only "get" | |
3890 | * if the node is not there already | |
3891 | */ | |
3892 | rn = route_node_get (rt, p); | |
3893 | } | |
3894 | ||
3895 | /* | |
3896 | * For ethernet routes, if there is an accompanying IP address, | |
3897 | * save it in the bi | |
3898 | */ | |
3899 | if ((AFI_ETHER == afi) && aux_prefix) | |
3900 | { | |
3901 | ||
a3b55c25 | 3902 | vnc_zlog_debug_verbose ("%s: setting BI's aux_prefix", __func__); |
65efcfce LB |
3903 | info_new->extra->vnc.import.aux_prefix = *aux_prefix; |
3904 | } | |
3905 | ||
a3b55c25 | 3906 | vnc_zlog_debug_verbose ("%s: inserting bi %p at prefix %s/%d #%d", |
65efcfce LB |
3907 | __func__, |
3908 | info_new, | |
3909 | rfapi_ntop (rn->p.family, &rn->p.u.prefix, buf, BUFSIZ), | |
3910 | rn->p.prefixlen, rn->lock); | |
3911 | ||
3912 | rfapiBgpInfoAttachSorted (rn, info_new, afi, SAFI_MPLS_VPN); | |
3913 | rfapiItBiIndexAdd (rn, info_new); | |
3914 | if (!rfapiGetUnAddrOfVpnBi (info_new, NULL)) | |
3915 | { | |
3916 | if (VALID_INTERIOR_TYPE (info_new->type)) | |
3917 | RFAPI_MONITOR_EXTERIOR (rn)->valid_interior_count++; | |
3918 | SET_FLAG (info_new->flags, BGP_INFO_VALID); | |
3919 | } | |
3920 | RFAPI_UPDATE_ITABLE_COUNT (info_new, import_table, afi, 1); | |
3921 | vnc_import_bgp_exterior_add_route_interior (bgp, import_table, rn, | |
3922 | info_new); | |
3923 | ||
3924 | if (import_table == bgp->rfapi->it_ce) | |
3925 | vnc_direct_bgp_add_route_ce (bgp, rn, info_new); | |
3926 | ||
a3b55c25 LB |
3927 | if (VNC_DEBUG(VERBOSE)) |
3928 | { | |
3929 | vnc_zlog_debug_verbose ("%s: showing IT node", __func__); | |
3930 | rfapiShowItNode (NULL, rn); /* debug */ | |
3931 | } | |
65efcfce LB |
3932 | |
3933 | rfapiMonitorEncapAdd (import_table, &vn_prefix, rn, info_new); | |
3934 | ||
3935 | if (!rfapiGetUnAddrOfVpnBi (info_new, &un_prefix)) | |
3936 | { | |
3937 | ||
3938 | /* | |
3939 | * if we have a valid UN address (either via Encap route | |
3940 | * or via tunnel attribute), then we should attempt | |
3941 | * to move any monitors at less-specific nodes to this node | |
3942 | */ | |
3943 | rfapiMonitorMoveLonger (rn); | |
3944 | ||
3945 | un_prefix_valid = 1; | |
3946 | ||
3947 | } | |
3948 | ||
3949 | /* | |
3950 | * 101129 Enhancement: if we add a route (implication: it is not | |
3951 | * in holddown), delete all other routes from this nve at this | |
3952 | * node that are in holddown, regardless of peer. | |
3953 | * | |
3954 | * Reasons it's OK to do that: | |
3955 | * | |
3956 | * - if the holddown route being deleted originally came from BGP VPN, | |
3957 | * it is already gone from BGP (implication of holddown), so there | |
3958 | * won't be any added inconsistency with the BGP RIB. | |
3959 | * | |
3960 | * - once a fresh route is added at a prefix, any routes in holddown | |
3961 | * at that prefix will not show up in RFP responses, so deleting | |
3962 | * the holddown routes won't affect the contents of responses. | |
3963 | * | |
3964 | * - lifetimes are supposed to be consistent, so there should not | |
3965 | * be a case where the fresh route has a shorter lifetime than | |
3966 | * the holddown route, so we don't expect the fresh route to | |
3967 | * disappear and complete its holddown time before the existing | |
3968 | * holddown routes time out. Therefore, we won't have a situation | |
3969 | * where we expect the existing holddown routes to be hidden and | |
3970 | * then to reappear sometime later (as holddown routes) in a | |
3971 | * RFP response. | |
3972 | * | |
3973 | * Among other things, this would enable us to skirt the problem | |
3974 | * of local holddown routes that refer to NVE descriptors that | |
3975 | * have already been closed (if the same NVE triggers a subsequent | |
3976 | * rfapi_open(), the new peer is different and doesn't match the | |
3977 | * peer of the holddown route, so the stale holddown route still | |
3978 | * hangs around until it times out instead of just being replaced | |
3979 | * by the fresh route). | |
3980 | */ | |
3981 | /* | |
3982 | * We know that the new bi will have been inserted before any routes | |
3983 | * in holddown, so we can skip any that came before it | |
3984 | */ | |
3985 | for (bi = info_new->next; bi; bi = next) | |
3986 | { | |
3987 | ||
3988 | struct prefix pfx_vn; | |
3989 | struct prefix pfx_un; | |
3990 | int un_match = 0; | |
3991 | int remote_peer_match = 0; | |
3992 | ||
3993 | next = bi->next; | |
3994 | ||
3995 | /* | |
3996 | * Must be holddown | |
3997 | */ | |
3998 | if (!CHECK_FLAG (bi->flags, BGP_INFO_REMOVED)) | |
3999 | continue; | |
4000 | ||
4001 | /* | |
4002 | * Must match VN address (nexthop of VPN route) | |
4003 | */ | |
4004 | if (rfapiGetNexthop (bi->attr, &pfx_vn)) | |
4005 | continue; | |
4006 | if (!prefix_same (&pfx_vn, &vn_prefix)) | |
4007 | continue; | |
4008 | ||
4009 | if (un_prefix_valid && /* new route UN addr */ | |
4010 | !rfapiGetUnAddrOfVpnBi (bi, &pfx_un) && /* old route UN addr */ | |
4011 | prefix_same (&pfx_un, &un_prefix)) | |
4012 | { /* compare */ | |
4013 | un_match = 1; | |
4014 | } | |
4015 | if (!RFAPI_LOCAL_BI (bi) && !RFAPI_LOCAL_BI (info_new) && | |
4016 | sockunion_same (&bi->peer->su, &info_new->peer->su)) | |
4017 | { | |
4018 | /* old & new are both remote, same peer */ | |
4019 | remote_peer_match = 1; | |
4020 | } | |
4021 | ||
4022 | if (!un_match & !remote_peer_match) | |
4023 | continue; | |
4024 | ||
a3b55c25 | 4025 | vnc_zlog_debug_verbose ("%s: removing holddown bi matching NVE of new route", |
65efcfce LB |
4026 | __func__); |
4027 | if (bi->extra->vnc.import.timer) | |
4028 | { | |
4029 | struct thread *t = (struct thread *) bi->extra->vnc.import.timer; | |
4030 | struct rfapi_withdraw *wcb = t->arg; | |
4031 | ||
4032 | XFREE (MTYPE_RFAPI_WITHDRAW, wcb); | |
4033 | thread_cancel (t); | |
4034 | } | |
4035 | rfapiExpireVpnNow (import_table, rn, bi, 0); | |
4036 | } | |
4037 | ||
4038 | if (!original_had_routes) | |
4039 | { | |
4040 | /* | |
4041 | * We went from 0 usable routes to 1 usable route. Perform the | |
4042 | * "Adding a Route" export process. | |
4043 | */ | |
4044 | vnc_direct_bgp_add_prefix (bgp, import_table, rn); | |
4045 | vnc_zebra_add_prefix (bgp, import_table, rn); | |
4046 | } | |
4047 | else | |
4048 | { | |
4049 | /* | |
4050 | * Check for nexthop change event | |
4051 | * Note: the prefix_same() test below detects two situations: | |
4052 | * 1. route is replaced, new route has different nexthop | |
4053 | * 2. new route is added (original_nexthop is 0) | |
4054 | */ | |
4055 | struct prefix new_nexthop; | |
4056 | ||
4057 | rfapiGetNexthop (attr, &new_nexthop); | |
4058 | if (!prefix_same (&original_nexthop, &new_nexthop)) | |
4059 | { | |
4060 | /* | |
4061 | * nexthop change event | |
4062 | * vnc_direct_bgp_add_prefix() will recompute VN addr ecommunity | |
4063 | */ | |
4064 | vnc_direct_bgp_add_prefix (bgp, import_table, rn); | |
4065 | } | |
4066 | } | |
4067 | ||
4068 | if (!(bgp->rfapi_cfg->flags & BGP_VNC_CONFIG_CALLBACK_DISABLE)) | |
4069 | { | |
4070 | for (n = rn; n; n = n->parent) | |
4071 | { | |
4072 | //rfapiDoRouteCallback(import_table, n, NULL); | |
4073 | } | |
4074 | rfapiMonitorItNodeChanged (import_table, rn, NULL); | |
4075 | } | |
4076 | RFAPI_CHECK_REFCOUNT (rn, SAFI_MPLS_VPN, 0); | |
4077 | VNC_ITRCCK; | |
4078 | } | |
4079 | ||
4080 | static rfapi_bi_filtered_import_f * | |
4081 | rfapiBgpInfoFilteredImportFunction (safi_t safi) | |
4082 | { | |
4083 | switch (safi) | |
4084 | { | |
4085 | case SAFI_MPLS_VPN: | |
65efcfce LB |
4086 | return rfapiBgpInfoFilteredImportVPN; |
4087 | ||
4088 | case SAFI_ENCAP: | |
4089 | return rfapiBgpInfoFilteredImportEncap; | |
4090 | } | |
4091 | zlog_err ("%s: bad safi %d", __func__, safi); | |
4092 | return NULL; | |
4093 | } | |
4094 | ||
4095 | void | |
4096 | rfapiProcessUpdate ( | |
4097 | struct peer *peer, | |
4098 | void *rfd, /* set when looped from RFP/RFAPI */ | |
4099 | struct prefix *p, | |
4100 | struct prefix_rd *prd, | |
4101 | struct attr *attr, | |
4102 | afi_t afi, | |
4103 | safi_t safi, | |
4104 | u_char type, | |
4105 | u_char sub_type, | |
4106 | uint32_t *label) | |
4107 | { | |
4108 | struct bgp *bgp; | |
4109 | struct rfapi *h; | |
4110 | struct rfapi_import_table *it; | |
4111 | int has_ip_route = 1; | |
4112 | uint32_t lni = 0; | |
4113 | ||
4114 | bgp = bgp_get_default (); /* assume 1 instance for now */ | |
4115 | assert (bgp); | |
4116 | ||
4117 | h = bgp->rfapi; | |
4118 | assert (h); | |
4119 | ||
4120 | /* | |
4121 | * look at high-order byte of RD. FF means MAC | |
4122 | * address is present (VNC L2VPN) | |
4123 | */ | |
4124 | if ((safi == SAFI_MPLS_VPN) && | |
4125 | (decode_rd_type(prd->val) == RD_TYPE_VNC_ETH)) | |
4126 | { | |
4127 | struct prefix pfx_mac_buf; | |
4128 | struct prefix pfx_nexthop_buf; | |
4129 | int rc; | |
4130 | ||
4131 | /* | |
4132 | * Set flag if prefix and nexthop are the same - don't | |
4133 | * add the route to normal IP-based import tables | |
4134 | */ | |
4135 | if (!rfapiGetNexthop (attr, &pfx_nexthop_buf)) | |
4136 | { | |
4137 | if (!prefix_cmp (&pfx_nexthop_buf, p)) | |
4138 | { | |
4139 | has_ip_route = 0; | |
4140 | } | |
4141 | } | |
4142 | ||
4143 | memset (&pfx_mac_buf, 0, sizeof (pfx_mac_buf)); | |
4144 | pfx_mac_buf.family = AF_ETHERNET; | |
4145 | pfx_mac_buf.prefixlen = 48; | |
4146 | memcpy (&pfx_mac_buf.u.prefix_eth.octet, prd->val + 2, 6); | |
4147 | ||
4148 | /* | |
4149 | * Find rt containing LNI (Logical Network ID), which | |
4150 | * _should_ always be present when mac address is present | |
4151 | */ | |
4152 | rc = rfapiEcommunityGetLNI (attr->extra->ecommunity, &lni); | |
4153 | ||
a3b55c25 | 4154 | vnc_zlog_debug_verbose |
65efcfce LB |
4155 | ("%s: rfapiEcommunityGetLNI returned %d, lni=%d, attr=%p, attr->extra=%p", |
4156 | __func__, rc, lni, attr, attr->extra); | |
4157 | if (attr && attr->extra && !rc) | |
4158 | { | |
4159 | it = rfapiMacImportTableGet (bgp, lni); | |
4160 | ||
4161 | rfapiBgpInfoFilteredImportVPN ( | |
4162 | it, | |
4163 | FIF_ACTION_UPDATE, | |
4164 | peer, | |
4165 | rfd, | |
4166 | &pfx_mac_buf, /* prefix */ | |
4167 | p, /* aux prefix: IP addr */ | |
4168 | AFI_ETHER, | |
4169 | prd, | |
4170 | attr, | |
4171 | type, | |
4172 | sub_type, | |
4173 | label); | |
4174 | } | |
4175 | ||
4176 | } | |
4177 | ||
4178 | if (!has_ip_route) | |
4179 | return; | |
4180 | ||
4181 | /* | |
4182 | * Iterate over all import tables; do a filtered import | |
4183 | * for the afi/safi combination | |
4184 | */ | |
4185 | for (it = h->imports; it; it = it->next) | |
4186 | { | |
4187 | (*rfapiBgpInfoFilteredImportFunction (safi)) ( | |
4188 | it, | |
4189 | FIF_ACTION_UPDATE, | |
4190 | peer, | |
4191 | rfd, | |
4192 | p, /* prefix */ | |
4193 | NULL, | |
4194 | afi, | |
4195 | prd, | |
4196 | attr, | |
4197 | type, | |
4198 | sub_type, | |
4199 | label); | |
4200 | } | |
4201 | ||
9cabb64b | 4202 | if (safi == SAFI_MPLS_VPN) |
65efcfce LB |
4203 | { |
4204 | vnc_direct_bgp_rh_add_route (bgp, afi, p, peer, attr); | |
4205 | } | |
4206 | ||
4207 | if (safi == SAFI_MPLS_VPN) | |
4208 | { | |
4209 | rfapiBgpInfoFilteredImportVPN ( | |
4210 | bgp->rfapi->it_ce, | |
4211 | FIF_ACTION_UPDATE, | |
4212 | peer, | |
4213 | rfd, | |
4214 | p, /* prefix */ | |
4215 | NULL, | |
4216 | afi, | |
4217 | prd, | |
4218 | attr, | |
4219 | type, | |
4220 | sub_type, | |
4221 | label); | |
4222 | } | |
4223 | } | |
4224 | ||
4225 | ||
4226 | void | |
4227 | rfapiProcessWithdraw ( | |
4228 | struct peer *peer, | |
4229 | void *rfd, | |
4230 | struct prefix *p, | |
4231 | struct prefix_rd *prd, | |
4232 | struct attr *attr, | |
4233 | afi_t afi, | |
4234 | safi_t safi, | |
4235 | u_char type, | |
4236 | int kill) | |
4237 | { | |
4238 | struct bgp *bgp; | |
4239 | struct rfapi *h; | |
4240 | struct rfapi_import_table *it; | |
4241 | ||
4242 | bgp = bgp_get_default (); /* assume 1 instance for now */ | |
4243 | assert (bgp); | |
4244 | ||
4245 | h = bgp->rfapi; | |
4246 | assert (h); | |
4247 | ||
4248 | /* | |
4249 | * look at high-order byte of RD. FF means MAC | |
4250 | * address is present (VNC L2VPN) | |
4251 | */ | |
4252 | if (h->import_mac != NULL && safi == SAFI_MPLS_VPN && | |
4253 | decode_rd_type(prd->val) == RD_TYPE_VNC_ETH) | |
4254 | { | |
4255 | struct prefix pfx_mac_buf; | |
4256 | void *cursor = NULL; | |
4257 | int rc; | |
4258 | ||
4259 | memset (&pfx_mac_buf, 0, sizeof (pfx_mac_buf)); | |
4260 | pfx_mac_buf.family = AF_ETHERNET; | |
4261 | pfx_mac_buf.prefixlen = 48; | |
4262 | memcpy (&pfx_mac_buf.u.prefix_eth, prd->val + 2, 6); | |
4263 | ||
4264 | /* | |
4265 | * withdraw does not contain attrs, so we don't have | |
4266 | * access to the route's LNI, which would ordinarily | |
4267 | * select the specific mac-based import table. Instead, | |
4268 | * we must iterate over all mac-based tables and rely | |
4269 | * on the RD to match. | |
4270 | * | |
4271 | * If this approach is too slow, add an index where | |
4272 | * key is {RD, peer} and value is the import table | |
4273 | */ | |
4274 | for (rc = skiplist_next (h->import_mac, NULL, (void **) &it, &cursor); | |
4275 | rc == 0; | |
4276 | rc = skiplist_next (h->import_mac, NULL, (void **) &it, &cursor)) | |
4277 | { | |
4278 | ||
4279 | #if DEBUG_L2_EXTRA | |
a3b55c25 | 4280 | vnc_zlog_debug_verbose |
65efcfce LB |
4281 | ("%s: calling rfapiBgpInfoFilteredImportVPN(it=%p, afi=AFI_ETHER)", |
4282 | __func__, it); | |
4283 | #endif | |
4284 | ||
4285 | rfapiBgpInfoFilteredImportVPN ( | |
4286 | it, | |
4287 | (kill ? FIF_ACTION_KILL : FIF_ACTION_WITHDRAW), | |
4288 | peer, | |
4289 | rfd, | |
4290 | &pfx_mac_buf, /* prefix */ | |
4291 | p, /* aux_prefix: IP */ | |
4292 | AFI_ETHER, | |
4293 | prd, | |
4294 | attr, | |
4295 | type, | |
4296 | 0, | |
4297 | NULL); /* sub_type & label unused for withdraw */ | |
4298 | } | |
4299 | } | |
4300 | ||
4301 | /* | |
4302 | * XXX For the case where the withdraw involves an L2 | |
4303 | * route with no IP information, we rely on the lack | |
4304 | * of RT-list intersection to filter out the withdraw | |
4305 | * from the IP-based import tables below | |
4306 | */ | |
4307 | ||
4308 | /* | |
4309 | * Iterate over all import tables; do a filtered import | |
4310 | * for the afi/safi combination | |
4311 | */ | |
4312 | ||
4313 | for (it = h->imports; it; it = it->next) | |
4314 | { | |
4315 | (*rfapiBgpInfoFilteredImportFunction (safi)) ( | |
4316 | it, | |
4317 | (kill ? FIF_ACTION_KILL : FIF_ACTION_WITHDRAW), | |
4318 | peer, | |
4319 | rfd, | |
4320 | p, /* prefix */ | |
4321 | NULL, | |
4322 | afi, | |
4323 | prd, | |
4324 | attr, | |
4325 | type, | |
4326 | 0, | |
4327 | NULL); /* sub_type & label unused for withdraw */ | |
4328 | } | |
4329 | ||
4330 | /* TBD the deletion should happen after the lifetime expires */ | |
9cabb64b | 4331 | if (safi == SAFI_MPLS_VPN) |
65efcfce LB |
4332 | vnc_direct_bgp_rh_del_route (bgp, afi, p, peer); |
4333 | ||
4334 | if (safi == SAFI_MPLS_VPN) | |
4335 | { | |
4336 | rfapiBgpInfoFilteredImportVPN ( | |
4337 | bgp->rfapi->it_ce, | |
4338 | (kill ? FIF_ACTION_KILL : FIF_ACTION_WITHDRAW), | |
4339 | peer, | |
4340 | rfd, | |
4341 | p, /* prefix */ | |
4342 | NULL, | |
4343 | afi, | |
4344 | prd, | |
4345 | attr, | |
4346 | type, | |
4347 | 0, | |
4348 | NULL); /* sub_type & label unused for withdraw */ | |
4349 | } | |
4350 | } | |
4351 | ||
4352 | /* | |
4353 | * TBD optimized withdraw timer algorithm for case of many | |
4354 | * routes expiring at the same time due to peer drop. | |
4355 | */ | |
4356 | /* | |
4357 | * 1. Visit all BIs in all ENCAP import tables. | |
4358 | * | |
4359 | * a. If a bi's peer is the failed peer, remove the bi. | |
4360 | * b. If the removed ENCAP bi was first in the list of | |
4361 | * BIs at this ENCAP node, loop over all monitors | |
4362 | * at this node: | |
4363 | * | |
4364 | * (1) for each ENCAP monitor, loop over all its | |
4365 | * VPN node monitors and set their RFAPI_MON_FLAG_NEEDCALLBACK | |
4366 | * flags. | |
4367 | * | |
4368 | * 2. Visit all BIs in all VPN import tables. | |
4369 | * a. If a bi's peer is the failed peer, remove the bi. | |
4370 | * b. loop over all the VPN node monitors and set their | |
4371 | * RFAPI_MON_FLAG_NEEDCALLBACK flags | |
4372 | * c. If there are no BIs left at this VPN node, | |
4373 | * | |
4374 | */ | |
4375 | ||
4376 | ||
4377 | /* surprise, this gets called from peer_delete(), from rfapi_close() */ | |
4378 | static void | |
4379 | rfapiProcessPeerDownRt ( | |
4380 | struct peer *peer, | |
4381 | struct rfapi_import_table *import_table, | |
4382 | afi_t afi, | |
4383 | safi_t safi) | |
4384 | { | |
4385 | struct route_node *rn; | |
4386 | struct bgp_info *bi; | |
4387 | struct route_table *rt; | |
4388 | int (*timer_service_func) (struct thread *); | |
4389 | ||
4390 | assert (afi == AFI_IP || afi == AFI_IP6); | |
4391 | ||
4392 | VNC_ITRCCK; | |
4393 | ||
4394 | switch (safi) | |
4395 | { | |
4396 | case SAFI_MPLS_VPN: | |
4397 | rt = import_table->imported_vpn[afi]; | |
4398 | timer_service_func = rfapiWithdrawTimerVPN; | |
4399 | break; | |
4400 | case SAFI_ENCAP: | |
4401 | rt = import_table->imported_encap[afi]; | |
4402 | timer_service_func = rfapiWithdrawTimerEncap; | |
4403 | break; | |
4404 | default: | |
4405 | assert (0); | |
4406 | } | |
4407 | ||
4408 | ||
4409 | for (rn = route_top (rt); rn; rn = route_next (rn)) | |
4410 | { | |
4411 | for (bi = rn->info; bi; bi = bi->next) | |
4412 | { | |
4413 | if (bi->peer == peer) | |
4414 | { | |
4415 | ||
4416 | if (CHECK_FLAG (bi->flags, BGP_INFO_REMOVED)) | |
4417 | { | |
4418 | /* already in holddown, skip */ | |
4419 | continue; | |
4420 | } | |
4421 | ||
4422 | if (safi == SAFI_MPLS_VPN) | |
4423 | { | |
4424 | RFAPI_UPDATE_ITABLE_COUNT (bi, import_table, afi, -1); | |
4425 | import_table->holddown_count[afi] += 1; | |
4426 | } | |
4427 | rfapiBiStartWithdrawTimer (import_table, rn, bi, | |
4428 | afi, safi, | |
4429 | timer_service_func); | |
4430 | } | |
4431 | } | |
4432 | } | |
4433 | VNC_ITRCCK; | |
4434 | } | |
4435 | ||
4436 | /* | |
4437 | * This gets called when a peer connection drops. We have to remove | |
4438 | * all the routes from this peer. | |
4439 | * | |
4440 | * Current approach is crude. TBD Optimize by setting fewer timers and | |
4441 | * grouping withdrawn routes so we can generate callbacks more | |
4442 | * efficiently. | |
4443 | */ | |
4444 | void | |
4445 | rfapiProcessPeerDown (struct peer *peer) | |
4446 | { | |
4447 | struct bgp *bgp; | |
4448 | struct rfapi *h; | |
4449 | struct rfapi_import_table *it; | |
4450 | ||
4451 | /* | |
4452 | * If this peer is a "dummy" peer structure atached to a RFAPI | |
4453 | * nve_descriptor, we don't need to walk the import tables | |
4454 | * because the routes are already withdrawn by rfapi_close() | |
4455 | */ | |
4456 | if (CHECK_FLAG (peer->flags, PEER_FLAG_IS_RFAPI_HD)) | |
4457 | return; | |
4458 | ||
4459 | /* | |
4460 | * 1. Visit all BIs in all ENCAP import tables. | |
4461 | * Start withdraw timer on the BIs that match peer. | |
4462 | * | |
4463 | * 2. Visit All BIs in all VPN import tables. | |
4464 | * Start withdraw timer on the BIs that match peer. | |
4465 | */ | |
4466 | ||
4467 | bgp = bgp_get_default (); /* assume 1 instance for now */ | |
34620e24 DS |
4468 | if (!bgp) |
4469 | return; | |
65efcfce LB |
4470 | |
4471 | h = bgp->rfapi; | |
4472 | assert (h); | |
4473 | ||
4474 | for (it = h->imports; it; it = it->next) | |
4475 | { | |
4476 | rfapiProcessPeerDownRt (peer, it, AFI_IP, SAFI_ENCAP); | |
4477 | rfapiProcessPeerDownRt (peer, it, AFI_IP6, SAFI_ENCAP); | |
4478 | rfapiProcessPeerDownRt (peer, it, AFI_IP, SAFI_MPLS_VPN); | |
4479 | rfapiProcessPeerDownRt (peer, it, AFI_IP6, SAFI_MPLS_VPN); | |
4480 | } | |
4481 | ||
4482 | if (h->it_ce) | |
4483 | { | |
4484 | rfapiProcessPeerDownRt (peer, h->it_ce, AFI_IP, SAFI_MPLS_VPN); | |
4485 | rfapiProcessPeerDownRt (peer, h->it_ce, AFI_IP6, SAFI_MPLS_VPN); | |
4486 | } | |
4487 | } | |
4488 | ||
4489 | /* | |
4490 | * Import an entire RIB (for an afi/safi) to an import table RIB, | |
4491 | * filtered according to the import table's RT list | |
4492 | * | |
4493 | * TBD: does this function need additions to match rfapiProcessUpdate() | |
4494 | * for, e.g., L2 handling? | |
4495 | */ | |
4496 | static void | |
4497 | rfapiBgpTableFilteredImport ( | |
4498 | struct bgp *bgp, | |
4499 | struct rfapi_import_table *it, | |
4500 | afi_t afi, | |
4501 | safi_t safi) | |
4502 | { | |
4503 | struct bgp_node *rn1; | |
4504 | struct bgp_node *rn2; | |
4505 | ||
4506 | /* Only these SAFIs have 2-level RIBS */ | |
4507 | assert (safi == SAFI_MPLS_VPN || safi == SAFI_ENCAP); | |
4508 | ||
4509 | /* | |
4510 | * Now visit all the rd nodes and the nodes of all the | |
4511 | * route tables attached to them, and import the routes | |
4512 | * if they have matching route targets | |
4513 | */ | |
4514 | for (rn1 = bgp_table_top (bgp->rib[afi][safi]); | |
4515 | rn1; rn1 = bgp_route_next (rn1)) | |
4516 | { | |
4517 | ||
4518 | if (rn1->info) | |
4519 | { | |
4520 | for (rn2 = bgp_table_top (rn1->info); | |
4521 | rn2; rn2 = bgp_route_next (rn2)) | |
4522 | { | |
4523 | ||
4524 | struct bgp_info *bi; | |
4525 | ||
4526 | for (bi = rn2->info; bi; bi = bi->next) | |
4527 | { | |
4528 | u_int32_t label = 0; | |
4529 | ||
4530 | if (CHECK_FLAG (bi->flags, BGP_INFO_REMOVED)) | |
4531 | continue; | |
4532 | ||
4533 | if (bi->extra) | |
4534 | label = decode_label (bi->extra->tag); | |
4535 | (*rfapiBgpInfoFilteredImportFunction (safi)) ( | |
4536 | it, /* which import table */ | |
4537 | FIF_ACTION_UPDATE, | |
4538 | bi->peer, | |
4539 | NULL, | |
4540 | &rn2->p, /* prefix */ | |
4541 | NULL, | |
4542 | afi, | |
4543 | (struct prefix_rd *) &rn1->p, | |
4544 | bi->attr, | |
4545 | bi->type, | |
4546 | bi->sub_type, | |
4547 | &label); | |
4548 | } | |
4549 | } | |
4550 | } | |
4551 | } | |
4552 | } | |
4553 | ||
4554 | ||
4555 | /* per-bgp-instance rfapi data */ | |
4556 | struct rfapi * | |
4557 | bgp_rfapi_new (struct bgp *bgp) | |
4558 | { | |
4559 | struct rfapi *h; | |
4560 | int afi; | |
4561 | struct rfapi_rfp_cfg *cfg = NULL; | |
4562 | struct rfapi_rfp_cb_methods *cbm = NULL; | |
4563 | ||
4564 | assert (bgp->rfapi_cfg == NULL); | |
4565 | ||
4566 | h = (struct rfapi *) XCALLOC (MTYPE_RFAPI, sizeof (struct rfapi)); | |
4567 | ||
4568 | for (afi = AFI_IP; afi < AFI_MAX; afi++) | |
4569 | { | |
4570 | /* ugly, to deal with addition of delegates, part of 0.99.24.1 merge */ | |
4571 | h->un[afi].delegate = route_table_get_default_delegate (); | |
4572 | } | |
4573 | ||
4574 | /* | |
4575 | * initialize the ce import table | |
4576 | */ | |
4577 | h->it_ce = | |
4578 | XCALLOC (MTYPE_RFAPI_IMPORTTABLE, sizeof (struct rfapi_import_table)); | |
4579 | h->it_ce->imported_vpn[AFI_IP] = route_table_init (); | |
4580 | h->it_ce->imported_vpn[AFI_IP6] = route_table_init (); | |
4581 | h->it_ce->imported_encap[AFI_IP] = route_table_init (); | |
4582 | h->it_ce->imported_encap[AFI_IP6] = route_table_init (); | |
4583 | rfapiBgpTableFilteredImport (bgp, h->it_ce, AFI_IP, SAFI_MPLS_VPN); | |
4584 | rfapiBgpTableFilteredImport (bgp, h->it_ce, AFI_IP6, SAFI_MPLS_VPN); | |
4585 | ||
4586 | /* | |
4587 | * Set up work queue for deferred rfapi_close operations | |
4588 | */ | |
4589 | h->deferred_close_q = work_queue_new (bm->master, "rfapi deferred close"); | |
4590 | h->deferred_close_q->spec.workfunc = rfapi_deferred_close_workfunc; | |
4591 | h->deferred_close_q->spec.data = h; | |
4592 | ||
4593 | h->rfp = rfp_start (bm->master, &cfg, &cbm); | |
4594 | bgp->rfapi_cfg = bgp_rfapi_cfg_new (cfg); | |
4595 | if (cbm != NULL) | |
4596 | { | |
4597 | h->rfp_methods = *cbm; | |
4598 | } | |
4599 | return h; | |
4600 | } | |
4601 | ||
4602 | void | |
4603 | bgp_rfapi_destroy (struct bgp *bgp, struct rfapi *h) | |
4604 | { | |
4605 | if (bgp == NULL || h == NULL) | |
4606 | return; | |
4607 | ||
4608 | if (h->resolve_nve_nexthop) | |
4609 | { | |
4610 | skiplist_free (h->resolve_nve_nexthop); | |
4611 | h->resolve_nve_nexthop = NULL; | |
4612 | } | |
4613 | ||
4614 | route_table_finish (h->it_ce->imported_vpn[AFI_IP]); | |
4615 | route_table_finish (h->it_ce->imported_vpn[AFI_IP6]); | |
4616 | route_table_finish (h->it_ce->imported_encap[AFI_IP]); | |
4617 | route_table_finish (h->it_ce->imported_encap[AFI_IP6]); | |
4618 | ||
4619 | if (h->import_mac) | |
4620 | { | |
4621 | struct rfapi_import_table *it; | |
4622 | void *cursor; | |
4623 | int rc; | |
4624 | ||
4625 | for (cursor = NULL, | |
4626 | rc = skiplist_next (h->import_mac, NULL, (void **) &it, &cursor); | |
4627 | !rc; | |
4628 | rc = skiplist_next (h->import_mac, NULL, (void **) &it, &cursor)) | |
4629 | { | |
4630 | ||
4631 | rfapiImportTableFlush (it); | |
4632 | XFREE (MTYPE_RFAPI_IMPORTTABLE, it); | |
4633 | } | |
4634 | skiplist_free (h->import_mac); | |
4635 | h->import_mac = NULL; | |
4636 | } | |
4637 | ||
4638 | work_queue_free (h->deferred_close_q); | |
4639 | ||
4640 | if (h->rfp != NULL) | |
4641 | rfp_stop (h->rfp); | |
4642 | XFREE (MTYPE_RFAPI_IMPORTTABLE, h->it_ce); | |
4643 | XFREE (MTYPE_RFAPI, h); | |
4644 | } | |
4645 | ||
4646 | struct rfapi_import_table * | |
4647 | rfapiImportTableRefAdd (struct bgp *bgp, struct ecommunity *rt_import_list) | |
4648 | { | |
4649 | struct rfapi *h; | |
4650 | struct rfapi_import_table *it; | |
4651 | afi_t afi; | |
4652 | ||
4653 | h = bgp->rfapi; | |
4654 | assert (h); | |
4655 | ||
4656 | for (it = h->imports; it; it = it->next) | |
4657 | { | |
4658 | if (ecommunity_cmp (it->rt_import_list, rt_import_list)) | |
4659 | break; | |
4660 | } | |
4661 | ||
a3b55c25 | 4662 | vnc_zlog_debug_verbose ("%s: matched it=%p", __func__, it); |
65efcfce LB |
4663 | |
4664 | if (!it) | |
4665 | { | |
4666 | it = | |
4667 | XCALLOC (MTYPE_RFAPI_IMPORTTABLE, sizeof (struct rfapi_import_table)); | |
4668 | assert (it); | |
4669 | it->next = h->imports; | |
4670 | h->imports = it; | |
4671 | ||
4672 | it->rt_import_list = ecommunity_dup (rt_import_list); | |
4673 | it->monitor_exterior_orphans = | |
4674 | skiplist_new (0, NULL, (void (*)(void *)) prefix_free); | |
4675 | ||
4676 | /* | |
4677 | * fill import route tables from RIBs | |
4678 | * | |
4679 | * Potential area for optimization. If this occurs when | |
4680 | * tables are large (e.g., the operator adds a nve group | |
4681 | * with a new RT list to a running system), it could take | |
4682 | * a while. | |
4683 | * | |
4684 | */ | |
4685 | for (afi = AFI_IP; afi < AFI_MAX; ++afi) | |
4686 | { | |
4687 | ||
4688 | it->imported_vpn[afi] = route_table_init (); | |
4689 | it->imported_encap[afi] = route_table_init (); | |
4690 | ||
4691 | rfapiBgpTableFilteredImport (bgp, it, afi, SAFI_MPLS_VPN); | |
4692 | rfapiBgpTableFilteredImport (bgp, it, afi, SAFI_ENCAP); | |
4693 | ||
4694 | vnc_import_bgp_exterior_redist_enable_it (bgp, afi, it); | |
4695 | } | |
4696 | } | |
4697 | ||
4698 | it->refcount += 1; | |
4699 | ||
4700 | return it; | |
4701 | } | |
4702 | ||
4703 | /* | |
4704 | * skiplist element free function | |
4705 | */ | |
4706 | static void | |
4707 | delete_rem_pfx_na_free (void *na) | |
4708 | { | |
4709 | uint32_t *pCounter = ((struct rfapi_nve_addr *) na)->info; | |
4710 | ||
4711 | *pCounter += 1; | |
4712 | XFREE (MTYPE_RFAPI_NVE_ADDR, na); | |
4713 | } | |
4714 | ||
4715 | /* | |
4716 | * Common deleter for IP and MAC import tables | |
4717 | */ | |
4718 | static void | |
4719 | rfapiDeleteRemotePrefixesIt ( | |
4720 | struct bgp *bgp, | |
4721 | struct rfapi_import_table *it, | |
4722 | struct prefix *un, | |
4723 | struct prefix *vn, | |
4724 | struct prefix *p, | |
4725 | int delete_active, | |
4726 | int delete_holddown, | |
4727 | uint32_t *pARcount, | |
4728 | uint32_t *pAHcount, | |
4729 | uint32_t *pHRcount, | |
4730 | uint32_t *pHHcount, | |
4731 | struct skiplist *uniq_active_nves, | |
4732 | struct skiplist *uniq_holddown_nves) | |
4733 | { | |
4734 | afi_t afi; | |
4735 | ||
4736 | #if DEBUG_L2_EXTRA | |
4737 | { | |
4738 | char buf_pfx[BUFSIZ]; | |
4739 | ||
4740 | if (p) | |
4741 | { | |
4742 | prefix2str (p, buf_pfx, BUFSIZ); | |
4743 | } | |
4744 | else | |
4745 | { | |
4746 | buf_pfx[0] = '*'; | |
4747 | buf_pfx[1] = 0; | |
4748 | } | |
4749 | ||
a3b55c25 | 4750 | vnc_zlog_debug_verbose ("%s: entry, p=%s, delete_active=%d, delete_holddown=%d", |
65efcfce LB |
4751 | __func__, buf_pfx, delete_active, delete_holddown); |
4752 | } | |
4753 | #endif | |
4754 | ||
4755 | for (afi = AFI_IP; afi < AFI_MAX; ++afi) | |
4756 | { | |
4757 | ||
4758 | struct route_table *rt; | |
4759 | struct route_node *rn; | |
4760 | ||
4761 | if (p && (family2afi (p->family) != afi)) | |
4762 | { | |
4763 | continue; | |
4764 | } | |
4765 | ||
4766 | rt = it->imported_vpn[afi]; | |
4767 | if (!rt) | |
4768 | continue; | |
4769 | ||
a3b55c25 | 4770 | vnc_zlog_debug_verbose ("%s: scanning rt for afi=%d", __func__, afi); |
65efcfce LB |
4771 | |
4772 | for (rn = route_top (rt); rn; rn = route_next (rn)) | |
4773 | { | |
4774 | struct bgp_info *bi; | |
4775 | struct bgp_info *next; | |
4776 | ||
4777 | if (VNC_DEBUG(IMPORT_DEL_REMOTE)) | |
4778 | { | |
4779 | char p1line[BUFSIZ]; | |
4780 | char p2line[BUFSIZ]; | |
4781 | ||
4782 | prefix2str (p, p1line, BUFSIZ); | |
4783 | prefix2str (&rn->p, p2line, BUFSIZ); | |
a3b55c25 | 4784 | vnc_zlog_debug_any ("%s: want %s, have %s", __func__, p1line, p2line); |
65efcfce LB |
4785 | } |
4786 | ||
4787 | if (p && prefix_cmp (p, &rn->p)) | |
4788 | continue; | |
4789 | ||
4790 | { | |
4791 | char buf_pfx[BUFSIZ]; | |
4792 | prefix2str (&rn->p, buf_pfx, BUFSIZ); | |
a3b55c25 | 4793 | vnc_zlog_debug_verbose ("%s: rn pfx=%s", __func__, buf_pfx); |
65efcfce LB |
4794 | } |
4795 | ||
4796 | /* TBD is this valid for afi == AFI_ETHER? */ | |
4797 | RFAPI_CHECK_REFCOUNT (rn, SAFI_MPLS_VPN, 1); | |
4798 | ||
4799 | for (bi = rn->info; bi; bi = next) | |
4800 | { | |
4801 | next = bi->next; | |
4802 | ||
4803 | struct prefix qpt; | |
4804 | struct prefix qct; | |
4805 | int qpt_valid = 0; | |
4806 | int qct_valid = 0; | |
4807 | int is_active = 0; | |
4808 | ||
a3b55c25 | 4809 | vnc_zlog_debug_verbose ("%s: examining bi %p", __func__, bi); |
65efcfce LB |
4810 | |
4811 | if (bi->attr) | |
4812 | { | |
4813 | if (!rfapiGetNexthop (bi->attr, &qpt)) | |
4814 | qpt_valid = 1; | |
4815 | } | |
4816 | if (vn) | |
4817 | { | |
4818 | if (!qpt_valid || !prefix_match (vn, &qpt)) | |
4819 | { | |
4820 | #if DEBUG_L2_EXTRA | |
a3b55c25 | 4821 | vnc_zlog_debug_verbose |
65efcfce LB |
4822 | ("%s: continue at vn && !qpt_valid || !prefix_match(vn, &qpt)", |
4823 | __func__); | |
4824 | #endif | |
4825 | continue; | |
4826 | } | |
4827 | } | |
4828 | ||
4829 | if (!rfapiGetUnAddrOfVpnBi (bi, &qct)) | |
4830 | qct_valid = 1; | |
4831 | ||
4832 | if (un) | |
4833 | { | |
4834 | if (!qct_valid || !prefix_match (un, &qct)) | |
4835 | { | |
4836 | #if DEBUG_L2_EXTRA | |
a3b55c25 | 4837 | vnc_zlog_debug_verbose |
65efcfce LB |
4838 | ("%s: continue at un && !qct_valid || !prefix_match(un, &qct)", |
4839 | __func__); | |
4840 | #endif | |
4841 | continue; | |
4842 | } | |
4843 | } | |
4844 | ||
4845 | ||
4846 | /* | |
4847 | * Blow bi away | |
4848 | */ | |
4849 | /* | |
4850 | * If this route is waiting to be deleted because of | |
4851 | * a previous withdraw, we must cancel its timer. | |
4852 | */ | |
4853 | if (CHECK_FLAG (bi->flags, BGP_INFO_REMOVED)) | |
4854 | { | |
4855 | if (!delete_holddown) | |
4856 | continue; | |
4857 | if (bi->extra->vnc.import.timer) | |
4858 | { | |
4859 | ||
4860 | struct thread *t = | |
4861 | (struct thread *) bi->extra->vnc.import.timer; | |
4862 | struct rfapi_withdraw *wcb = t->arg; | |
4863 | ||
4864 | wcb->import_table->holddown_count[afi] -= 1; | |
4865 | RFAPI_UPDATE_ITABLE_COUNT (bi, wcb->import_table, afi, | |
4866 | 1); | |
4867 | XFREE (MTYPE_RFAPI_WITHDRAW, wcb); | |
4868 | thread_cancel (t); | |
4869 | } | |
4870 | } | |
4871 | else | |
4872 | { | |
4873 | if (!delete_active) | |
4874 | continue; | |
4875 | is_active = 1; | |
4876 | } | |
4877 | ||
a3b55c25 | 4878 | vnc_zlog_debug_verbose |
65efcfce LB |
4879 | ("%s: deleting bi %p (qct_valid=%d, qpt_valid=%d, delete_holddown=%d, delete_active=%d)", |
4880 | __func__, bi, qct_valid, qpt_valid, delete_holddown, | |
4881 | delete_active); | |
4882 | ||
4883 | ||
4884 | /* | |
4885 | * add nve to list | |
4886 | */ | |
4887 | if (qct_valid && qpt_valid) | |
4888 | { | |
4889 | ||
4890 | struct rfapi_nve_addr na; | |
4891 | struct rfapi_nve_addr *nap; | |
4892 | ||
4893 | memset (&na, 0, sizeof (na)); | |
4894 | assert (!rfapiQprefix2Raddr (&qct, &na.un)); | |
4895 | assert (!rfapiQprefix2Raddr (&qpt, &na.vn)); | |
4896 | ||
4897 | if (skiplist_search ((is_active ? uniq_active_nves : | |
4898 | uniq_holddown_nves), &na, | |
4899 | (void **) &nap)) | |
4900 | { | |
4901 | char line[BUFSIZ]; | |
4902 | ||
4903 | nap = XCALLOC (MTYPE_RFAPI_NVE_ADDR, | |
4904 | sizeof (struct rfapi_nve_addr)); | |
4905 | assert (nap); | |
4906 | *nap = na; | |
4907 | nap->info = is_active ? pAHcount : pHHcount; | |
4908 | skiplist_insert ((is_active ? uniq_active_nves : | |
4909 | uniq_holddown_nves), nap, nap); | |
4910 | ||
4911 | rfapiNveAddr2Str (nap, line, BUFSIZ); | |
4912 | } | |
4913 | } | |
4914 | ||
4915 | vnc_direct_bgp_rh_del_route (bgp, afi, &rn->p, bi->peer); | |
4916 | ||
4917 | RFAPI_UPDATE_ITABLE_COUNT (bi, it, afi, -1); | |
4918 | it->holddown_count[afi] += 1; | |
4919 | rfapiExpireVpnNow (it, rn, bi, 1); | |
4920 | ||
a3b55c25 | 4921 | vnc_zlog_debug_verbose ("%s: incrementing count (is_active=%d)", |
65efcfce LB |
4922 | __func__, is_active); |
4923 | ||
4924 | if (is_active) | |
4925 | ++ * pARcount; | |
4926 | else | |
4927 | ++ * pHRcount; | |
4928 | } | |
4929 | } | |
4930 | } | |
4931 | } | |
4932 | ||
4933 | ||
4934 | /* | |
4935 | * For use by the "clear vnc prefixes" command | |
4936 | */ | |
4937 | /*------------------------------------------ | |
4938 | * rfapiDeleteRemotePrefixes | |
4939 | * | |
4940 | * UI helper: For use by the "clear vnc prefixes" command | |
4941 | * | |
4942 | * input: | |
4943 | * un if set, tunnel must match this prefix | |
4944 | * vn if set, nexthop prefix must match this prefix | |
4945 | * p if set, prefix must match this prefix | |
4946 | * | |
4947 | * output | |
4948 | * pARcount number of active routes deleted | |
4949 | * pAHcount number of active nves deleted | |
4950 | * pHRcount number of holddown routes deleted | |
4951 | * pHHcount number of holddown nves deleted | |
4952 | * | |
4953 | * return value: | |
4954 | * void | |
4955 | --------------------------------------------*/ | |
4956 | void | |
4957 | rfapiDeleteRemotePrefixes ( | |
4958 | struct prefix *un, | |
4959 | struct prefix *vn, | |
4960 | struct prefix *p, | |
4961 | int delete_active, | |
4962 | int delete_holddown, | |
4963 | uint32_t *pARcount, | |
4964 | uint32_t *pAHcount, | |
4965 | uint32_t *pHRcount, | |
4966 | uint32_t *pHHcount) | |
4967 | { | |
4968 | struct bgp *bgp; | |
4969 | struct rfapi *h; | |
4970 | struct rfapi_import_table *it; | |
4971 | uint32_t deleted_holddown_route_count = 0; | |
4972 | uint32_t deleted_active_route_count = 0; | |
4973 | uint32_t deleted_holddown_nve_count = 0; | |
4974 | uint32_t deleted_active_nve_count = 0; | |
4975 | struct skiplist *uniq_holddown_nves; | |
4976 | struct skiplist *uniq_active_nves; | |
4977 | ||
4978 | VNC_ITRCCK; | |
4979 | ||
4980 | bgp = bgp_get_default (); /* assume 1 instance for now */ | |
4981 | /* If no bgp instantiated yet, no vnc prefixes exist */ | |
4982 | if (!bgp) | |
4983 | return; | |
4984 | ||
4985 | h = bgp->rfapi; | |
4986 | assert (h); | |
4987 | ||
4988 | uniq_holddown_nves = | |
4989 | skiplist_new (0, rfapi_nve_addr_cmp, delete_rem_pfx_na_free); | |
4990 | uniq_active_nves = | |
4991 | skiplist_new (0, rfapi_nve_addr_cmp, delete_rem_pfx_na_free); | |
4992 | ||
4993 | /* | |
4994 | * Iterate over all import tables; do a filtered import | |
4995 | * for the afi/safi combination | |
4996 | */ | |
4997 | ||
4998 | for (it = h->imports; it; it = it->next) | |
4999 | { | |
5000 | ||
a3b55c25 | 5001 | vnc_zlog_debug_verbose |
65efcfce LB |
5002 | ("%s: calling rfapiDeleteRemotePrefixesIt() on (IP) import %p", |
5003 | __func__, it); | |
5004 | ||
5005 | rfapiDeleteRemotePrefixesIt ( | |
5006 | bgp, | |
5007 | it, | |
5008 | un, | |
5009 | vn, | |
5010 | p, | |
5011 | delete_active, | |
5012 | delete_holddown, | |
5013 | &deleted_active_route_count, | |
5014 | &deleted_active_nve_count, | |
5015 | &deleted_holddown_route_count, | |
5016 | &deleted_holddown_nve_count, | |
5017 | uniq_active_nves, | |
5018 | uniq_holddown_nves); | |
5019 | } | |
5020 | ||
5021 | /* | |
5022 | * Now iterate over L2 import tables | |
5023 | */ | |
5024 | if (h->import_mac && !(p && (p->family != AF_ETHERNET))) | |
5025 | { | |
5026 | ||
5027 | void *cursor = NULL; | |
5028 | int rc; | |
5029 | ||
5030 | for (cursor = NULL, | |
5031 | rc = skiplist_next (h->import_mac, NULL, (void **) &it, &cursor); | |
5032 | !rc; | |
5033 | rc = skiplist_next (h->import_mac, NULL, (void **) &it, &cursor)) | |
5034 | { | |
5035 | ||
a3b55c25 | 5036 | vnc_zlog_debug_verbose |
65efcfce LB |
5037 | ("%s: calling rfapiDeleteRemotePrefixesIt() on import_mac %p", |
5038 | __func__, it); | |
5039 | ||
5040 | rfapiDeleteRemotePrefixesIt ( | |
5041 | bgp, | |
5042 | it, | |
5043 | un, | |
5044 | vn, | |
5045 | p, | |
5046 | delete_active, | |
5047 | delete_holddown, | |
5048 | &deleted_active_route_count, | |
5049 | &deleted_active_nve_count, | |
5050 | &deleted_holddown_route_count, | |
5051 | &deleted_holddown_nve_count, | |
5052 | uniq_active_nves, | |
5053 | uniq_holddown_nves); | |
5054 | } | |
5055 | } | |
5056 | ||
5057 | /* | |
5058 | * our custom element freeing function above counts as it deletes | |
5059 | */ | |
5060 | skiplist_free (uniq_holddown_nves); | |
5061 | skiplist_free (uniq_active_nves); | |
5062 | ||
5063 | if (pARcount) | |
5064 | *pARcount = deleted_active_route_count; | |
5065 | if (pAHcount) | |
5066 | *pAHcount = deleted_active_nve_count; | |
5067 | if (pHRcount) | |
5068 | *pHRcount = deleted_holddown_route_count; | |
5069 | if (pHHcount) | |
5070 | *pHHcount = deleted_holddown_nve_count; | |
5071 | ||
5072 | VNC_ITRCCK; | |
5073 | } | |
5074 | ||
5075 | /*------------------------------------------ | |
5076 | * rfapiCountRemoteRoutes | |
5077 | * | |
5078 | * UI helper: count VRF routes from BGP side | |
5079 | * | |
5080 | * input: | |
5081 | * | |
5082 | * output | |
5083 | * pALRcount count of active local routes | |
5084 | * pARRcount count of active remote routes | |
5085 | * pHRcount count of holddown routes | |
5086 | * pIRcount count of direct imported routes | |
5087 | * | |
5088 | * return value: | |
5089 | * void | |
5090 | --------------------------------------------*/ | |
5091 | void | |
5092 | rfapiCountAllItRoutes (int *pALRcount, /* active local routes */ | |
5093 | int *pARRcount, /* active remote routes */ | |
5094 | int *pHRcount, /* holddown routes */ | |
5095 | int *pIRcount) /* imported routes */ | |
5096 | { | |
5097 | struct bgp *bgp; | |
5098 | struct rfapi *h; | |
5099 | struct rfapi_import_table *it; | |
5100 | afi_t afi; | |
5101 | ||
5102 | int total_active_local = 0; | |
5103 | int total_active_remote = 0; | |
5104 | int total_holddown = 0; | |
5105 | int total_imported = 0; | |
5106 | ||
5107 | bgp = bgp_get_default (); /* assume 1 instance for now */ | |
5108 | assert (bgp); | |
5109 | ||
5110 | h = bgp->rfapi; | |
5111 | assert (h); | |
5112 | ||
5113 | /* | |
5114 | * Iterate over all import tables; do a filtered import | |
5115 | * for the afi/safi combination | |
5116 | */ | |
5117 | ||
5118 | for (it = h->imports; it; it = it->next) | |
5119 | { | |
5120 | ||
5121 | for (afi = AFI_IP; afi < AFI_MAX; ++afi) | |
5122 | { | |
5123 | ||
5124 | total_active_local += it->local_count[afi]; | |
5125 | total_active_remote += it->remote_count[afi]; | |
5126 | total_holddown += it->holddown_count[afi]; | |
5127 | total_imported += it->imported_count[afi]; | |
5128 | ||
5129 | } | |
5130 | } | |
5131 | ||
5132 | void *cursor; | |
5133 | int rc; | |
5134 | ||
5135 | if (h->import_mac) | |
5136 | { | |
5137 | for (cursor = NULL, | |
5138 | rc = skiplist_next (h->import_mac, NULL, (void **) &it, &cursor); | |
5139 | !rc; | |
5140 | rc = skiplist_next (h->import_mac, NULL, (void **) &it, &cursor)) | |
5141 | { | |
5142 | ||
5143 | total_active_local += it->local_count[AFI_ETHER]; | |
5144 | total_active_remote += it->remote_count[AFI_ETHER]; | |
5145 | total_holddown += it->holddown_count[AFI_ETHER]; | |
5146 | total_imported += it->imported_count[AFI_ETHER]; | |
5147 | ||
5148 | } | |
5149 | } | |
5150 | ||
5151 | ||
5152 | if (pALRcount) | |
5153 | { | |
5154 | *pALRcount = total_active_local; | |
5155 | } | |
5156 | if (pARRcount) | |
5157 | { | |
5158 | *pARRcount = total_active_remote; | |
5159 | } | |
5160 | if (pHRcount) | |
5161 | { | |
5162 | *pHRcount = total_holddown; | |
5163 | } | |
5164 | if (pIRcount) | |
5165 | { | |
5166 | *pIRcount = total_imported; | |
5167 | } | |
5168 | } | |
5169 | ||
5170 | /*------------------------------------------ | |
5171 | * rfapiGetHolddownFromLifetime | |
5172 | * | |
5173 | * calculate holddown value based on lifetime | |
5174 | * | |
5175 | * input: | |
5176 | * lifetime lifetime | |
5177 | * | |
5178 | * return value: | |
5179 | * Holddown value based on lifetime, holddown_factor, | |
5180 | * and RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY | |
5181 | * | |
5182 | --------------------------------------------*/ | |
5183 | /* hold down time maxes out at RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY */ | |
5184 | uint32_t | |
5185 | rfapiGetHolddownFromLifetime (uint32_t lifetime) | |
5186 | { | |
5187 | uint32_t factor; | |
5188 | struct bgp *bgp; | |
5189 | ||
5190 | bgp = bgp_get_default (); | |
5191 | if (bgp && bgp->rfapi_cfg) | |
5192 | factor = bgp->rfapi_cfg->rfp_cfg.holddown_factor; | |
5193 | else | |
5194 | factor = RFAPI_RFP_CFG_DEFAULT_HOLDDOWN_FACTOR; | |
5195 | ||
5196 | if (factor < 100 || lifetime < RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY) | |
5197 | lifetime = lifetime * factor / 100; | |
5198 | if (lifetime < RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY) | |
5199 | return lifetime; | |
5200 | else | |
5201 | return RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY; | |
5202 | } |