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