]>
Commit | Line | Data |
---|---|---|
718e3744 | 1 | /* Routing Information Base. |
2 | * Copyright (C) 1997, 98, 99, 2001 Kunihiro Ishiguro | |
3 | * | |
4 | * This file is part of GNU Zebra. | |
5 | * | |
6 | * GNU Zebra is free software; you can redistribute it and/or modify it | |
7 | * under the terms of the GNU General Public License as published by the | |
8 | * Free Software Foundation; either version 2, or (at your option) any | |
9 | * later version. | |
10 | * | |
11 | * GNU Zebra is distributed in the hope that it will be useful, but | |
12 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | * General Public License for more details. | |
15 | * | |
896014f4 DL |
16 | * You should have received a copy of the GNU General Public License along |
17 | * with this program; see the file COPYING; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
718e3744 | 19 | */ |
20 | ||
21 | #include <zebra.h> | |
22 | ||
718e3744 | 23 | #include "command.h" |
43e52561 QY |
24 | #include "if.h" |
25 | #include "linklist.h" | |
718e3744 | 26 | #include "log.h" |
43e52561 QY |
27 | #include "memory.h" |
28 | #include "mpls.h" | |
29 | #include "nexthop.h" | |
30 | #include "prefix.h" | |
7514fb77 PJ |
31 | #include "prefix.h" |
32 | #include "routemap.h" | |
43e52561 | 33 | #include "sockunion.h" |
05737783 | 34 | #include "srcdest_table.h" |
43e52561 QY |
35 | #include "table.h" |
36 | #include "thread.h" | |
37 | #include "vrf.h" | |
38 | #include "workqueue.h" | |
50d89650 | 39 | #include "nexthop_group_private.h" |
00dffa8c | 40 | #include "frr_pthread.h" |
718e3744 | 41 | |
89272910 | 42 | #include "zebra/zebra_router.h" |
43e52561 QY |
43 | #include "zebra/connected.h" |
44 | #include "zebra/debug.h" | |
45 | #include "zebra/interface.h" | |
46 | #include "zebra/redistribute.h" | |
718e3744 | 47 | #include "zebra/rib.h" |
48 | #include "zebra/rt.h" | |
43e52561 QY |
49 | #include "zebra/zapi_msg.h" |
50 | #include "zebra/zebra_errors.h" | |
51 | #include "zebra/zebra_memory.h" | |
7c551956 | 52 | #include "zebra/zebra_ns.h" |
fb018d25 | 53 | #include "zebra/zebra_rnh.h" |
43e52561 QY |
54 | #include "zebra/zebra_routemap.h" |
55 | #include "zebra/zebra_vrf.h" | |
6134fd82 | 56 | #include "zebra/zebra_vxlan.h" |
7cdb1a84 MS |
57 | #include "zebra/zapi_msg.h" |
58 | #include "zebra/zebra_dplane.h" | |
59 | ||
2a99ab95 SW |
60 | DEFINE_MTYPE_STATIC(ZEBRA, RIB_UPDATE_CTX, "Rib update context object"); |
61 | ||
7cdb1a84 MS |
62 | /* |
63 | * Event, list, and mutex for delivery of dataplane results | |
64 | */ | |
65 | static pthread_mutex_t dplane_mutex; | |
66 | static struct thread *t_dplane; | |
25779064 | 67 | static struct dplane_ctx_q rib_dplane_q; |
718e3744 | 68 | |
d62a17ae | 69 | DEFINE_HOOK(rib_update, (struct route_node * rn, const char *reason), |
70 | (rn, reason)) | |
4f8ea50c | 71 | |
6baf7bb8 DS |
72 | /* Should we allow non Quagga processes to delete our routes */ |
73 | extern int allow_delete; | |
74 | ||
718e3744 | 75 | /* Each route type's string and default distance value. */ |
d62a17ae | 76 | static const struct { |
77 | int key; | |
8dc7a759 | 78 | uint8_t distance; |
73547a75 | 79 | uint8_t meta_q_map; |
d62a17ae | 80 | } route_info[ZEBRA_ROUTE_MAX] = { |
e22e8001 SW |
81 | [ZEBRA_ROUTE_NHG] = {ZEBRA_ROUTE_NHG, 255 /* Uneeded for nhg's */, 0}, |
82 | [ZEBRA_ROUTE_SYSTEM] = {ZEBRA_ROUTE_SYSTEM, 0, 5}, | |
83 | [ZEBRA_ROUTE_KERNEL] = {ZEBRA_ROUTE_KERNEL, 0, 1}, | |
84 | [ZEBRA_ROUTE_CONNECT] = {ZEBRA_ROUTE_CONNECT, 0, 1}, | |
85 | [ZEBRA_ROUTE_STATIC] = {ZEBRA_ROUTE_STATIC, 1, 2}, | |
86 | [ZEBRA_ROUTE_RIP] = {ZEBRA_ROUTE_RIP, 120, 3}, | |
87 | [ZEBRA_ROUTE_RIPNG] = {ZEBRA_ROUTE_RIPNG, 120, 3}, | |
88 | [ZEBRA_ROUTE_OSPF] = {ZEBRA_ROUTE_OSPF, 110, 3}, | |
89 | [ZEBRA_ROUTE_OSPF6] = {ZEBRA_ROUTE_OSPF6, 110, 3}, | |
90 | [ZEBRA_ROUTE_ISIS] = {ZEBRA_ROUTE_ISIS, 115, 3}, | |
91 | [ZEBRA_ROUTE_BGP] = {ZEBRA_ROUTE_BGP, 20 /* IBGP is 200. */, 4}, | |
92 | [ZEBRA_ROUTE_PIM] = {ZEBRA_ROUTE_PIM, 255, 5}, | |
93 | [ZEBRA_ROUTE_EIGRP] = {ZEBRA_ROUTE_EIGRP, 90, 3}, | |
94 | [ZEBRA_ROUTE_NHRP] = {ZEBRA_ROUTE_NHRP, 10, 3}, | |
95 | [ZEBRA_ROUTE_HSLS] = {ZEBRA_ROUTE_HSLS, 255, 5}, | |
96 | [ZEBRA_ROUTE_OLSR] = {ZEBRA_ROUTE_OLSR, 255, 5}, | |
97 | [ZEBRA_ROUTE_TABLE] = {ZEBRA_ROUTE_TABLE, 150, 2}, | |
98 | [ZEBRA_ROUTE_LDP] = {ZEBRA_ROUTE_LDP, 150, 5}, | |
99 | [ZEBRA_ROUTE_VNC] = {ZEBRA_ROUTE_VNC, 20, 4}, | |
100 | [ZEBRA_ROUTE_VNC_DIRECT] = {ZEBRA_ROUTE_VNC_DIRECT, 20, 4}, | |
101 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = {ZEBRA_ROUTE_VNC_DIRECT_RH, 20, 4}, | |
102 | [ZEBRA_ROUTE_BGP_DIRECT] = {ZEBRA_ROUTE_BGP_DIRECT, 20, 4}, | |
103 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = {ZEBRA_ROUTE_BGP_DIRECT_EXT, 20, 4}, | |
104 | [ZEBRA_ROUTE_BABEL] = {ZEBRA_ROUTE_BABEL, 100, 3}, | |
105 | [ZEBRA_ROUTE_SHARP] = {ZEBRA_ROUTE_SHARP, 150, 5}, | |
106 | [ZEBRA_ROUTE_PBR] = {ZEBRA_ROUTE_PBR, 200, 5}, | |
107 | [ZEBRA_ROUTE_BFD] = {ZEBRA_ROUTE_BFD, 255, 5}, | |
108 | [ZEBRA_ROUTE_OPENFABRIC] = {ZEBRA_ROUTE_OPENFABRIC, 115, 3}, | |
109 | [ZEBRA_ROUTE_VRRP] = {ZEBRA_ROUTE_VRRP, 255, 5} | |
d6abd8b0 | 110 | /* Any new route type added to zebra, should be mirrored here */ |
c710b277 | 111 | |
d62a17ae | 112 | /* no entry/default: 150 */ |
718e3744 | 113 | }; |
6b0655a2 | 114 | |
d62a17ae | 115 | static void __attribute__((format(printf, 5, 6))) |
116 | _rnode_zlog(const char *_func, vrf_id_t vrf_id, struct route_node *rn, | |
117 | int priority, const char *msgfmt, ...) | |
2263a412 | 118 | { |
d62a17ae | 119 | char buf[SRCDEST2STR_BUFFER + sizeof(" (MRIB)")]; |
120 | char msgbuf[512]; | |
121 | va_list ap; | |
2263a412 | 122 | |
d62a17ae | 123 | va_start(ap, msgfmt); |
124 | vsnprintf(msgbuf, sizeof(msgbuf), msgfmt, ap); | |
125 | va_end(ap); | |
2263a412 | 126 | |
d62a17ae | 127 | if (rn) { |
128 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
129 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
cb653491 | 130 | |
d62a17ae | 131 | if (info->safi == SAFI_MULTICAST) |
eab4a5c2 | 132 | strlcat(buf, " (MRIB)", sizeof(buf)); |
d62a17ae | 133 | } else { |
134 | snprintf(buf, sizeof(buf), "{(route_node *) NULL}"); | |
135 | } | |
2263a412 | 136 | |
d62a17ae | 137 | zlog(priority, "%s: %d:%s: %s", _func, vrf_id, buf, msgbuf); |
2263a412 DL |
138 | } |
139 | ||
d62a17ae | 140 | #define rnode_debug(node, vrf_id, ...) \ |
2263a412 | 141 | _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__) |
d62a17ae | 142 | #define rnode_info(node, ...) \ |
2263a412 DL |
143 | _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__) |
144 | ||
fd289fc8 | 145 | uint8_t route_distance(int type) |
40c7bdb0 | 146 | { |
fd289fc8 | 147 | uint8_t distance; |
40c7bdb0 | 148 | |
d62a17ae | 149 | if ((unsigned)type >= array_size(route_info)) |
150 | distance = 150; | |
151 | else | |
152 | distance = route_info[type].distance; | |
40c7bdb0 | 153 | |
d62a17ae | 154 | return distance; |
40c7bdb0 | 155 | } |
156 | ||
d7c0a89a | 157 | int is_zebra_valid_kernel_table(uint32_t table_id) |
7a4bb9c5 | 158 | { |
8f500a1c | 159 | #ifdef linux |
d62a17ae | 160 | if ((table_id == RT_TABLE_UNSPEC) || (table_id == RT_TABLE_LOCAL) |
161 | || (table_id == RT_TABLE_COMPAT)) | |
162 | return 0; | |
8f500a1c RW |
163 | #endif |
164 | ||
d62a17ae | 165 | return 1; |
7a4bb9c5 DS |
166 | } |
167 | ||
d7c0a89a | 168 | int is_zebra_main_routing_table(uint32_t table_id) |
7a4bb9c5 | 169 | { |
c447ad08 | 170 | if (table_id == RT_TABLE_MAIN) |
d62a17ae | 171 | return 1; |
172 | return 0; | |
7a4bb9c5 DS |
173 | } |
174 | ||
86391e56 | 175 | int zebra_check_addr(const struct prefix *p) |
0aabccc0 | 176 | { |
d62a17ae | 177 | if (p->family == AF_INET) { |
d7c0a89a | 178 | uint32_t addr; |
0aabccc0 | 179 | |
d62a17ae | 180 | addr = p->u.prefix4.s_addr; |
181 | addr = ntohl(addr); | |
0aabccc0 | 182 | |
d62a17ae | 183 | if (IPV4_NET127(addr) || IN_CLASSD(addr) |
184 | || IPV4_LINKLOCAL(addr)) | |
185 | return 0; | |
186 | } | |
187 | if (p->family == AF_INET6) { | |
188 | if (IN6_IS_ADDR_LOOPBACK(&p->u.prefix6)) | |
189 | return 0; | |
190 | if (IN6_IS_ADDR_LINKLOCAL(&p->u.prefix6)) | |
191 | return 0; | |
192 | } | |
193 | return 1; | |
0aabccc0 DD |
194 | } |
195 | ||
6e26278c DS |
196 | /** |
197 | * copy_nexthop - copy a nexthop to the rib structure. | |
198 | */ | |
d62a17ae | 199 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) |
6e26278c | 200 | { |
c415d895 MS |
201 | assert(!re->nhe->nhg.nexthop); |
202 | copy_nexthops(&re->nhe->nhg.nexthop, nh, NULL); | |
718e3744 | 203 | } |
204 | ||
5463ce26 MS |
205 | static void route_entry_attach_ref(struct route_entry *re, |
206 | struct nhg_hash_entry *new) | |
718e3744 | 207 | { |
0eb97b86 | 208 | re->nhe = new; |
5463ce26 | 209 | re->nhe_id = new->id; |
718e3744 | 210 | |
5463ce26 | 211 | zebra_nhg_increment_ref(new); |
718e3744 | 212 | } |
213 | ||
5463ce26 | 214 | int route_entry_update_nhe(struct route_entry *re, struct nhg_hash_entry *new) |
718e3744 | 215 | { |
5463ce26 MS |
216 | struct nhg_hash_entry *old = NULL; |
217 | int ret = 0; | |
718e3744 | 218 | |
5463ce26 | 219 | if (new == NULL) { |
c415d895 MS |
220 | if (re->nhe) |
221 | zebra_nhg_decrement_ref(re->nhe); | |
222 | re->nhe = NULL; | |
5463ce26 MS |
223 | goto done; |
224 | } | |
718e3744 | 225 | |
5463ce26 | 226 | if (re->nhe_id != new->id) { |
c415d895 | 227 | old = re->nhe; |
718e3744 | 228 | |
5463ce26 | 229 | route_entry_attach_ref(re, new); |
718e3744 | 230 | |
5463ce26 MS |
231 | if (old) |
232 | zebra_nhg_decrement_ref(old); | |
84a89a8d | 233 | } else if (!re->nhe) |
5463ce26 MS |
234 | /* This is the first time it's being attached */ |
235 | route_entry_attach_ref(re, new); | |
595db7f1 | 236 | |
5463ce26 MS |
237 | done: |
238 | return ret; | |
595db7f1 | 239 | } |
240 | ||
d62a17ae | 241 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
242 | union g_addr *addr, struct route_node **rn_out) | |
243 | { | |
244 | struct prefix p; | |
245 | struct route_table *table; | |
246 | struct route_node *rn; | |
5f7a4718 | 247 | struct route_entry *match = NULL; |
d62a17ae | 248 | |
249 | /* Lookup table. */ | |
250 | table = zebra_vrf_table(afi, safi, vrf_id); | |
251 | if (!table) | |
252 | return 0; | |
253 | ||
254 | memset(&p, 0, sizeof(struct prefix)); | |
255 | p.family = afi; | |
256 | if (afi == AFI_IP) { | |
257 | p.u.prefix4 = addr->ipv4; | |
258 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
259 | } else { | |
260 | p.u.prefix6 = addr->ipv6; | |
261 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
262 | } | |
718e3744 | 263 | |
d62a17ae | 264 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 265 | |
d62a17ae | 266 | while (rn) { |
5f7a4718 DS |
267 | rib_dest_t *dest; |
268 | ||
d62a17ae | 269 | route_unlock_node(rn); |
718e3744 | 270 | |
5f7a4718 | 271 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
272 | if (dest && dest->selected_fib |
273 | && !CHECK_FLAG(dest->selected_fib->status, | |
274 | ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 275 | match = dest->selected_fib; |
718e3744 | 276 | |
d62a17ae | 277 | /* If there is no selected route or matched route is EGP, go up |
278 | tree. */ | |
279 | if (!match) { | |
280 | do { | |
281 | rn = rn->parent; | |
282 | } while (rn && rn->info == NULL); | |
283 | if (rn) | |
284 | route_lock_node(rn); | |
285 | } else { | |
286 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
677c1dd5 DS |
287 | if (!CHECK_FLAG(match->status, |
288 | ROUTE_ENTRY_INSTALLED)) | |
d62a17ae | 289 | return NULL; |
290 | } | |
291 | ||
292 | if (rn_out) | |
293 | *rn_out = rn; | |
294 | return match; | |
295 | } | |
16814f96 | 296 | } |
d62a17ae | 297 | return NULL; |
298 | } | |
299 | ||
300 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
301 | struct in_addr addr, | |
302 | struct route_node **rn_out) | |
303 | { | |
304 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
305 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
306 | union g_addr gaddr = {.ipv4 = addr}; | |
307 | ||
526052fb | 308 | switch (zrouter.ipv4_multicast_mode) { |
d62a17ae | 309 | case MCAST_MRIB_ONLY: |
310 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
311 | rn_out); | |
312 | case MCAST_URIB_ONLY: | |
313 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
314 | case MCAST_NO_CONFIG: | |
315 | case MCAST_MIX_MRIB_FIRST: | |
316 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
317 | &m_rn); | |
318 | if (!mre) | |
319 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
320 | &gaddr, &u_rn); | |
321 | break; | |
322 | case MCAST_MIX_DISTANCE: | |
323 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
324 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
325 | if (mre && ure) | |
326 | re = ure->distance < mre->distance ? ure : mre; | |
327 | else if (mre) | |
328 | re = mre; | |
329 | else if (ure) | |
330 | re = ure; | |
331 | break; | |
332 | case MCAST_MIX_PFXLEN: | |
333 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
334 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
335 | if (mre && ure) | |
336 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
337 | else if (mre) | |
338 | re = mre; | |
339 | else if (ure) | |
340 | re = ure; | |
341 | break; | |
718e3744 | 342 | } |
33550aa8 | 343 | |
d62a17ae | 344 | if (rn_out) |
345 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 346 | |
d62a17ae | 347 | if (IS_ZEBRA_DEBUG_RIB) { |
348 | char buf[BUFSIZ]; | |
349 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 350 | |
32391aff DS |
351 | zlog_debug("%s: %s: vrf: %u found %s, using %s", |
352 | __func__, buf, vrf_id, | |
d62a17ae | 353 | mre ? (ure ? "MRIB+URIB" : "MRIB") |
354 | : ure ? "URIB" : "nothing", | |
355 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
356 | } | |
357 | return re; | |
4623d897 DL |
358 | } |
359 | ||
d62a17ae | 360 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 361 | { |
d62a17ae | 362 | struct route_table *table; |
363 | struct route_node *rn; | |
5f7a4718 | 364 | struct route_entry *match = NULL; |
5f7a4718 | 365 | rib_dest_t *dest; |
718e3744 | 366 | |
d62a17ae | 367 | /* Lookup table. */ |
368 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
369 | if (!table) | |
370 | return 0; | |
718e3744 | 371 | |
d62a17ae | 372 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 373 | |
d62a17ae | 374 | /* No route for this prefix. */ |
375 | if (!rn) | |
376 | return NULL; | |
718e3744 | 377 | |
d62a17ae | 378 | /* Unlock node. */ |
379 | route_unlock_node(rn); | |
5f7a4718 | 380 | dest = rib_dest_from_rnode(rn); |
718e3744 | 381 | |
996c9314 LB |
382 | if (dest && dest->selected_fib |
383 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 384 | match = dest->selected_fib; |
718e3744 | 385 | |
d62a17ae | 386 | if (!match) |
387 | return NULL; | |
718e3744 | 388 | |
d62a17ae | 389 | if (match->type == ZEBRA_ROUTE_CONNECT) |
390 | return match; | |
f9e1b38e | 391 | |
677c1dd5 DS |
392 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_INSTALLED)) |
393 | return match; | |
718e3744 | 394 | |
d62a17ae | 395 | return NULL; |
718e3744 | 396 | } |
397 | ||
a64448ba DS |
398 | /* |
399 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
400 | * (nexthops) must have a label. | |
401 | */ | |
d62a17ae | 402 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 403 | { |
d62a17ae | 404 | struct nexthop *nexthop = NULL; |
a64448ba | 405 | |
d62a17ae | 406 | if (re->type != ZEBRA_ROUTE_BGP) |
407 | return 0; | |
a64448ba | 408 | |
c415d895 | 409 | for (ALL_NEXTHOPS(re->nhe->nhg, nexthop)) |
d62a17ae | 410 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) |
411 | return 0; | |
6b0655a2 | 412 | |
d62a17ae | 413 | return 1; |
a64448ba | 414 | } |
718e3744 | 415 | |
6ae24471 DS |
416 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
417 | * is only used for IPv4. | |
418 | */ | |
0c555cc6 DS |
419 | void rib_install_kernel(struct route_node *rn, struct route_entry *re, |
420 | struct route_entry *old) | |
718e3744 | 421 | { |
d62a17ae | 422 | struct nexthop *nexthop; |
423 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 424 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
97f5b441 MS |
425 | const struct prefix *p, *src_p; |
426 | enum zebra_dplane_result ret; | |
427 | ||
428 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
718e3744 | 429 | |
d62a17ae | 430 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 431 | |
d62a17ae | 432 | if (info->safi != SAFI_UNICAST) { |
c415d895 | 433 | for (ALL_NEXTHOPS(re->nhe->nhg, nexthop)) |
d62a17ae | 434 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 435 | return; |
d62a17ae | 436 | } |
718e3744 | 437 | |
9ef49038 | 438 | |
8dfbc657 SW |
439 | /* |
440 | * Install the resolved nexthop object first. | |
441 | */ | |
c415d895 | 442 | zebra_nhg_install_kernel(re->nhe); |
8dfbc657 | 443 | |
2063a814 DS |
444 | /* |
445 | * If this is a replace to a new RE let the originator of the RE | |
446 | * know that they've lost | |
447 | */ | |
9a9f8926 | 448 | if (old && (old != re) && (old->type != re->type)) |
28610f7e | 449 | zsend_route_notify_owner(old, p, ZAPI_ROUTE_BETTER_ADMIN_WON); |
25b9cb0c | 450 | |
97f5b441 MS |
451 | /* Update fib selection */ |
452 | dest->selected_fib = re; | |
453 | ||
d62a17ae | 454 | /* |
455 | * Make sure we update the FPM any time we send new information to | |
456 | * the kernel. | |
457 | */ | |
458 | hook_call(rib_update, rn, "installing in kernel"); | |
97f5b441 MS |
459 | |
460 | /* Send add or update */ | |
8263d1d0 | 461 | if (old) |
97f5b441 | 462 | ret = dplane_route_update(rn, re, old); |
5709131c | 463 | else |
97f5b441 | 464 | ret = dplane_route_add(rn, re); |
97f5b441 MS |
465 | |
466 | switch (ret) { | |
ea1c14f6 | 467 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
60f98b23 | 468 | SET_FLAG(re->status, ROUTE_ENTRY_QUEUED); |
efe6c026 MS |
469 | |
470 | if (old) { | |
60f98b23 | 471 | SET_FLAG(old->status, ROUTE_ENTRY_QUEUED); |
efe6c026 MS |
472 | |
473 | /* Free old FIB nexthop group */ | |
474 | if (old->fib_ng.nexthop) { | |
475 | nexthops_free(old->fib_ng.nexthop); | |
476 | old->fib_ng.nexthop = NULL; | |
477 | } | |
efe6c026 MS |
478 | } |
479 | ||
97f5b441 MS |
480 | if (zvrf) |
481 | zvrf->installs_queued++; | |
1e885672 | 482 | break; |
ea1c14f6 | 483 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
484 | { |
485 | char str[SRCDEST2STR_BUFFER]; | |
486 | ||
487 | srcdest_rnode2str(rn, str, sizeof(str)); | |
488 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
489 | "%u:%s: Failed to enqueue dataplane install", | |
490 | re->vrf_id, str); | |
1e885672 | 491 | break; |
97f5b441 | 492 | } |
ea1c14f6 | 493 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
97f5b441 MS |
494 | if (zvrf) |
495 | zvrf->installs++; | |
1e885672 DS |
496 | break; |
497 | } | |
d62a17ae | 498 | |
0c555cc6 | 499 | return; |
718e3744 | 500 | } |
501 | ||
502 | /* Uninstall the route from kernel. */ | |
0c555cc6 | 503 | void rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 504 | { |
d62a17ae | 505 | struct nexthop *nexthop; |
506 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 507 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
05737783 | 508 | |
d62a17ae | 509 | if (info->safi != SAFI_UNICAST) { |
677c1dd5 | 510 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
c415d895 | 511 | for (ALL_NEXTHOPS(re->nhe->nhg, nexthop)) |
d6792f9d | 512 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 513 | return; |
d62a17ae | 514 | } |
416ec78d | 515 | |
d62a17ae | 516 | /* |
517 | * Make sure we update the FPM any time we send new information to | |
97f5b441 | 518 | * the dataplane. |
d62a17ae | 519 | */ |
520 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
97f5b441 MS |
521 | |
522 | switch (dplane_route_delete(rn, re)) { | |
ea1c14f6 | 523 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
97f5b441 MS |
524 | if (zvrf) |
525 | zvrf->removals_queued++; | |
1e885672 | 526 | break; |
ea1c14f6 | 527 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
528 | { |
529 | char str[SRCDEST2STR_BUFFER]; | |
530 | ||
531 | srcdest_rnode2str(rn, str, sizeof(str)); | |
532 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
533 | "%u:%s: Failed to enqueue dataplane uninstall", | |
534 | re->vrf_id, str); | |
1e885672 | 535 | break; |
97f5b441 | 536 | } |
ea1c14f6 | 537 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
1e885672 DS |
538 | if (zvrf) |
539 | zvrf->removals++; | |
540 | break; | |
541 | } | |
718e3744 | 542 | |
0c555cc6 | 543 | return; |
718e3744 | 544 | } |
545 | ||
546 | /* Uninstall the route from kernel. */ | |
d62a17ae | 547 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 548 | { |
d62a17ae | 549 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
5f7a4718 | 550 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
97f5b441 | 551 | struct nexthop *nexthop; |
416ec78d | 552 | |
5f7a4718 | 553 | if (dest && dest->selected_fib == re) { |
d62a17ae | 554 | if (info->safi == SAFI_UNICAST) |
555 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 556 | |
d62a17ae | 557 | /* If labeled-unicast route, uninstall transit LSP. */ |
558 | if (zebra_rib_labeled_unicast(re)) | |
559 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
7cdb1a84 | 560 | |
3cdba47a | 561 | rib_uninstall_kernel(rn, re); |
97f5b441 MS |
562 | |
563 | dest->selected_fib = NULL; | |
564 | ||
efe6c026 MS |
565 | /* Free FIB nexthop group, if present */ |
566 | if (re->fib_ng.nexthop) { | |
567 | nexthops_free(re->fib_ng.nexthop); | |
568 | re->fib_ng.nexthop = NULL; | |
569 | } | |
570 | ||
c415d895 | 571 | for (ALL_NEXTHOPS(re->nhe->nhg, nexthop)) |
97f5b441 | 572 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
d62a17ae | 573 | } |
446bb95e | 574 | |
d62a17ae | 575 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
86391e56 MS |
576 | const struct prefix *p, *src_p; |
577 | ||
d62a17ae | 578 | srcdest_rnode_prefixes(rn, &p, &src_p); |
05737783 | 579 | |
40f321c0 | 580 | redistribute_delete(p, src_p, re, NULL); |
d62a17ae | 581 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); |
582 | } | |
718e3744 | 583 | } |
584 | ||
9fd92e3c AS |
585 | /* |
586 | * rib_can_delete_dest | |
587 | * | |
2951a7a4 | 588 | * Returns true if the given dest can be deleted from the table. |
9fd92e3c | 589 | */ |
d62a17ae | 590 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 591 | { |
045207e2 | 592 | if (re_list_first(&dest->routes)) { |
d62a17ae | 593 | return 0; |
594 | } | |
9fd92e3c | 595 | |
c86ba6c2 DS |
596 | /* |
597 | * Unresolved rnh's are stored on the default route's list | |
598 | * | |
599 | * dest->rnode can also be the source prefix node in an | |
600 | * ipv6 sourcedest table. Fortunately the prefix of a | |
601 | * source prefix node can never be the default prefix. | |
602 | */ | |
603 | if (is_default_prefix(&dest->rnode->p)) | |
604 | return 0; | |
605 | ||
d62a17ae | 606 | /* |
607 | * Don't delete the dest if we have to update the FPM about this | |
608 | * prefix. | |
609 | */ | |
610 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
611 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
612 | return 0; | |
5adc2528 | 613 | |
d62a17ae | 614 | return 1; |
9fd92e3c AS |
615 | } |
616 | ||
699dae23 DS |
617 | void zebra_rib_evaluate_rn_nexthops(struct route_node *rn, uint32_t seq) |
618 | { | |
619 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
699dae23 DS |
620 | struct rnh *rnh; |
621 | ||
622 | /* | |
50872b08 | 623 | * We are storing the rnh's associated withb |
699dae23 DS |
624 | * the tracked nexthop as a list of the rn's. |
625 | * Unresolved rnh's are placed at the top | |
626 | * of the tree list.( 0.0.0.0/0 for v4 and 0::0/0 for v6 ) | |
627 | * As such for each rn we need to walk up the tree | |
628 | * and see if any rnh's need to see if they | |
629 | * would match a more specific route | |
630 | */ | |
631 | while (rn) { | |
50872b08 DS |
632 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
633 | char buf[PREFIX_STRLEN]; | |
634 | ||
15569c58 DA |
635 | zlog_debug( |
636 | "%s: %s Being examined for Nexthop Tracking Count: %zd", | |
637 | __func__, | |
638 | srcdest_rnode2str(rn, buf, sizeof(buf)), | |
639 | dest ? rnh_list_count(&dest->nht) : 0); | |
50872b08 | 640 | } |
699dae23 DS |
641 | if (!dest) { |
642 | rn = rn->parent; | |
643 | if (rn) | |
644 | dest = rib_dest_from_rnode(rn); | |
645 | continue; | |
646 | } | |
647 | /* | |
648 | * If we have any rnh's stored in the nht list | |
649 | * then we know that this route node was used for | |
650 | * nht resolution and as such we need to call the | |
651 | * nexthop tracking evaluation code | |
652 | */ | |
739c9c90 | 653 | frr_each_safe(rnh_list, &dest->nht, rnh) { |
699dae23 DS |
654 | struct zebra_vrf *zvrf = |
655 | zebra_vrf_lookup_by_id(rnh->vrf_id); | |
656 | struct prefix *p = &rnh->node->p; | |
657 | ||
50872b08 | 658 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
699dae23 DS |
659 | char buf1[PREFIX_STRLEN]; |
660 | char buf2[PREFIX_STRLEN]; | |
661 | ||
0a7be328 | 662 | zlog_debug("%u:%s has Nexthop(%s) Type: %s depending on it, evaluating %u:%u", |
699dae23 | 663 | zvrf->vrf->vrf_id, |
50872b08 | 664 | srcdest_rnode2str(rn, buf1, |
699dae23 DS |
665 | sizeof(buf1)), |
666 | prefix2str(p, buf2, sizeof(buf2)), | |
0a7be328 | 667 | rnh_type2str(rnh->type), |
699dae23 DS |
668 | seq, rnh->seqno); |
669 | } | |
670 | ||
671 | /* | |
672 | * If we have evaluated this node on this pass | |
673 | * already, due to following the tree up | |
674 | * then we know that we can move onto the next | |
675 | * rnh to process. | |
676 | * | |
677 | * Additionally we call zebra_evaluate_rnh | |
678 | * when we gc the dest. In this case we know | |
679 | * that there must be no other re's where | |
680 | * we were originally as such we know that | |
681 | * that sequence number is ok to respect. | |
682 | */ | |
50872b08 DS |
683 | if (rnh->seqno == seq) { |
684 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) | |
685 | zlog_debug( | |
686 | "\tNode processed and moved already"); | |
699dae23 | 687 | continue; |
50872b08 | 688 | } |
699dae23 DS |
689 | |
690 | rnh->seqno = seq; | |
691 | zebra_evaluate_rnh(zvrf, family2afi(p->family), 0, | |
692 | rnh->type, p); | |
693 | } | |
694 | ||
695 | rn = rn->parent; | |
696 | if (rn) | |
697 | dest = rib_dest_from_rnode(rn); | |
698 | } | |
699 | } | |
700 | ||
9fd92e3c AS |
701 | /* |
702 | * rib_gc_dest | |
703 | * | |
704 | * Garbage collect the rib dest corresponding to the given route node | |
705 | * if appropriate. | |
706 | * | |
2951a7a4 | 707 | * Returns true if the dest was deleted, false otherwise. |
9fd92e3c | 708 | */ |
d62a17ae | 709 | int rib_gc_dest(struct route_node *rn) |
9fd92e3c | 710 | { |
d62a17ae | 711 | rib_dest_t *dest; |
9fd92e3c | 712 | |
d62a17ae | 713 | dest = rib_dest_from_rnode(rn); |
714 | if (!dest) | |
715 | return 0; | |
9fd92e3c | 716 | |
d62a17ae | 717 | if (!rib_can_delete_dest(dest)) |
718 | return 0; | |
9fd92e3c | 719 | |
c9abf558 DS |
720 | if (IS_ZEBRA_DEBUG_RIB) { |
721 | struct zebra_vrf *zvrf; | |
722 | ||
723 | zvrf = rib_dest_vrf(dest); | |
d62a17ae | 724 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); |
c9abf558 | 725 | } |
9fd92e3c | 726 | |
699dae23 DS |
727 | zebra_rib_evaluate_rn_nexthops(rn, zebra_router_get_next_sequence()); |
728 | ||
d62a17ae | 729 | dest->rnode = NULL; |
aa57abfb | 730 | rnh_list_fini(&dest->nht); |
d62a17ae | 731 | XFREE(MTYPE_RIB_DEST, dest); |
732 | rn->info = NULL; | |
9fd92e3c | 733 | |
d62a17ae | 734 | /* |
735 | * Release the one reference that we keep on the route node. | |
736 | */ | |
737 | route_unlock_node(rn); | |
738 | return 1; | |
9fd92e3c AS |
739 | } |
740 | ||
d62a17ae | 741 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
742 | struct route_entry *new) | |
3e5c6e00 | 743 | { |
d62a17ae | 744 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 745 | |
d62a17ae | 746 | /* Update real nexthop. This may actually determine if nexthop is active |
747 | * or not. */ | |
c415d895 | 748 | if (!nexthop_group_active_nexthop_num(&(new->nhe->nhg))) { |
d62a17ae | 749 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
750 | return; | |
751 | } | |
3e5c6e00 | 752 | |
d62a17ae | 753 | if (IS_ZEBRA_DEBUG_RIB) { |
754 | char buf[SRCDEST2STR_BUFFER]; | |
755 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
756 | zlog_debug("%u:%s: Adding route rn %p, re %p (%s)", |
757 | zvrf_id(zvrf), buf, rn, new, | |
758 | zebra_route_string(new->type)); | |
d62a17ae | 759 | } |
3e5c6e00 | 760 | |
d62a17ae | 761 | /* If labeled-unicast route, install transit LSP. */ |
762 | if (zebra_rib_labeled_unicast(new)) | |
763 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 764 | |
3cdba47a | 765 | rib_install_kernel(rn, new, NULL); |
3e5c6e00 | 766 | |
d62a17ae | 767 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 768 | } |
769 | ||
d62a17ae | 770 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
771 | struct route_entry *old) | |
3e5c6e00 | 772 | { |
d62a17ae | 773 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 774 | |
d62a17ae | 775 | /* Uninstall from kernel. */ |
776 | if (IS_ZEBRA_DEBUG_RIB) { | |
777 | char buf[SRCDEST2STR_BUFFER]; | |
778 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
779 | zlog_debug("%u:%s: Deleting route rn %p, re %p (%s)", |
780 | zvrf_id(zvrf), buf, rn, old, | |
781 | zebra_route_string(old->type)); | |
d62a17ae | 782 | } |
3e5c6e00 | 783 | |
d62a17ae | 784 | /* If labeled-unicast route, uninstall transit LSP. */ |
785 | if (zebra_rib_labeled_unicast(old)) | |
786 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
787 | ||
3cdba47a | 788 | rib_uninstall_kernel(rn, old); |
d62a17ae | 789 | |
790 | /* Update nexthop for route, reset changed flag. */ | |
ef57f35f DL |
791 | /* Note: this code also handles the Linux case when an interface goes |
792 | * down, causing the kernel to delete routes without sending DELROUTE | |
793 | * notifications | |
794 | */ | |
6883bf8d | 795 | if (RIB_KERNEL_ROUTE(old)) |
212df1de PG |
796 | SET_FLAG(old->status, ROUTE_ENTRY_REMOVED); |
797 | else | |
798 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 799 | } |
800 | ||
801 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
802 | struct route_node *rn, | |
803 | struct route_entry *old, | |
804 | struct route_entry *new) | |
805 | { | |
d62a17ae | 806 | int nh_active = 0; |
d62a17ae | 807 | |
808 | /* | |
809 | * We have to install or update if a new route has been selected or | |
810 | * something has changed. | |
811 | */ | |
812 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
813 | hook_call(rib_update, rn, "updating existing route"); | |
814 | ||
815 | /* Update the nexthop; we could determine here that nexthop is | |
816 | * inactive. */ | |
c415d895 | 817 | if (nexthop_group_active_nexthop_num(&(new->nhe->nhg))) |
d62a17ae | 818 | nh_active = 1; |
819 | ||
820 | /* If nexthop is active, install the selected route, if | |
821 | * appropriate. If | |
822 | * the install succeeds, cleanup flags for prior route, if | |
823 | * different from | |
824 | * newly selected. | |
825 | */ | |
826 | if (nh_active) { | |
827 | if (IS_ZEBRA_DEBUG_RIB) { | |
828 | char buf[SRCDEST2STR_BUFFER]; | |
829 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
830 | if (new != old) | |
831 | zlog_debug( | |
2da33d6b | 832 | "%u:%s: Updating route rn %p, re %p (%s) old %p (%s)", |
d62a17ae | 833 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
834 | zebra_route_string(new->type), |
835 | old, | |
836 | zebra_route_string(old->type)); | |
d62a17ae | 837 | else |
838 | zlog_debug( | |
2da33d6b | 839 | "%u:%s: Updating route rn %p, re %p (%s)", |
d62a17ae | 840 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 841 | zebra_route_string(new->type)); |
d62a17ae | 842 | } |
843 | ||
844 | /* If labeled-unicast route, uninstall transit LSP. */ | |
845 | if (zebra_rib_labeled_unicast(old)) | |
846 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
847 | ||
3cdba47a DS |
848 | /* |
849 | * Non-system route should be installed. | |
850 | * If labeled-unicast route, install transit | |
851 | * LSP. | |
852 | */ | |
853 | if (zebra_rib_labeled_unicast(new)) | |
854 | zebra_mpls_lsp_install(zvrf, rn, new); | |
d62a17ae | 855 | |
3cdba47a | 856 | rib_install_kernel(rn, new, old); |
d62a17ae | 857 | } |
a64448ba | 858 | |
d62a17ae | 859 | /* |
860 | * If nexthop for selected route is not active or install | |
861 | * failed, we | |
862 | * may need to uninstall and delete for redistribution. | |
863 | */ | |
ed216282 | 864 | if (!nh_active) { |
d62a17ae | 865 | if (IS_ZEBRA_DEBUG_RIB) { |
866 | char buf[SRCDEST2STR_BUFFER]; | |
867 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
868 | if (new != old) | |
869 | zlog_debug( | |
2da33d6b | 870 | "%u:%s: Deleting route rn %p, re %p (%s) old %p (%s) - nexthop inactive", |
d62a17ae | 871 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
872 | zebra_route_string(new->type), |
873 | old, | |
874 | zebra_route_string(old->type)); | |
d62a17ae | 875 | else |
876 | zlog_debug( | |
2da33d6b | 877 | "%u:%s: Deleting route rn %p, re %p (%s) - nexthop inactive", |
d62a17ae | 878 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 879 | zebra_route_string(new->type)); |
d62a17ae | 880 | } |
881 | ||
882 | /* If labeled-unicast route, uninstall transit LSP. */ | |
883 | if (zebra_rib_labeled_unicast(old)) | |
884 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
885 | ||
3cdba47a | 886 | rib_uninstall_kernel(rn, old); |
d62a17ae | 887 | } |
888 | } else { | |
889 | /* | |
890 | * Same route selected; check if in the FIB and if not, | |
3cdba47a DS |
891 | * re-install. This is housekeeping code to deal with |
892 | * race conditions in kernel with linux netlink reporting | |
893 | * interface up before IPv4 or IPv6 protocol is ready | |
d62a17ae | 894 | * to add routes. |
895 | */ | |
3cdba47a DS |
896 | if (!CHECK_FLAG(new->status, ROUTE_ENTRY_INSTALLED) || |
897 | RIB_SYSTEM_ROUTE(new)) | |
677c1dd5 | 898 | rib_install_kernel(rn, new, NULL); |
d62a17ae | 899 | } |
900 | ||
901 | /* Update prior route. */ | |
6883bf8d | 902 | if (new != old) |
d62a17ae | 903 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 904 | |
d62a17ae | 905 | /* Clear changed flag. */ |
906 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 907 | } |
908 | ||
d62a17ae | 909 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
910 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
911 | struct route_entry *alternate) | |
912 | { | |
913 | if (current == NULL) | |
914 | return alternate; | |
915 | ||
916 | /* filter route selection in following order: | |
917 | * - connected beats other types | |
fec4ca19 | 918 | * - if both connected, loopback or vrf wins |
d62a17ae | 919 | * - lower distance beats higher |
920 | * - lower metric beats higher for equal distance | |
921 | * - last, hence oldest, route wins tie break. | |
922 | */ | |
923 | ||
fec4ca19 DS |
924 | /* Connected routes. Check to see if either are a vrf |
925 | * or loopback interface. If not, pick the last connected | |
d62a17ae | 926 | * route of the set of lowest metric connected routes. |
927 | */ | |
928 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
fec4ca19 DS |
929 | if (current->type != ZEBRA_ROUTE_CONNECT) |
930 | return alternate; | |
931 | ||
932 | /* both are connected. are either loop or vrf? */ | |
933 | struct nexthop *nexthop = NULL; | |
934 | ||
c415d895 | 935 | for (ALL_NEXTHOPS(alternate->nhe->nhg, nexthop)) { |
873fde8c | 936 | struct interface *ifp = if_lookup_by_index( |
937 | nexthop->ifindex, alternate->vrf_id); | |
938 | ||
939 | if (ifp && if_is_loopback_or_vrf(ifp)) | |
fec4ca19 DS |
940 | return alternate; |
941 | } | |
942 | ||
c415d895 | 943 | for (ALL_NEXTHOPS(current->nhe->nhg, nexthop)) { |
873fde8c | 944 | struct interface *ifp = if_lookup_by_index( |
945 | nexthop->ifindex, current->vrf_id); | |
946 | ||
947 | if (ifp && if_is_loopback_or_vrf(ifp)) | |
fec4ca19 DS |
948 | return current; |
949 | } | |
950 | ||
951 | /* Neither are loop or vrf so pick best metric */ | |
952 | if (alternate->metric <= current->metric) | |
d62a17ae | 953 | return alternate; |
954 | ||
955 | return current; | |
956 | } | |
3e5c6e00 | 957 | |
d62a17ae | 958 | if (current->type == ZEBRA_ROUTE_CONNECT) |
959 | return current; | |
3e5c6e00 | 960 | |
d62a17ae | 961 | /* higher distance loses */ |
962 | if (alternate->distance < current->distance) | |
963 | return alternate; | |
964 | if (current->distance < alternate->distance) | |
965 | return current; | |
3e5c6e00 | 966 | |
d62a17ae | 967 | /* metric tie-breaks equal distance */ |
968 | if (alternate->metric <= current->metric) | |
969 | return alternate; | |
3e5c6e00 | 970 | |
d62a17ae | 971 | return current; |
3e5c6e00 | 972 | } |
973 | ||
e22e8001 SW |
974 | /* Core function for processing nexthop group contexts's off metaq */ |
975 | static void rib_nhg_process(struct nhg_ctx *ctx) | |
976 | { | |
977 | nhg_ctx_process(ctx); | |
978 | } | |
979 | ||
d62a17ae | 980 | /* Core function for processing routing information base. */ |
981 | static void rib_process(struct route_node *rn) | |
982 | { | |
983 | struct route_entry *re; | |
984 | struct route_entry *next; | |
985 | struct route_entry *old_selected = NULL; | |
986 | struct route_entry *new_selected = NULL; | |
987 | struct route_entry *old_fib = NULL; | |
988 | struct route_entry *new_fib = NULL; | |
989 | struct route_entry *best = NULL; | |
990 | char buf[SRCDEST2STR_BUFFER]; | |
991 | rib_dest_t *dest; | |
992 | struct zebra_vrf *zvrf = NULL; | |
86391e56 MS |
993 | const struct prefix *p, *src_p; |
994 | ||
d62a17ae | 995 | srcdest_rnode_prefixes(rn, &p, &src_p); |
996 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
997 | ||
998 | assert(rn); | |
999 | ||
1000 | dest = rib_dest_from_rnode(rn); | |
1001 | if (dest) { | |
1002 | zvrf = rib_dest_vrf(dest); | |
1003 | vrf_id = zvrf_id(zvrf); | |
1004 | } | |
bab85d4f | 1005 | |
d62a17ae | 1006 | if (IS_ZEBRA_DEBUG_RIB) |
1007 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1008 | |
d62a17ae | 1009 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1010 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1011 | |
607425e5 DS |
1012 | /* |
1013 | * we can have rn's that have a NULL info pointer | |
1014 | * (dest). As such let's not let the deref happen | |
1015 | * additionally we know RNODE_FOREACH_RE_SAFE | |
1016 | * will not iterate so we are ok. | |
1017 | */ | |
1018 | if (dest) | |
1019 | old_fib = dest->selected_fib; | |
5f7a4718 | 1020 | |
a2addae8 | 1021 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1022 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1023 | zlog_debug( | |
2da33d6b DS |
1024 | "%u:%s: Examine re %p (%s) status %x flags %x dist %d metric %d", |
1025 | vrf_id, buf, re, zebra_route_string(re->type), | |
1026 | re->status, re->flags, re->distance, | |
1027 | re->metric); | |
d62a17ae | 1028 | |
d62a17ae | 1029 | /* Currently selected re. */ |
1030 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1031 | assert(old_selected == NULL); | |
1032 | old_selected = re; | |
1033 | } | |
bab85d4f | 1034 | |
d62a17ae | 1035 | /* Skip deleted entries from selection */ |
1036 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1037 | continue; | |
1038 | ||
1039 | /* Skip unreachable nexthop. */ | |
1040 | /* This first call to nexthop_active_update is merely to | |
6883bf8d DS |
1041 | * determine if there's any change to nexthops associated |
1042 | * with this RIB entry. Now, rib_process() can be invoked due | |
1043 | * to an external event such as link down or due to | |
1044 | * next-hop-tracking evaluation. In the latter case, | |
d62a17ae | 1045 | * a decision has already been made that the NHs have changed. |
6883bf8d DS |
1046 | * So, no need to invoke a potentially expensive call again. |
1047 | * Further, since the change might be in a recursive NH which | |
1048 | * is not caught in the nexthop_active_update() code. Thus, we | |
1049 | * might miss changes to recursive NHs. | |
d62a17ae | 1050 | */ |
6883bf8d | 1051 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) |
99eabcec | 1052 | && !nexthop_active_update(rn, re)) { |
d62a17ae | 1053 | if (re->type == ZEBRA_ROUTE_TABLE) { |
1054 | /* XXX: HERE BE DRAGONS!!!!! | |
1055 | * In all honesty, I have not yet figured out | |
6883bf8d DS |
1056 | * what this part does or why the |
1057 | * ROUTE_ENTRY_CHANGED test above is correct | |
d62a17ae | 1058 | * or why we need to delete a route here, and |
6883bf8d DS |
1059 | * also not whether this concerns both selected |
1060 | * and fib route, or only selected | |
1061 | * or only fib | |
1062 | * | |
1063 | * This entry was denied by the 'ip protocol | |
1064 | * table' route-map, we need to delete it */ | |
d62a17ae | 1065 | if (re != old_selected) { |
1066 | if (IS_ZEBRA_DEBUG_RIB) | |
1067 | zlog_debug( | |
32391aff | 1068 | "%s: %u:%s: imported via import-table but denied " |
d62a17ae | 1069 | "by the ip protocol table route-map", |
32391aff | 1070 | __func__, vrf_id, buf); |
d62a17ae | 1071 | rib_unlink(rn, re); |
1072 | } else | |
1073 | SET_FLAG(re->status, | |
1074 | ROUTE_ENTRY_REMOVED); | |
1075 | } | |
1076 | ||
1077 | continue; | |
1078 | } | |
bab85d4f | 1079 | |
d62a17ae | 1080 | /* Infinite distance. */ |
3332f4f0 DS |
1081 | if (re->distance == DISTANCE_INFINITY && |
1082 | re->type != ZEBRA_ROUTE_KERNEL) { | |
d62a17ae | 1083 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); |
1084 | continue; | |
1085 | } | |
bab85d4f | 1086 | |
d62a17ae | 1087 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1088 | best = rib_choose_best(new_fib, re); | |
1089 | if (new_fib && best != new_fib) | |
1090 | UNSET_FLAG(new_fib->status, | |
1091 | ROUTE_ENTRY_CHANGED); | |
1092 | new_fib = best; | |
1093 | } else { | |
1094 | best = rib_choose_best(new_selected, re); | |
1095 | if (new_selected && best != new_selected) | |
1096 | UNSET_FLAG(new_selected->status, | |
1097 | ROUTE_ENTRY_CHANGED); | |
1098 | new_selected = best; | |
1099 | } | |
1100 | if (best != re) | |
1101 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1102 | } /* RNODE_FOREACH_RE */ | |
1103 | ||
1104 | /* If no FIB override route, use the selected route also for FIB */ | |
1105 | if (new_fib == NULL) | |
1106 | new_fib = new_selected; | |
1107 | ||
1108 | /* After the cycle is finished, the following pointers will be set: | |
1109 | * old_selected --- RE entry currently having SELECTED | |
1110 | * new_selected --- RE entry that is newly SELECTED | |
1111 | * old_fib --- RE entry currently in kernel FIB | |
1112 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1113 | * | |
1114 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1115 | * the zclients. new_fib (which can be new_selected) will be installed | |
1116 | * in kernel. | |
1117 | */ | |
1118 | ||
1119 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1120 | zlog_debug( | |
1121 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1122 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1123 | (void *)old_fib, (void *)new_fib); | |
1124 | } | |
446bb95e | 1125 | |
d62a17ae | 1126 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1127 | * fib == selected */ | |
9d303b37 DL |
1128 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, |
1129 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1130 | |
1131 | /* Update fib according to selection results */ | |
1132 | if (new_fib && old_fib) | |
1133 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1134 | else if (new_fib) | |
1135 | rib_process_add_fib(zvrf, rn, new_fib); | |
1136 | else if (old_fib) | |
1137 | rib_process_del_fib(zvrf, rn, old_fib); | |
1138 | ||
8cb41cd6 | 1139 | /* Update SELECTED entry */ |
d62a17ae | 1140 | if (old_selected != new_selected || selected_changed) { |
93bdadae | 1141 | |
6883bf8d | 1142 | if (new_selected && new_selected != new_fib) |
d62a17ae | 1143 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); |
41ec9222 | 1144 | |
3cdba47a | 1145 | if (new_selected) |
5af4b346 MS |
1146 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); |
1147 | ||
d62a17ae | 1148 | if (old_selected) { |
40f321c0 MS |
1149 | /* |
1150 | * If we're removing the old entry, we should tell | |
1151 | * redist subscribers about that *if* they aren't | |
1152 | * going to see a redist for the new entry. | |
1153 | */ | |
1154 | if (!new_selected || CHECK_FLAG(old_selected->status, | |
1155 | ROUTE_ENTRY_REMOVED)) | |
1156 | redistribute_delete(p, src_p, | |
1157 | old_selected, | |
1158 | new_selected); | |
1159 | ||
d62a17ae | 1160 | if (old_selected != new_selected) |
1161 | UNSET_FLAG(old_selected->flags, | |
1162 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1163 | } |
d62a17ae | 1164 | } |
3e5c6e00 | 1165 | |
d62a17ae | 1166 | /* Remove all RE entries queued for removal */ |
a2addae8 | 1167 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1168 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { |
1169 | if (IS_ZEBRA_DEBUG_RIB) { | |
1170 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1171 | (void *)rn, (void *)re); | |
1172 | } | |
1173 | rib_unlink(rn, re); | |
1174 | } | |
1175 | } | |
4d38fdb4 | 1176 | |
d62a17ae | 1177 | /* |
1178 | * Check if the dest can be deleted now. | |
1179 | */ | |
1180 | rib_gc_dest(rn); | |
e96f9203 DO |
1181 | } |
1182 | ||
a1494c25 DS |
1183 | static void zebra_rib_evaluate_mpls(struct route_node *rn) |
1184 | { | |
1185 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
1186 | struct zebra_vrf *zvrf = vrf_info_lookup(VRF_DEFAULT); | |
1187 | ||
1188 | if (!dest) | |
1189 | return; | |
1190 | ||
1191 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_LSPS)) { | |
1192 | if (IS_ZEBRA_DEBUG_MPLS) | |
1193 | zlog_debug( | |
1194 | "%u: Scheduling all LSPs upon RIB completion", | |
1195 | zvrf_id(zvrf)); | |
1196 | zebra_mpls_lsp_schedule(zvrf); | |
1197 | mpls_unmark_lsps_for_processing(rn); | |
1198 | } | |
1199 | } | |
1200 | ||
e5ac2adf MS |
1201 | /* |
1202 | * Utility to match route with dplane context data | |
1203 | */ | |
1204 | static bool rib_route_match_ctx(const struct route_entry *re, | |
25779064 MS |
1205 | const struct zebra_dplane_ctx *ctx, |
1206 | bool is_update) | |
e5ac2adf MS |
1207 | { |
1208 | bool result = false; | |
1209 | ||
1210 | if (is_update) { | |
1211 | /* | |
1212 | * In 'update' case, we test info about the 'previous' or | |
1213 | * 'old' route | |
1214 | */ | |
1215 | if ((re->type == dplane_ctx_get_old_type(ctx)) && | |
1216 | (re->instance == dplane_ctx_get_old_instance(ctx))) { | |
1217 | result = true; | |
1218 | ||
1219 | /* TODO -- we're using this extra test, but it's not | |
1220 | * exactly clear why. | |
1221 | */ | |
1222 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1223 | (re->distance != dplane_ctx_get_old_distance(ctx) || | |
1224 | re->tag != dplane_ctx_get_old_tag(ctx))) { | |
1225 | result = false; | |
1226 | } | |
1227 | } | |
1228 | ||
1229 | } else { | |
1230 | /* | |
1231 | * Ordinary, single-route case using primary context info | |
1232 | */ | |
1233 | if ((dplane_ctx_get_op(ctx) != DPLANE_OP_ROUTE_DELETE) && | |
1234 | CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1235 | /* Skip route that's been deleted */ | |
1236 | goto done; | |
1237 | } | |
1238 | ||
1239 | if ((re->type == dplane_ctx_get_type(ctx)) && | |
1240 | (re->instance == dplane_ctx_get_instance(ctx))) { | |
1241 | result = true; | |
1242 | ||
1243 | /* TODO -- we're using this extra test, but it's not | |
1244 | * exactly clear why. | |
1245 | */ | |
1246 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1247 | (re->distance != dplane_ctx_get_distance(ctx) || | |
1248 | re->tag != dplane_ctx_get_tag(ctx))) { | |
1249 | result = false; | |
1250 | } | |
1251 | } | |
1252 | } | |
1253 | ||
1254 | done: | |
1255 | ||
1256 | return (result); | |
1257 | } | |
1258 | ||
3cdba47a DS |
1259 | static void zebra_rib_fixup_system(struct route_node *rn) |
1260 | { | |
1261 | struct route_entry *re; | |
1262 | ||
1263 | RNODE_FOREACH_RE(rn, re) { | |
1264 | struct nexthop *nhop; | |
1265 | ||
1266 | if (!RIB_SYSTEM_ROUTE(re)) | |
1267 | continue; | |
1268 | ||
1269 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1270 | continue; | |
1271 | ||
1272 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
6d0ee6a0 | 1273 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); |
3cdba47a | 1274 | |
c415d895 | 1275 | for (ALL_NEXTHOPS(re->nhe->nhg, nhop)) { |
3cdba47a DS |
1276 | if (CHECK_FLAG(nhop->flags, NEXTHOP_FLAG_RECURSIVE)) |
1277 | continue; | |
1278 | ||
1279 | SET_FLAG(nhop->flags, NEXTHOP_FLAG_FIB); | |
1280 | } | |
1281 | } | |
1282 | } | |
1283 | ||
e5ac2adf | 1284 | /* |
78bf56b0 MS |
1285 | * Update a route from a dplane context. This consolidates common code |
1286 | * that can be used in processing of results from FIB updates, and in | |
1287 | * async notification processing. | |
efe6c026 | 1288 | * The return is 'true' if the installed nexthops changed; 'false' otherwise. |
e5ac2adf | 1289 | */ |
efe6c026 MS |
1290 | static bool rib_update_re_from_ctx(struct route_entry *re, |
1291 | struct route_node *rn, | |
1292 | struct zebra_dplane_ctx *ctx) | |
78bf56b0 | 1293 | { |
941e261c | 1294 | char dest_str[PREFIX_STRLEN] = ""; |
efe6c026 | 1295 | char nh_str[NEXTHOP_STRLEN]; |
78bf56b0 | 1296 | struct nexthop *nexthop, *ctx_nexthop; |
941e261c MS |
1297 | bool matched; |
1298 | const struct nexthop_group *ctxnhg; | |
efe6c026 MS |
1299 | bool is_selected = false; /* Is 're' currently the selected re? */ |
1300 | bool changed_p = false; /* Change to nexthops? */ | |
1301 | rib_dest_t *dest; | |
78bf56b0 | 1302 | |
941e261c MS |
1303 | /* Note well: only capturing the prefix string if debug is enabled here; |
1304 | * unconditional log messages will have to generate the string. | |
1305 | */ | |
1306 | if (IS_ZEBRA_DEBUG_RIB) | |
1307 | prefix2str(&(rn->p), dest_str, sizeof(dest_str)); | |
78bf56b0 | 1308 | |
efe6c026 MS |
1309 | dest = rib_dest_from_rnode(rn); |
1310 | if (dest) | |
1311 | is_selected = (re == dest->selected_fib); | |
1312 | ||
1313 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1314 | zlog_debug("update_from_ctx: %u:%s: %sSELECTED", | |
1315 | re->vrf_id, dest_str, (is_selected ? "" : "NOT ")); | |
1316 | ||
1317 | /* Update zebra's nexthop FIB flag for each nexthop that was installed. | |
941e261c | 1318 | * If the installed set differs from the set requested by the rib/owner, |
efe6c026 MS |
1319 | * we use the fib-specific nexthop-group to record the actual FIB |
1320 | * status. | |
78bf56b0 | 1321 | */ |
941e261c | 1322 | |
986a6617 SW |
1323 | /* Check both fib group and notif group for equivalence. |
1324 | * | |
1325 | * Let's assume the nexthops are ordered here to save time. | |
941e261c | 1326 | */ |
986a6617 SW |
1327 | if (nexthop_group_equal(&re->fib_ng, dplane_ctx_get_ng(ctx)) == false) { |
1328 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1329 | zlog_debug( | |
1330 | "%u:%s update_from_ctx: notif nh and fib nh mismatch", | |
1331 | re->vrf_id, dest_str); | |
941e261c MS |
1332 | } |
1333 | ||
986a6617 SW |
1334 | matched = false; |
1335 | } else | |
1336 | matched = true; | |
941e261c MS |
1337 | |
1338 | /* If the new FIB set matches the existing FIB set, we're done. */ | |
1339 | if (matched) { | |
1340 | if (IS_ZEBRA_DEBUG_RIB) | |
1341 | zlog_debug("%u:%s update_from_ctx(): existing fib nhg, no change", | |
1342 | re->vrf_id, dest_str); | |
1343 | goto done; | |
1344 | ||
1345 | } else if (re->fib_ng.nexthop) { | |
1346 | /* | |
1347 | * Free stale fib list and move on to check the rib nhg. | |
1348 | */ | |
1349 | if (IS_ZEBRA_DEBUG_RIB) | |
1350 | zlog_debug("%u:%s update_from_ctx(): replacing fib nhg", | |
1351 | re->vrf_id, dest_str); | |
1352 | nexthops_free(re->fib_ng.nexthop); | |
1353 | re->fib_ng.nexthop = NULL; | |
efe6c026 MS |
1354 | |
1355 | /* Note that the installed nexthops have changed */ | |
1356 | changed_p = true; | |
941e261c MS |
1357 | } else { |
1358 | if (IS_ZEBRA_DEBUG_RIB) | |
1359 | zlog_debug("%u:%s update_from_ctx(): no fib nhg", | |
1360 | re->vrf_id, dest_str); | |
1361 | } | |
1362 | ||
1363 | /* | |
1364 | * Compare with the rib nexthop group. The comparison here is different: | |
1365 | * the RIB group may be a superset of the list installed in the FIB. We | |
1366 | * walk the RIB group, looking for the 'installable' candidate | |
1367 | * nexthops, and then check those against the set | |
1368 | * that is actually installed. | |
986a6617 SW |
1369 | * |
1370 | * Assume nexthops are ordered here as well. | |
941e261c MS |
1371 | */ |
1372 | matched = true; | |
986a6617 SW |
1373 | |
1374 | ctx_nexthop = dplane_ctx_get_ng(ctx)->nexthop; | |
1375 | ||
1f6a5aca MS |
1376 | /* Nothing installed - we can skip some of the checking/comparison |
1377 | * of nexthops. | |
1378 | */ | |
1379 | if (ctx_nexthop == NULL) { | |
1380 | changed_p = true; | |
1381 | goto no_nexthops; | |
1382 | } | |
1383 | ||
986a6617 SW |
1384 | /* Get the first `installed` one to check against. |
1385 | * If the dataplane doesn't set these to be what was actually installed, | |
0eb97b86 | 1386 | * it will just be whatever was in re->nhe->nhg? |
986a6617 SW |
1387 | */ |
1388 | if (CHECK_FLAG(ctx_nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
1389 | || !CHECK_FLAG(ctx_nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
1390 | ctx_nexthop = nexthop_next_active_resolved(ctx_nexthop); | |
1391 | ||
c415d895 | 1392 | for (ALL_NEXTHOPS(re->nhe->nhg, nexthop)) { |
78bf56b0 | 1393 | |
5695d9ac | 1394 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) |
78bf56b0 MS |
1395 | continue; |
1396 | ||
941e261c MS |
1397 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) |
1398 | continue; | |
5695d9ac | 1399 | |
941e261c | 1400 | /* Check for a FIB nexthop corresponding to the RIB nexthop */ |
986a6617 SW |
1401 | if (nexthop_same(ctx_nexthop, nexthop) == false) { |
1402 | /* If the FIB doesn't know about the nexthop, | |
1403 | * it's not installed | |
1404 | */ | |
efe6c026 MS |
1405 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1406 | nexthop2str(nexthop, nh_str, sizeof(nh_str)); | |
986a6617 SW |
1407 | zlog_debug( |
1408 | "update_from_ctx: no notif match for rib nh %s", | |
1409 | nh_str); | |
efe6c026 | 1410 | } |
941e261c | 1411 | matched = false; |
efe6c026 MS |
1412 | |
1413 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1414 | changed_p = true; | |
1415 | ||
1416 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
11260e70 MS |
1417 | |
1418 | /* Keep checking nexthops */ | |
1419 | continue; | |
5695d9ac | 1420 | } |
78bf56b0 | 1421 | |
efe6c026 MS |
1422 | if (CHECK_FLAG(ctx_nexthop->flags, NEXTHOP_FLAG_FIB)) { |
1423 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1424 | changed_p = true; | |
1425 | ||
5695d9ac | 1426 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
efe6c026 MS |
1427 | } else { |
1428 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1429 | changed_p = true; | |
1430 | ||
5695d9ac | 1431 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
efe6c026 | 1432 | } |
986a6617 SW |
1433 | |
1434 | ctx_nexthop = nexthop_next_active_resolved(ctx_nexthop); | |
78bf56b0 MS |
1435 | } |
1436 | ||
941e261c MS |
1437 | /* If all nexthops were processed, we're done */ |
1438 | if (matched) { | |
1439 | if (IS_ZEBRA_DEBUG_RIB) | |
efe6c026 MS |
1440 | zlog_debug("%u:%s update_from_ctx(): rib nhg matched, changed '%s'", |
1441 | re->vrf_id, dest_str, | |
1442 | (changed_p ? "true" : "false")); | |
941e261c MS |
1443 | goto done; |
1444 | } | |
78bf56b0 | 1445 | |
1f6a5aca MS |
1446 | no_nexthops: |
1447 | ||
941e261c MS |
1448 | /* FIB nexthop set differs from the RIB set: |
1449 | * create a fib-specific nexthop-group | |
1450 | */ | |
1451 | if (IS_ZEBRA_DEBUG_RIB) | |
efe6c026 MS |
1452 | zlog_debug("%u:%s update_from_ctx(): changed %s, adding new fib nhg", |
1453 | re->vrf_id, dest_str, | |
1454 | (changed_p ? "true" : "false")); | |
941e261c MS |
1455 | |
1456 | ctxnhg = dplane_ctx_get_ng(ctx); | |
188a00e0 MS |
1457 | |
1458 | if (ctxnhg->nexthop) | |
1459 | copy_nexthops(&(re->fib_ng.nexthop), ctxnhg->nexthop, NULL); | |
1460 | else { | |
1461 | /* Bit of a special case when the fib has _no_ installed | |
1462 | * nexthops. | |
1463 | */ | |
1464 | nexthop = nexthop_new(); | |
1465 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
50d89650 | 1466 | _nexthop_add(&(re->fib_ng.nexthop), nexthop); |
188a00e0 | 1467 | } |
941e261c MS |
1468 | |
1469 | done: | |
efe6c026 | 1470 | return changed_p; |
78bf56b0 MS |
1471 | } |
1472 | ||
e5ac2adf | 1473 | /* |
fad4d69c MS |
1474 | * Helper to locate a zebra route-node from a dplane context. This is used |
1475 | * when processing dplane results, e.g. Note well: the route-node is returned | |
1476 | * with a ref held - route_unlock_node() must be called eventually. | |
e5ac2adf | 1477 | */ |
fad4d69c MS |
1478 | static struct route_node * |
1479 | rib_find_rn_from_ctx(const struct zebra_dplane_ctx *ctx) | |
e5ac2adf MS |
1480 | { |
1481 | struct route_table *table = NULL; | |
1482 | struct route_node *rn = NULL; | |
e5ac2adf MS |
1483 | const struct prefix *dest_pfx, *src_pfx; |
1484 | ||
1485 | /* Locate rn and re(s) from ctx */ | |
1486 | ||
c7c0b007 SW |
1487 | table = zebra_vrf_lookup_table_with_table_id( |
1488 | dplane_ctx_get_afi(ctx), dplane_ctx_get_safi(ctx), | |
1489 | dplane_ctx_get_vrf(ctx), dplane_ctx_get_table(ctx)); | |
e5ac2adf MS |
1490 | if (table == NULL) { |
1491 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fad4d69c | 1492 | zlog_debug("Failed to find route for ctx: no table for afi %d, safi %d, vrf %u", |
e5ac2adf MS |
1493 | dplane_ctx_get_afi(ctx), |
1494 | dplane_ctx_get_safi(ctx), | |
1495 | dplane_ctx_get_vrf(ctx)); | |
1496 | } | |
1497 | goto done; | |
1498 | } | |
1499 | ||
fad4d69c MS |
1500 | dest_pfx = dplane_ctx_get_dest(ctx); |
1501 | src_pfx = dplane_ctx_get_src(ctx); | |
1502 | ||
1503 | rn = srcdest_rnode_get(table, dest_pfx, | |
1504 | src_pfx ? (struct prefix_ipv6 *)src_pfx : NULL); | |
1505 | ||
1506 | done: | |
1507 | return rn; | |
1508 | } | |
1509 | ||
1510 | ||
97f5b441 | 1511 | |
fad4d69c MS |
1512 | /* |
1513 | * Route-update results processing after async dataplane update. | |
1514 | */ | |
1515 | static void rib_process_result(struct zebra_dplane_ctx *ctx) | |
1516 | { | |
1517 | struct zebra_vrf *zvrf = NULL; | |
1518 | struct route_node *rn = NULL; | |
1519 | struct route_entry *re = NULL, *old_re = NULL, *rib; | |
1520 | bool is_update = false; | |
1521 | char dest_str[PREFIX_STRLEN] = ""; | |
1522 | enum dplane_op_e op; | |
1523 | enum zebra_dplane_result status; | |
1524 | const struct prefix *dest_pfx, *src_pfx; | |
1525 | uint32_t seq; | |
efe6c026 | 1526 | bool fib_changed = false; |
97f5b441 | 1527 | |
fad4d69c | 1528 | zvrf = vrf_info_lookup(dplane_ctx_get_vrf(ctx)); |
e5ac2adf MS |
1529 | dest_pfx = dplane_ctx_get_dest(ctx); |
1530 | ||
1531 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1532 | * unconditional log messages will have to generate the string. | |
1533 | */ | |
5709131c | 1534 | if (IS_ZEBRA_DEBUG_DPLANE) |
e5ac2adf | 1535 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
e5ac2adf | 1536 | |
fad4d69c MS |
1537 | /* Locate rn and re(s) from ctx */ |
1538 | rn = rib_find_rn_from_ctx(ctx); | |
e5ac2adf MS |
1539 | if (rn == NULL) { |
1540 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fe2c53d4 | 1541 | zlog_debug("Failed to process dplane results: no route for %u:%s", |
e5ac2adf MS |
1542 | dplane_ctx_get_vrf(ctx), dest_str); |
1543 | } | |
1544 | goto done; | |
1545 | } | |
1546 | ||
1547 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); | |
1548 | ||
1549 | op = dplane_ctx_get_op(ctx); | |
1550 | status = dplane_ctx_get_status(ctx); | |
1551 | ||
c831033f | 1552 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
f183e380 | 1553 | zlog_debug("%u:%s Processing dplane ctx %p, op %s result %s", |
e5ac2adf | 1554 | dplane_ctx_get_vrf(ctx), dest_str, ctx, |
f183e380 | 1555 | dplane_op2str(op), dplane_res2str(status)); |
e5ac2adf | 1556 | |
e5ac2adf MS |
1557 | /* |
1558 | * Update is a bit of a special case, where we may have both old and new | |
1559 | * routes to post-process. | |
1560 | */ | |
1561 | is_update = dplane_ctx_is_update(ctx); | |
1562 | ||
1563 | /* | |
1564 | * Take a pass through the routes, look for matches with the context | |
1565 | * info. | |
1566 | */ | |
1567 | RNODE_FOREACH_RE(rn, rib) { | |
1568 | ||
1569 | if (re == NULL) { | |
5709131c | 1570 | if (rib_route_match_ctx(rib, ctx, false)) |
e5ac2adf | 1571 | re = rib; |
e5ac2adf MS |
1572 | } |
1573 | ||
1574 | /* Check for old route match */ | |
1575 | if (is_update && (old_re == NULL)) { | |
5709131c | 1576 | if (rib_route_match_ctx(rib, ctx, true /*is_update*/)) |
e5ac2adf | 1577 | old_re = rib; |
e5ac2adf MS |
1578 | } |
1579 | ||
1580 | /* Have we found the routes we need to work on? */ | |
5709131c | 1581 | if (re && ((!is_update || old_re))) |
e5ac2adf | 1582 | break; |
e5ac2adf MS |
1583 | } |
1584 | ||
699dae23 DS |
1585 | seq = dplane_ctx_get_seq(ctx); |
1586 | ||
e5ac2adf MS |
1587 | /* |
1588 | * Check sequence number(s) to detect stale results before continuing | |
1589 | */ | |
60f98b23 | 1590 | if (re) { |
699dae23 | 1591 | if (re->dplane_sequence != seq) { |
60f98b23 DS |
1592 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
1593 | zlog_debug("%u:%s Stale dplane result for re %p", | |
1594 | dplane_ctx_get_vrf(ctx), | |
1595 | dest_str, re); | |
1596 | } else | |
1597 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); | |
e5ac2adf MS |
1598 | } |
1599 | ||
60f98b23 DS |
1600 | if (old_re) { |
1601 | if (old_re->dplane_sequence != dplane_ctx_get_old_seq(ctx)) { | |
1602 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
1603 | zlog_debug("%u:%s Stale dplane result for old_re %p", | |
1604 | dplane_ctx_get_vrf(ctx), | |
1605 | dest_str, old_re); | |
1606 | } else | |
9ef0c6ba | 1607 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_QUEUED); |
e5ac2adf MS |
1608 | } |
1609 | ||
12e7fe3a | 1610 | switch (op) { |
12e7fe3a DS |
1611 | case DPLANE_OP_ROUTE_INSTALL: |
1612 | case DPLANE_OP_ROUTE_UPDATE: | |
1613 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
677c1dd5 DS |
1614 | if (re) { |
1615 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
1616 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1617 | } | |
b9f0e5ee DS |
1618 | /* |
1619 | * On an update operation from the same route type | |
1620 | * context retrieval currently has no way to know | |
1621 | * which was the old and which was the new. | |
1622 | * So don't unset our flags that we just set. | |
1623 | * We know redistribution is ok because the | |
1624 | * old_re in this case is used for nothing | |
1625 | * more than knowing whom to contact if necessary. | |
1626 | */ | |
1627 | if (old_re && old_re != re) { | |
677c1dd5 DS |
1628 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
1629 | UNSET_FLAG(old_re->status, | |
1630 | ROUTE_ENTRY_INSTALLED); | |
1631 | } | |
f183e380 | 1632 | |
78bf56b0 | 1633 | /* Update zebra route based on the results in |
941e261c | 1634 | * the context struct. |
78bf56b0 | 1635 | */ |
941e261c | 1636 | if (re) { |
efe6c026 MS |
1637 | fib_changed = |
1638 | rib_update_re_from_ctx(re, rn, ctx); | |
1639 | ||
1640 | if (!fib_changed) { | |
1641 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
1642 | zlog_debug("%u:%s no fib change for re", | |
1643 | dplane_ctx_get_vrf( | |
1644 | ctx), | |
1645 | dest_str); | |
12e7fe3a DS |
1646 | } |
1647 | ||
941e261c MS |
1648 | /* Redistribute */ |
1649 | redistribute_update(dest_pfx, src_pfx, | |
228a811a | 1650 | re, old_re); |
f183e380 MS |
1651 | } |
1652 | ||
3cdba47a DS |
1653 | /* |
1654 | * System routes are weird in that they | |
1655 | * allow multiple to be installed that match | |
1656 | * to the same prefix, so after we get the | |
1657 | * result we need to clean them up so that | |
1658 | * we can actually use them. | |
1659 | */ | |
1660 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
1661 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
1662 | zebra_rib_fixup_system(rn); | |
1663 | ||
f4c6e2a8 | 1664 | if (zvrf) |
12e7fe3a | 1665 | zvrf->installs++; |
f183e380 | 1666 | |
12e7fe3a | 1667 | /* Notify route owner */ |
677c1dd5 | 1668 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_INSTALLED); |
e5ac2adf | 1669 | |
12e7fe3a | 1670 | } else { |
7a230a9d | 1671 | if (re) { |
677c1dd5 | 1672 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); |
7a230a9d DS |
1673 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
1674 | } if (old_re) | |
677c1dd5 DS |
1675 | SET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
1676 | if (re) | |
1677 | zsend_route_notify_owner(re, dest_pfx, | |
1678 | ZAPI_ROUTE_FAIL_INSTALL); | |
97f5b441 | 1679 | |
12e7fe3a DS |
1680 | zlog_warn("%u:%s: Route install failed", |
1681 | dplane_ctx_get_vrf(ctx), | |
1682 | prefix2str(dest_pfx, | |
1683 | dest_str, sizeof(dest_str))); | |
1684 | } | |
1685 | break; | |
1686 | case DPLANE_OP_ROUTE_DELETE: | |
677c1dd5 DS |
1687 | if (re) |
1688 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
1689 | /* |
1690 | * In the delete case, the zebra core datastructs were | |
1691 | * updated (or removed) at the time the delete was issued, | |
1692 | * so we're just notifying the route owner. | |
e5ac2adf | 1693 | */ |
12e7fe3a | 1694 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { |
677c1dd5 DS |
1695 | if (re) { |
1696 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1697 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
1698 | } | |
12e7fe3a | 1699 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_REMOVED); |
e5ac2adf | 1700 | |
12e7fe3a DS |
1701 | if (zvrf) |
1702 | zvrf->removals++; | |
1703 | } else { | |
677c1dd5 DS |
1704 | if (re) |
1705 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
1706 | zsend_route_notify_owner_ctx(ctx, |
1707 | ZAPI_ROUTE_REMOVE_FAIL); | |
e5ac2adf | 1708 | |
12e7fe3a DS |
1709 | zlog_warn("%u:%s: Route Deletion failure", |
1710 | dplane_ctx_get_vrf(ctx), | |
1711 | prefix2str(dest_pfx, | |
1712 | dest_str, sizeof(dest_str))); | |
1713 | } | |
3cdba47a DS |
1714 | |
1715 | /* | |
1716 | * System routes are weird in that they | |
1717 | * allow multiple to be installed that match | |
1718 | * to the same prefix, so after we get the | |
1719 | * result we need to clean them up so that | |
1720 | * we can actually use them. | |
1721 | */ | |
1722 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
1723 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
1724 | zebra_rib_fixup_system(rn); | |
12e7fe3a | 1725 | break; |
d37f4d6c MS |
1726 | default: |
1727 | break; | |
e5ac2adf | 1728 | } |
699dae23 DS |
1729 | |
1730 | zebra_rib_evaluate_rn_nexthops(rn, seq); | |
a1494c25 | 1731 | zebra_rib_evaluate_mpls(rn); |
e5ac2adf MS |
1732 | done: |
1733 | ||
5f27bcba DS |
1734 | if (rn) |
1735 | route_unlock_node(rn); | |
1736 | ||
e5ac2adf MS |
1737 | /* Return context to dataplane module */ |
1738 | dplane_ctx_fini(&ctx); | |
1739 | } | |
1740 | ||
54818e3b MS |
1741 | /* |
1742 | * Handle notification from async dataplane: the dataplane has detected | |
1743 | * some change to a route, and notifies zebra so that the control plane | |
1744 | * can reflect that change. | |
1745 | */ | |
1746 | static void rib_process_dplane_notify(struct zebra_dplane_ctx *ctx) | |
1747 | { | |
1748 | struct route_node *rn = NULL; | |
efe6c026 MS |
1749 | struct route_entry *re = NULL; |
1750 | struct nexthop *nexthop; | |
54818e3b MS |
1751 | char dest_str[PREFIX_STRLEN] = ""; |
1752 | const struct prefix *dest_pfx, *src_pfx; | |
941e261c | 1753 | rib_dest_t *dest; |
efe6c026 | 1754 | bool fib_changed = false; |
941e261c | 1755 | bool debug_p = IS_ZEBRA_DEBUG_DPLANE | IS_ZEBRA_DEBUG_RIB; |
efe6c026 | 1756 | int start_count, end_count; |
54818e3b MS |
1757 | dest_pfx = dplane_ctx_get_dest(ctx); |
1758 | ||
1759 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1760 | * unconditional log messages will have to generate the string. | |
1761 | */ | |
941e261c | 1762 | if (debug_p) |
54818e3b MS |
1763 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
1764 | ||
1765 | /* Locate rn and re(s) from ctx */ | |
1766 | rn = rib_find_rn_from_ctx(ctx); | |
1767 | if (rn == NULL) { | |
941e261c MS |
1768 | if (debug_p) { |
1769 | zlog_debug("Failed to process dplane notification: no routes for %u:%s", | |
54818e3b MS |
1770 | dplane_ctx_get_vrf(ctx), dest_str); |
1771 | } | |
1772 | goto done; | |
1773 | } | |
1774 | ||
941e261c | 1775 | dest = rib_dest_from_rnode(rn); |
54818e3b MS |
1776 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); |
1777 | ||
efe6c026 | 1778 | if (debug_p) |
54818e3b MS |
1779 | zlog_debug("%u:%s Processing dplane notif ctx %p", |
1780 | dplane_ctx_get_vrf(ctx), dest_str, ctx); | |
1781 | ||
1782 | /* | |
1783 | * Take a pass through the routes, look for matches with the context | |
1784 | * info. | |
1785 | */ | |
efe6c026 MS |
1786 | RNODE_FOREACH_RE(rn, re) { |
1787 | if (rib_route_match_ctx(re, ctx, false /*!update*/)) | |
54818e3b MS |
1788 | break; |
1789 | } | |
1790 | ||
1791 | /* No match? Nothing we can do */ | |
1792 | if (re == NULL) { | |
941e261c MS |
1793 | if (debug_p) |
1794 | zlog_debug("%u:%s Unable to process dplane notification: no entry for type %s", | |
54818e3b | 1795 | dplane_ctx_get_vrf(ctx), dest_str, |
941e261c MS |
1796 | zebra_route_string( |
1797 | dplane_ctx_get_type(ctx))); | |
54818e3b MS |
1798 | |
1799 | goto done; | |
1800 | } | |
1801 | ||
272e8903 MS |
1802 | /* Ensure we clear the QUEUED flag */ |
1803 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); | |
1804 | ||
4c0b5436 MS |
1805 | /* Is this a notification that ... matters? We mostly care about |
1806 | * the route that is currently selected for installation; we may also | |
1807 | * get an un-install notification, and handle that too. | |
941e261c MS |
1808 | */ |
1809 | if (re != dest->selected_fib) { | |
4c0b5436 MS |
1810 | /* |
1811 | * If we need to, clean up after a delete that was part of | |
1812 | * an update operation. | |
941e261c | 1813 | */ |
4c0b5436 MS |
1814 | end_count = 0; |
1815 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), nexthop)) { | |
1816 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1817 | end_count++; | |
1818 | } | |
1819 | ||
1820 | /* If no nexthops or none installed, ensure that this re | |
1821 | * gets its 'installed' flag cleared. | |
1822 | */ | |
1823 | if (end_count == 0) { | |
1824 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED)) | |
1825 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1826 | if (debug_p) | |
1827 | zlog_debug("%u:%s dplane notif, uninstalled type %s route", | |
1828 | dplane_ctx_get_vrf(ctx), dest_str, | |
1829 | zebra_route_string( | |
1830 | dplane_ctx_get_type(ctx))); | |
1831 | } else { | |
1832 | /* At least report on the event. */ | |
1833 | if (debug_p) | |
1834 | zlog_debug("%u:%s dplane notif, but type %s not selected_fib", | |
1835 | dplane_ctx_get_vrf(ctx), dest_str, | |
1836 | zebra_route_string( | |
1837 | dplane_ctx_get_type(ctx))); | |
1838 | } | |
941e261c MS |
1839 | goto done; |
1840 | } | |
1841 | ||
efe6c026 MS |
1842 | /* We'll want to determine whether the installation status of the |
1843 | * route has changed: we'll check the status before processing, | |
1844 | * and then again if there's been a change. | |
1845 | */ | |
1846 | start_count = 0; | |
1c30d64b MS |
1847 | |
1848 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED)) { | |
1849 | for (ALL_NEXTHOPS_PTR(rib_active_nhg(re), nexthop)) { | |
1850 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1851 | start_count++; | |
1852 | } | |
efe6c026 MS |
1853 | } |
1854 | ||
54818e3b MS |
1855 | /* Update zebra's nexthop FIB flags based on the context struct's |
1856 | * nexthops. | |
1857 | */ | |
efe6c026 | 1858 | fib_changed = rib_update_re_from_ctx(re, rn, ctx); |
54818e3b | 1859 | |
efe6c026 MS |
1860 | if (!fib_changed) { |
1861 | if (debug_p) | |
1c30d64b | 1862 | zlog_debug("%u:%s dplane notification: rib_update returns FALSE", |
54818e3b MS |
1863 | dplane_ctx_get_vrf(ctx), dest_str); |
1864 | } | |
1865 | ||
188a00e0 MS |
1866 | /* |
1867 | * Perform follow-up work if the actual status of the prefix | |
efe6c026 | 1868 | * changed. |
941e261c MS |
1869 | */ |
1870 | ||
efe6c026 MS |
1871 | end_count = 0; |
1872 | for (ALL_NEXTHOPS_PTR(rib_active_nhg(re), nexthop)) { | |
1873 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1874 | end_count++; | |
1875 | } | |
941e261c | 1876 | |
188a00e0 MS |
1877 | /* Various fib transitions: changed nexthops; from installed to |
1878 | * not-installed; or not-installed to installed. | |
efe6c026 | 1879 | */ |
188a00e0 | 1880 | if (start_count > 0 && end_count > 0) { |
11260e70 MS |
1881 | if (debug_p) |
1882 | zlog_debug("%u:%s applied nexthop changes from dplane notification", | |
1883 | dplane_ctx_get_vrf(ctx), dest_str); | |
188a00e0 MS |
1884 | |
1885 | /* Changed nexthops - update kernel/others */ | |
1886 | dplane_route_notif_update(rn, re, | |
1887 | DPLANE_OP_ROUTE_UPDATE, ctx); | |
1888 | ||
1889 | } else if (start_count == 0 && end_count > 0) { | |
efe6c026 MS |
1890 | if (debug_p) |
1891 | zlog_debug("%u:%s installed transition from dplane notification", | |
1892 | dplane_ctx_get_vrf(ctx), dest_str); | |
1893 | ||
1894 | /* We expect this to be the selected route, so we want | |
272e8903 | 1895 | * to tell others about this transition. |
efe6c026 MS |
1896 | */ |
1897 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1898 | ||
188a00e0 MS |
1899 | /* Changed nexthops - update kernel/others */ |
1900 | dplane_route_notif_update(rn, re, DPLANE_OP_ROUTE_INSTALL, ctx); | |
1901 | ||
efe6c026 MS |
1902 | /* Redistribute, lsp, and nht update */ |
1903 | redistribute_update(dest_pfx, src_pfx, re, NULL); | |
1904 | ||
efe6c026 MS |
1905 | } else if (start_count > 0 && end_count == 0) { |
1906 | if (debug_p) | |
1907 | zlog_debug("%u:%s un-installed transition from dplane notification", | |
1908 | dplane_ctx_get_vrf(ctx), dest_str); | |
1909 | ||
1910 | /* Transition from _something_ installed to _nothing_ | |
1911 | * installed. | |
1912 | */ | |
1913 | /* We expect this to be the selected route, so we want | |
1914 | * to tell others about this transistion. | |
1915 | */ | |
1916 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1917 | ||
188a00e0 MS |
1918 | /* Changed nexthops - update kernel/others */ |
1919 | dplane_route_notif_update(rn, re, DPLANE_OP_ROUTE_DELETE, ctx); | |
1920 | ||
efe6c026 | 1921 | /* Redistribute, lsp, and nht update */ |
40f321c0 | 1922 | redistribute_delete(dest_pfx, src_pfx, re, NULL); |
9287b4c5 | 1923 | } |
efe6c026 | 1924 | |
9287b4c5 MS |
1925 | /* Make any changes visible for lsp and nexthop-tracking processing */ |
1926 | zebra_rib_evaluate_rn_nexthops( | |
1927 | rn, zebra_router_get_next_sequence()); | |
efe6c026 | 1928 | |
9287b4c5 | 1929 | zebra_rib_evaluate_mpls(rn); |
941e261c | 1930 | |
54818e3b | 1931 | done: |
941e261c MS |
1932 | if (rn) |
1933 | route_unlock_node(rn); | |
1934 | ||
54818e3b MS |
1935 | /* Return context to dataplane module */ |
1936 | dplane_ctx_fini(&ctx); | |
1937 | } | |
1938 | ||
e22e8001 | 1939 | static void process_subq_nhg(struct listnode *lnode) |
e96f9203 | 1940 | { |
e22e8001 SW |
1941 | struct nhg_ctx *ctx = NULL; |
1942 | uint8_t qindex = route_info[ZEBRA_ROUTE_NHG].meta_q_map; | |
5110a0c6 | 1943 | |
e22e8001 SW |
1944 | ctx = listgetdata(lnode); |
1945 | ||
1946 | if (!ctx) | |
1947 | return; | |
1948 | ||
1949 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1950 | zlog_debug("NHG Context id=%u dequeued from sub-queue %u", | |
1951 | ctx->id, qindex); | |
1952 | ||
1953 | rib_nhg_process(ctx); | |
1954 | } | |
1955 | ||
1956 | static void process_subq_route(struct listnode *lnode, uint8_t qindex) | |
1957 | { | |
1958 | struct route_node *rnode = NULL; | |
1959 | rib_dest_t *dest = NULL; | |
1960 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 1961 | |
d62a17ae | 1962 | rnode = listgetdata(lnode); |
1963 | dest = rib_dest_from_rnode(rnode); | |
1964 | if (dest) | |
1965 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 1966 | |
d62a17ae | 1967 | rib_process(rnode); |
5110a0c6 | 1968 | |
d62a17ae | 1969 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1970 | char buf[SRCDEST2STR_BUFFER]; | |
efe6c026 | 1971 | |
d62a17ae | 1972 | srcdest_rnode2str(rnode, buf, sizeof(buf)); |
1973 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
1974 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
1975 | } | |
41ec9222 | 1976 | |
d62a17ae | 1977 | if (rnode->info) |
1978 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
1979 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 1980 | |
67b9467f | 1981 | #if 0 |
5110a0c6 SH |
1982 | else |
1983 | { | |
1984 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
1985 | __func__, rnode, rnode->lock); | |
1986 | zlog_backtrace(LOG_DEBUG); | |
1987 | } | |
67b9467f | 1988 | #endif |
d62a17ae | 1989 | route_unlock_node(rnode); |
e22e8001 SW |
1990 | } |
1991 | ||
1992 | /* Take a list of route_node structs and return 1, if there was a record | |
1993 | * picked from it and processed by rib_process(). Don't process more, | |
1994 | * than one RN record; operate only in the specified sub-queue. | |
1995 | */ | |
1996 | static unsigned int process_subq(struct list *subq, uint8_t qindex) | |
1997 | { | |
1998 | struct listnode *lnode = listhead(subq); | |
1999 | ||
2000 | if (!lnode) | |
2001 | return 0; | |
2002 | ||
2003 | if (qindex == route_info[ZEBRA_ROUTE_NHG].meta_q_map) | |
2004 | process_subq_nhg(lnode); | |
2005 | else | |
2006 | process_subq_route(lnode, qindex); | |
2007 | ||
d62a17ae | 2008 | list_delete_node(subq, lnode); |
e22e8001 | 2009 | |
d62a17ae | 2010 | return 1; |
e96f9203 DO |
2011 | } |
2012 | ||
2013 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from | |
d62a17ae | 2014 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
2015 | * data | |
e96f9203 DO |
2016 | * is pointed to the meta queue structure. |
2017 | */ | |
d62a17ae | 2018 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 2019 | { |
d62a17ae | 2020 | struct meta_queue *mq = data; |
2021 | unsigned i; | |
91f16812 MS |
2022 | uint32_t queue_len, queue_limit; |
2023 | ||
2024 | /* Ensure there's room for more dataplane updates */ | |
2025 | queue_limit = dplane_get_in_queue_limit(); | |
2026 | queue_len = dplane_get_in_queue_len(); | |
2027 | if (queue_len > queue_limit) { | |
2028 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
8b962e77 MS |
2029 | zlog_debug("rib queue: dplane queue len %u, limit %u, retrying", |
2030 | queue_len, queue_limit); | |
91f16812 MS |
2031 | |
2032 | /* Ensure that the meta-queue is actually enqueued */ | |
489a9614 | 2033 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2034 | work_queue_add(zrouter.ribq, zrouter.mq); |
91f16812 MS |
2035 | |
2036 | return WQ_QUEUE_BLOCKED; | |
2037 | } | |
5110a0c6 | 2038 | |
d62a17ae | 2039 | for (i = 0; i < MQ_SIZE; i++) |
2040 | if (process_subq(mq->subq[i], i)) { | |
2041 | mq->size--; | |
2042 | break; | |
2043 | } | |
2044 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
2045 | } |
2046 | ||
f52ed677 DS |
2047 | |
2048 | /* | |
2049 | * Look into the RN and queue it into the highest priority queue | |
2050 | * at this point in time for processing. | |
2051 | * | |
2052 | * We will enqueue a route node only once per invocation. | |
2053 | * | |
2054 | * There are two possibilities here that should be kept in mind. | |
2055 | * If the original invocation has not been pulled off for processing | |
2056 | * yet, A subsuquent invocation can have a route entry with a better | |
2057 | * meta queue index value and we can have a situation where | |
2058 | * we might have the same node enqueued 2 times. Not necessarily | |
2059 | * an optimal situation but it should be ok. | |
2060 | * | |
2061 | * The other possibility is that the original invocation has not | |
2062 | * been pulled off for processing yet, A subsusquent invocation | |
2063 | * doesn't have a route_entry with a better meta-queue and the | |
2064 | * original metaqueue index value will win and we'll end up with | |
2065 | * the route node enqueued once. | |
e96f9203 | 2066 | */ |
e22e8001 | 2067 | static int rib_meta_queue_add(struct meta_queue *mq, void *data) |
e96f9203 | 2068 | { |
e22e8001 | 2069 | struct route_node *rn = NULL; |
f52ed677 DS |
2070 | struct route_entry *re = NULL, *curr_re = NULL; |
2071 | uint8_t qindex = MQ_SIZE, curr_qindex = MQ_SIZE; | |
5110a0c6 | 2072 | |
e22e8001 SW |
2073 | rn = (struct route_node *)data; |
2074 | ||
f52ed677 DS |
2075 | RNODE_FOREACH_RE (rn, curr_re) { |
2076 | curr_qindex = route_info[curr_re->type].meta_q_map; | |
d62a17ae | 2077 | |
f52ed677 DS |
2078 | if (curr_qindex <= qindex) { |
2079 | re = curr_re; | |
2080 | qindex = curr_qindex; | |
d62a17ae | 2081 | } |
f52ed677 | 2082 | } |
5110a0c6 | 2083 | |
f52ed677 | 2084 | if (!re) |
e22e8001 | 2085 | return -1; |
5110a0c6 | 2086 | |
f52ed677 DS |
2087 | /* Invariant: at this point we always have rn->info set. */ |
2088 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
2089 | RIB_ROUTE_QUEUED(qindex))) { | |
d62a17ae | 2090 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2091 | rnode_debug(rn, re->vrf_id, | |
f52ed677 | 2092 | "rn %p is already queued in sub-queue %u", |
d62a17ae | 2093 | (void *)rn, qindex); |
e22e8001 | 2094 | return -1; |
d62a17ae | 2095 | } |
f52ed677 DS |
2096 | |
2097 | SET_FLAG(rib_dest_from_rnode(rn)->flags, RIB_ROUTE_QUEUED(qindex)); | |
2098 | listnode_add(mq->subq[qindex], rn); | |
2099 | route_lock_node(rn); | |
2100 | mq->size++; | |
2101 | ||
2102 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2103 | rnode_debug(rn, re->vrf_id, "queued rn %p into sub-queue %u", | |
2104 | (void *)rn, qindex); | |
e22e8001 SW |
2105 | |
2106 | return 0; | |
4d38fdb4 | 2107 | } |
2108 | ||
e22e8001 | 2109 | static int rib_meta_queue_nhg_add(struct meta_queue *mq, void *data) |
4d38fdb4 | 2110 | { |
e22e8001 SW |
2111 | struct nhg_ctx *ctx = NULL; |
2112 | uint8_t qindex = route_info[ZEBRA_ROUTE_NHG].meta_q_map; | |
fc328ac9 | 2113 | |
e22e8001 SW |
2114 | ctx = (struct nhg_ctx *)data; |
2115 | ||
2116 | if (!ctx) | |
2117 | return -1; | |
4d38fdb4 | 2118 | |
e22e8001 SW |
2119 | listnode_add(mq->subq[qindex], ctx); |
2120 | mq->size++; | |
2121 | ||
2122 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2123 | zlog_debug("NHG Context id=%u queued into sub-queue %u", | |
2124 | ctx->id, qindex); | |
2125 | ||
2126 | return 0; | |
2127 | } | |
2128 | ||
2129 | static int mq_add_handler(void *data, | |
2130 | int (*mq_add_func)(struct meta_queue *mq, void *data)) | |
2131 | { | |
489a9614 | 2132 | if (zrouter.ribq == NULL) { |
e914ccbe | 2133 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2134 | "%s: work_queue does not exist!", __func__); |
e22e8001 | 2135 | return -1; |
d62a17ae | 2136 | } |
2137 | ||
2138 | /* | |
2139 | * The RIB queue should normally be either empty or holding the only | |
2140 | * work_queue_item element. In the latter case this element would | |
2141 | * hold a pointer to the meta queue structure, which must be used to | |
2142 | * actually queue the route nodes to process. So create the MQ | |
2143 | * holder, if necessary, then push the work into it in any case. | |
2144 | * This semantics was introduced after 0.99.9 release. | |
2145 | */ | |
489a9614 | 2146 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2147 | work_queue_add(zrouter.ribq, zrouter.mq); |
e96f9203 | 2148 | |
e22e8001 SW |
2149 | return mq_add_func(zrouter.mq, data); |
2150 | } | |
fc328ac9 | 2151 | |
e22e8001 SW |
2152 | /* Add route_node to work queue and schedule processing */ |
2153 | int rib_queue_add(struct route_node *rn) | |
2154 | { | |
2155 | assert(rn); | |
2156 | ||
2157 | /* Pointless to queue a route_node with no RIB entries to add or remove | |
2158 | */ | |
2159 | if (!rnode_to_ribs(rn)) { | |
2160 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
2161 | __func__, (void *)rn, rn->lock); | |
2162 | zlog_backtrace(LOG_DEBUG); | |
2163 | return -1; | |
2164 | } | |
2165 | ||
2166 | return mq_add_handler(rn, &rib_meta_queue_add); | |
2167 | } | |
2168 | ||
2169 | int rib_queue_nhg_add(struct nhg_ctx *ctx) | |
2170 | { | |
2171 | assert(ctx); | |
2172 | ||
2173 | return mq_add_handler(ctx, &rib_meta_queue_nhg_add); | |
4d38fdb4 | 2174 | } |
2175 | ||
5110a0c6 SH |
2176 | /* Create new meta queue. |
2177 | A destructor function doesn't seem to be necessary here. | |
2178 | */ | |
d62a17ae | 2179 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 2180 | { |
d62a17ae | 2181 | struct meta_queue *new; |
2182 | unsigned i; | |
5110a0c6 | 2183 | |
d62a17ae | 2184 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
e96f9203 | 2185 | |
d62a17ae | 2186 | for (i = 0; i < MQ_SIZE; i++) { |
2187 | new->subq[i] = list_new(); | |
2188 | assert(new->subq[i]); | |
2189 | } | |
5110a0c6 | 2190 | |
d62a17ae | 2191 | return new; |
e96f9203 DO |
2192 | } |
2193 | ||
d62a17ae | 2194 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 2195 | { |
d62a17ae | 2196 | unsigned i; |
5a8dfcd8 | 2197 | |
d62a17ae | 2198 | for (i = 0; i < MQ_SIZE; i++) |
6a154c88 | 2199 | list_delete(&mq->subq[i]); |
5a8dfcd8 | 2200 | |
d62a17ae | 2201 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
2202 | } |
2203 | ||
4d38fdb4 | 2204 | /* initialise zebra rib work queue */ |
2561d12e | 2205 | static void rib_queue_init(void) |
4d38fdb4 | 2206 | { |
489a9614 DS |
2207 | if (!(zrouter.ribq = work_queue_new(zrouter.master, |
2208 | "route_node processing"))) { | |
e914ccbe | 2209 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2210 | "%s: could not initialise work queue!", __func__); |
d62a17ae | 2211 | return; |
2212 | } | |
4d38fdb4 | 2213 | |
d62a17ae | 2214 | /* fill in the work queue spec */ |
489a9614 DS |
2215 | zrouter.ribq->spec.workfunc = &meta_queue_process; |
2216 | zrouter.ribq->spec.errorfunc = NULL; | |
46a4e345 | 2217 | zrouter.ribq->spec.completion_func = NULL; |
d62a17ae | 2218 | /* XXX: TODO: These should be runtime configurable via vty */ |
489a9614 DS |
2219 | zrouter.ribq->spec.max_retries = 3; |
2220 | zrouter.ribq->spec.hold = ZEBRA_RIB_PROCESS_HOLD_TIME; | |
2221 | zrouter.ribq->spec.retry = ZEBRA_RIB_PROCESS_RETRY_TIME; | |
d62a17ae | 2222 | |
ea45a4e7 | 2223 | if (!(zrouter.mq = meta_queue_new())) { |
e914ccbe | 2224 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2225 | "%s: could not initialise meta queue!", __func__); |
d62a17ae | 2226 | return; |
2227 | } | |
2228 | return; | |
718e3744 | 2229 | } |
2230 | ||
434434f7 DS |
2231 | rib_dest_t *zebra_rib_create_dest(struct route_node *rn) |
2232 | { | |
2233 | rib_dest_t *dest; | |
2234 | ||
2235 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); | |
aa57abfb | 2236 | rnh_list_init(&dest->nht); |
434434f7 DS |
2237 | route_lock_node(rn); /* rn route table reference */ |
2238 | rn->info = dest; | |
2239 | dest->rnode = rn; | |
2240 | ||
2241 | return dest; | |
2242 | } | |
2243 | ||
6d691129 PJ |
2244 | /* RIB updates are processed via a queue of pointers to route_nodes. |
2245 | * | |
2246 | * The queue length is bounded by the maximal size of the routing table, | |
2247 | * as a route_node will not be requeued, if already queued. | |
2248 | * | |
f0f77c9a DS |
2249 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
2250 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 2251 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 2252 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 2253 | * |
f0f77c9a | 2254 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
2255 | * |
2256 | * rib_addnode | |
f0f77c9a DS |
2257 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
2258 | * |-------->| | best RE, if required | |
3c0755dc PJ |
2259 | * | | |
2260 | * static_install->|->rib_addqueue...... -> rib_process | |
2261 | * | | | |
2262 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
2263 | * |-> set ROUTE_ENTRY_REMOVE | |
2264 | * rib_delnode (RE freed) | |
6d691129 | 2265 | * |
9fd92e3c AS |
2266 | * The 'info' pointer of a route_node points to a rib_dest_t |
2267 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
2268 | * dest is created on-demand by rib_link() and is kept around at least | |
2269 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 2270 | * |
6d691129 PJ |
2271 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
2272 | * | |
2273 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
2274 | * - dest attached to route_node: |
2275 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
2276 | * - route_node processing queue |
2277 | * - managed by: rib_addqueue, rib_process. | |
2278 | * | |
2279 | */ | |
d62a17ae | 2280 | |
f0f77c9a | 2281 | /* Add RE to head of the route node. */ |
d62a17ae | 2282 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
2283 | { | |
d62a17ae | 2284 | rib_dest_t *dest; |
2285 | afi_t afi; | |
2286 | const char *rmap_name; | |
9fd92e3c | 2287 | |
d62a17ae | 2288 | assert(re && rn); |
9fd92e3c | 2289 | |
d62a17ae | 2290 | dest = rib_dest_from_rnode(rn); |
2291 | if (!dest) { | |
2292 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2293 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 2294 | |
434434f7 | 2295 | dest = zebra_rib_create_dest(rn); |
d62a17ae | 2296 | } |
2263a412 | 2297 | |
045207e2 | 2298 | re_list_add_head(&dest->routes, re); |
d62a17ae | 2299 | |
2300 | afi = (rn->p.family == AF_INET) | |
2301 | ? AFI_IP | |
2302 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
fe257ae7 DS |
2303 | if (is_zebra_import_table_enabled(afi, re->vrf_id, re->table)) { |
2304 | struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(re->vrf_id); | |
2305 | ||
d62a17ae | 2306 | rmap_name = zebra_get_import_table_route_map(afi, re->table); |
fe257ae7 | 2307 | zebra_add_import_table_entry(zvrf, rn, re, rmap_name); |
d62a17ae | 2308 | } else if (process) |
2309 | rib_queue_add(rn); | |
2310 | } | |
2311 | ||
7e24fdf3 DS |
2312 | static void rib_addnode(struct route_node *rn, |
2313 | struct route_entry *re, int process) | |
d62a17ae | 2314 | { |
2315 | /* RE node has been un-removed before route-node is processed. | |
2316 | * route_node must hence already be on the queue for processing.. | |
2317 | */ | |
2318 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
2319 | if (IS_ZEBRA_DEBUG_RIB) | |
2320 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
2321 | (void *)rn, (void *)re); | |
2322 | ||
2323 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2324 | return; | |
2325 | } | |
2326 | rib_link(rn, re, process); | |
6d691129 PJ |
2327 | } |
2328 | ||
9fd92e3c AS |
2329 | /* |
2330 | * rib_unlink | |
2331 | * | |
2332 | * Detach a rib structure from a route_node. | |
2333 | * | |
2334 | * Note that a call to rib_unlink() should be followed by a call to | |
2335 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
2336 | * longer required to be deleted. | |
2337 | */ | |
d62a17ae | 2338 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 2339 | { |
d62a17ae | 2340 | rib_dest_t *dest; |
8e766379 | 2341 | struct nhg_hash_entry *nhe = NULL; |
9fd92e3c | 2342 | |
d62a17ae | 2343 | assert(rn && re); |
6d691129 | 2344 | |
d62a17ae | 2345 | if (IS_ZEBRA_DEBUG_RIB) |
2346 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
2347 | (void *)re); | |
6d691129 | 2348 | |
d62a17ae | 2349 | dest = rib_dest_from_rnode(rn); |
6d691129 | 2350 | |
045207e2 | 2351 | re_list_del(&dest->routes, re); |
7a4bb9c5 | 2352 | |
2eb07de3 DS |
2353 | if (dest->selected_fib == re) |
2354 | dest->selected_fib = NULL; | |
2355 | ||
e22e8001 SW |
2356 | if (re->nhe_id) { |
2357 | nhe = zebra_nhg_lookup_id(re->nhe_id); | |
2358 | if (nhe) | |
2359 | zebra_nhg_decrement_ref(nhe); | |
c415d895 MS |
2360 | } else if (re->nhe->nhg.nexthop) |
2361 | nexthops_free(re->nhe->nhg.nexthop); | |
22bcedb2 | 2362 | |
ee5e8a48 MS |
2363 | nexthops_free(re->fib_ng.nexthop); |
2364 | ||
d62a17ae | 2365 | XFREE(MTYPE_RE, re); |
2366 | } | |
2367 | ||
2368 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2369 | { | |
2370 | afi_t afi; | |
2371 | ||
2372 | if (IS_ZEBRA_DEBUG_RIB) | |
2373 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2374 | (void *)rn, (void *)re); | |
2375 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2376 | ||
2377 | afi = (rn->p.family == AF_INET) | |
2378 | ? AFI_IP | |
2379 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
fe257ae7 DS |
2380 | if (is_zebra_import_table_enabled(afi, re->vrf_id, re->table)) { |
2381 | struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(re->vrf_id); | |
2382 | ||
2383 | zebra_del_import_table_entry(zvrf, rn, re); | |
d62a17ae | 2384 | /* Just clean up if non main table */ |
2385 | if (IS_ZEBRA_DEBUG_RIB) { | |
2386 | char buf[SRCDEST2STR_BUFFER]; | |
2387 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
2388 | zlog_debug("%u:%s: Freeing route rn %p, re %p (%s)", |
2389 | re->vrf_id, buf, rn, re, | |
2390 | zebra_route_string(re->type)); | |
d62a17ae | 2391 | } |
7a4bb9c5 | 2392 | |
d62a17ae | 2393 | rib_unlink(rn, re); |
2394 | } else { | |
2395 | rib_queue_add(rn); | |
2396 | } | |
718e3744 | 2397 | } |
2398 | ||
f0f77c9a | 2399 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2400 | * standard debug log. Calling function name and IP prefix in |
2401 | * question are passed as 1st and 2nd arguments. | |
2402 | */ | |
2403 | ||
d62a17ae | 2404 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2405 | union prefixconstptr src_pp, | |
2406 | const struct route_entry *re) | |
2407 | { | |
d62a17ae | 2408 | const struct prefix *src_p = src_pp.p; |
2409 | bool is_srcdst = src_p && src_p->prefixlen; | |
2410 | char straddr[PREFIX_STRLEN]; | |
2411 | char srcaddr[PREFIX_STRLEN]; | |
53c16fbe | 2412 | char nhname[PREFIX_STRLEN]; |
d62a17ae | 2413 | struct nexthop *nexthop; |
c479e756 | 2414 | struct vrf *vrf = vrf_lookup_by_id(re->vrf_id); |
d62a17ae | 2415 | |
c479e756 | 2416 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %s(%u)", func, |
d62a17ae | 2417 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), |
2418 | is_srcdst ? " from " : "", | |
2419 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2420 | : "", | |
c479e756 | 2421 | VRF_LOGNAME(vrf), re->vrf_id); |
cc54cfee | 2422 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", |
53c16fbe | 2423 | straddr, (unsigned long)re->uptime, re->type, re->instance, |
cc54cfee | 2424 | re->table); |
d62a17ae | 2425 | zlog_debug( |
2426 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
53c16fbe DS |
2427 | straddr, re->metric, re->mtu, re->distance, re->flags, re->status); |
2428 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", straddr, | |
c415d895 MS |
2429 | nexthop_group_nexthop_num(&(re->nhe->nhg)), |
2430 | nexthop_group_active_nexthop_num(&(re->nhe->nhg))); | |
d62a17ae | 2431 | |
c415d895 | 2432 | for (ALL_NEXTHOPS(re->nhe->nhg, nexthop)) { |
2d68a0f2 DS |
2433 | struct interface *ifp; |
2434 | struct vrf *vrf = vrf_lookup_by_id(nexthop->vrf_id); | |
2435 | ||
2436 | switch (nexthop->type) { | |
2437 | case NEXTHOP_TYPE_BLACKHOLE: | |
53c16fbe | 2438 | sprintf(nhname, "Blackhole"); |
2d68a0f2 DS |
2439 | break; |
2440 | case NEXTHOP_TYPE_IFINDEX: | |
2441 | ifp = if_lookup_by_index(nexthop->ifindex, | |
2442 | nexthop->vrf_id); | |
53c16fbe | 2443 | sprintf(nhname, "%s", ifp ? ifp->name : "Unknown"); |
2d68a0f2 DS |
2444 | break; |
2445 | case NEXTHOP_TYPE_IPV4: | |
2446 | /* fallthrough */ | |
2447 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
53c16fbe | 2448 | inet_ntop(AF_INET, &nexthop->gate, nhname, |
2d68a0f2 DS |
2449 | INET6_ADDRSTRLEN); |
2450 | break; | |
2451 | case NEXTHOP_TYPE_IPV6: | |
2452 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
53c16fbe | 2453 | inet_ntop(AF_INET6, &nexthop->gate, nhname, |
2d68a0f2 DS |
2454 | INET6_ADDRSTRLEN); |
2455 | break; | |
2456 | } | |
0641a955 | 2457 | zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s%s%s", |
53c16fbe | 2458 | straddr, (nexthop->rparent ? " NH" : "NH"), nhname, |
2d68a0f2 DS |
2459 | nexthop->ifindex, vrf ? vrf->name : "Unknown", |
2460 | nexthop->vrf_id, | |
d62a17ae | 2461 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) |
2462 | ? "ACTIVE " | |
2463 | : ""), | |
677c1dd5 | 2464 | (CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED) |
d62a17ae | 2465 | ? "FIB " |
2466 | : ""), | |
2467 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
df38b099 DS |
2468 | ? "RECURSIVE " |
2469 | : ""), | |
2470 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK) | |
2471 | ? "ONLINK " | |
2472 | : ""), | |
df38b099 DS |
2473 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_DUPLICATE) |
2474 | ? "DUPLICATE " | |
d62a17ae | 2475 | : "")); |
2476 | } | |
53c16fbe | 2477 | zlog_debug("%s: dump complete", straddr); |
dc95824a DO |
2478 | } |
2479 | ||
2480 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2481 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2482 | */ |
2483 | ||
d62a17ae | 2484 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2485 | { | |
2486 | struct route_table *table; | |
2487 | struct route_node *rn; | |
2488 | struct route_entry *re; | |
2489 | char prefix_buf[INET_ADDRSTRLEN]; | |
2490 | ||
2491 | /* Lookup table. */ | |
2492 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2493 | if (!table) { | |
e914ccbe | 2494 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2495 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2496 | vrf_id); | |
d62a17ae | 2497 | return; |
2498 | } | |
2499 | ||
2500 | /* Scan the RIB table for exactly matching RE entry. */ | |
2501 | rn = route_node_lookup(table, (struct prefix *)p); | |
2502 | ||
2503 | /* No route for this prefix. */ | |
2504 | if (!rn) { | |
32391aff | 2505 | zlog_debug("%s:%u lookup failed for %s", __func__, vrf_id, |
d62a17ae | 2506 | prefix2str((struct prefix *)p, prefix_buf, |
2507 | sizeof(prefix_buf))); | |
2508 | return; | |
2509 | } | |
2510 | ||
2511 | /* Unlock node. */ | |
2512 | route_unlock_node(rn); | |
2513 | ||
2514 | /* let's go */ | |
a2addae8 | 2515 | RNODE_FOREACH_RE (rn, re) { |
32391aff DS |
2516 | zlog_debug("%s:%u rn %p, re %p: %s, %s", |
2517 | __func__, vrf_id, | |
2518 | (void *)rn, (void *)re, | |
d62a17ae | 2519 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) |
2520 | ? "removed" | |
2521 | : "NOT removed"), | |
2522 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2523 | ? "selected" | |
2524 | : "NOT selected")); | |
2525 | route_entry_dump(p, NULL, re); | |
2526 | } | |
dc95824a DO |
2527 | } |
2528 | ||
20e5ff0a DO |
2529 | /* Check if requested address assignment will fail due to another |
2530 | * route being installed by zebra in FIB already. Take necessary | |
2531 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2532 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2533 | */ |
d62a17ae | 2534 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2535 | { | |
2536 | struct route_table *table; | |
2537 | struct route_node *rn; | |
5f7a4718 | 2538 | rib_dest_t *dest; |
d62a17ae | 2539 | |
2540 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
e914ccbe | 2541 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2542 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2543 | vrf_id); | |
d62a17ae | 2544 | return; |
2545 | } | |
2546 | ||
2547 | /* No matches would be the simplest case. */ | |
2548 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2549 | return; | |
2550 | ||
2551 | /* Unlock node. */ | |
2552 | route_unlock_node(rn); | |
2553 | ||
5f7a4718 | 2554 | dest = rib_dest_from_rnode(rn); |
d62a17ae | 2555 | /* Check all RE entries. In case any changes have to be done, requeue |
2556 | * the RN into RIBQ head. If the routing message about the new connected | |
2557 | * route (generated by the IP address we are going to assign very soon) | |
2558 | * comes before the RIBQ is processed, the new RE entry will join | |
2559 | * RIBQ record already on head. This is necessary for proper | |
2560 | * revalidation | |
2561 | * of the rest of the RE. | |
2562 | */ | |
3cdba47a | 2563 | if (dest->selected_fib) { |
5f7a4718 DS |
2564 | if (IS_ZEBRA_DEBUG_RIB) { |
2565 | char buf[PREFIX_STRLEN]; | |
2566 | ||
2567 | zlog_debug("%u:%s: freeing way for connected prefix", | |
2568 | dest->selected_fib->vrf_id, | |
2569 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2570 | route_entry_dump(&rn->p, NULL, dest->selected_fib); | |
d62a17ae | 2571 | } |
5f7a4718 | 2572 | rib_uninstall(rn, dest->selected_fib); |
d62a17ae | 2573 | rib_queue_add(rn); |
ade4a886 | 2574 | } |
20e5ff0a DO |
2575 | } |
2576 | ||
d62a17ae | 2577 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
0eb97b86 MS |
2578 | struct prefix_ipv6 *src_p, struct route_entry *re, |
2579 | struct nexthop_group *ng) | |
718e3744 | 2580 | { |
98cda54a | 2581 | struct nhg_hash_entry *nhe = NULL; |
d62a17ae | 2582 | struct route_table *table; |
2583 | struct route_node *rn; | |
40ecd8e4 | 2584 | struct route_entry *same = NULL; |
d62a17ae | 2585 | int ret = 0; |
b4c034b0 | 2586 | |
d62a17ae | 2587 | if (!re) |
2588 | return 0; | |
b4c034b0 | 2589 | |
1f610a1f | 2590 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
05737783 | 2591 | |
d62a17ae | 2592 | /* Lookup table. */ |
c7c0b007 SW |
2593 | table = zebra_vrf_get_table_with_table_id(afi, safi, re->vrf_id, |
2594 | re->table); | |
7865c65d | 2595 | if (!table) { |
0eb97b86 MS |
2596 | if (ng) |
2597 | nexthop_group_delete(&ng); | |
7865c65d | 2598 | XFREE(MTYPE_RE, re); |
d62a17ae | 2599 | return 0; |
7865c65d | 2600 | } |
cddf391b | 2601 | |
98cda54a SW |
2602 | if (re->nhe_id) { |
2603 | nhe = zebra_nhg_lookup_id(re->nhe_id); | |
2604 | ||
2605 | if (!nhe) { | |
2606 | flog_err( | |
2607 | EC_ZEBRA_TABLE_LOOKUP_FAILED, | |
2608 | "Zebra failed to find the nexthop hash entry for id=%u in a route entry", | |
2609 | re->nhe_id); | |
2610 | XFREE(MTYPE_RE, re); | |
2611 | return -1; | |
2612 | } | |
2613 | } else { | |
0eb97b86 | 2614 | nhe = zebra_nhg_rib_find(0, ng, afi); |
98cda54a SW |
2615 | |
2616 | /* | |
2617 | * The nexthops got copied over into an nhe, | |
2618 | * so free them now. | |
2619 | */ | |
0eb97b86 | 2620 | nexthop_group_delete(&ng); |
98cda54a SW |
2621 | |
2622 | if (!nhe) { | |
2623 | char buf[PREFIX_STRLEN] = ""; | |
2624 | char buf2[PREFIX_STRLEN] = ""; | |
2625 | ||
2626 | flog_err( | |
2627 | EC_ZEBRA_TABLE_LOOKUP_FAILED, | |
2628 | "Zebra failed to find or create a nexthop hash entry for %s%s%s", | |
2629 | prefix2str(p, buf, sizeof(buf)), | |
2630 | src_p ? " from " : "", | |
2631 | src_p ? prefix2str(src_p, buf2, sizeof(buf2)) | |
2632 | : ""); | |
2633 | ||
2634 | XFREE(MTYPE_RE, re); | |
2635 | return -1; | |
2636 | } | |
98cda54a SW |
2637 | } |
2638 | ||
4b87c90d SW |
2639 | /* |
2640 | * Attach the re to the nhe's nexthop group. | |
2641 | * | |
2642 | * TODO: This will need to change when we start getting IDs from upper | |
2643 | * level protocols, as the refcnt might be wrong, since it checks | |
2644 | * if old_id != new_id. | |
2645 | */ | |
5463ce26 | 2646 | route_entry_update_nhe(re, nhe); |
98cda54a | 2647 | |
d62a17ae | 2648 | /* Make it sure prefixlen is applied to the prefix. */ |
2649 | apply_mask(p); | |
2650 | if (src_p) | |
2651 | apply_mask_ipv6(src_p); | |
718e3744 | 2652 | |
d62a17ae | 2653 | /* Set default distance by route type. */ |
eea2ef07 | 2654 | if (re->distance == 0) |
0492eea0 | 2655 | re->distance = route_distance(re->type); |
718e3744 | 2656 | |
d62a17ae | 2657 | /* Lookup route node.*/ |
2658 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2659 | |
40ecd8e4 DS |
2660 | /* |
2661 | * If same type of route are installed, treat it as a implicit | |
2662 | * withdraw. | |
2663 | * If the user has specified the No route replace semantics | |
2664 | * for the install don't do a route replace. | |
2665 | */ | |
a2addae8 | 2666 | RNODE_FOREACH_RE (rn, same) { |
d62a17ae | 2667 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) |
2668 | continue; | |
41ec9222 | 2669 | |
eb327fa5 RW |
2670 | if (same->type != re->type) |
2671 | continue; | |
2672 | if (same->instance != re->instance) | |
2673 | continue; | |
996c9314 LB |
2674 | if (same->type == ZEBRA_ROUTE_KERNEL |
2675 | && same->metric != re->metric) | |
eb327fa5 | 2676 | continue; |
40ecd8e4 DS |
2677 | |
2678 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && | |
2679 | same->distance != re->distance) | |
2680 | continue; | |
2681 | ||
844b3a87 | 2682 | /* |
40ecd8e4 DS |
2683 | * We should allow duplicate connected routes |
2684 | * because of IPv6 link-local routes and unnumbered | |
2685 | * interfaces on Linux. | |
844b3a87 RW |
2686 | */ |
2687 | if (same->type != ZEBRA_ROUTE_CONNECT) | |
d62a17ae | 2688 | break; |
2689 | } | |
718e3744 | 2690 | |
cf363e1b MS |
2691 | /* If this route is kernel/connected route, notify the dataplane. */ |
2692 | if (RIB_SYSTEM_ROUTE(re)) { | |
2693 | /* Notify dataplane */ | |
2694 | dplane_sys_route_add(rn, re); | |
2695 | } | |
2696 | ||
d62a17ae | 2697 | /* Link new re to node.*/ |
2698 | if (IS_ZEBRA_DEBUG_RIB) { | |
2da33d6b DS |
2699 | rnode_debug(rn, re->vrf_id, |
2700 | "Inserting route rn %p, re %p (%s) existing %p", | |
2701 | rn, re, zebra_route_string(re->type), same); | |
718e3744 | 2702 | |
d62a17ae | 2703 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2704 | route_entry_dump(p, src_p, re); | |
718e3744 | 2705 | } |
6883bf8d DS |
2706 | |
2707 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 2708 | rib_addnode(rn, re, 1); |
2709 | ret = 1; | |
6b0655a2 | 2710 | |
d62a17ae | 2711 | /* Free implicit route.*/ |
2712 | if (same) { | |
2713 | rib_delnode(rn, same); | |
2714 | ret = -1; | |
2715 | } | |
718e3744 | 2716 | |
d62a17ae | 2717 | route_unlock_node(rn); |
2718 | return ret; | |
2719 | } | |
2720 | ||
2721 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
d7c0a89a | 2722 | unsigned short instance, int flags, struct prefix *p, |
fd36be7e | 2723 | struct prefix_ipv6 *src_p, const struct nexthop *nh, |
bc541126 SW |
2724 | uint32_t nhe_id, uint32_t table_id, uint32_t metric, |
2725 | uint8_t distance, bool fromkernel) | |
d62a17ae | 2726 | { |
2727 | struct route_table *table; | |
2728 | struct route_node *rn; | |
2729 | struct route_entry *re; | |
2730 | struct route_entry *fib = NULL; | |
2731 | struct route_entry *same = NULL; | |
fd36be7e | 2732 | struct nexthop *rtnh; |
d62a17ae | 2733 | char buf2[INET6_ADDRSTRLEN]; |
5f7a4718 | 2734 | rib_dest_t *dest; |
d62a17ae | 2735 | |
1f610a1f | 2736 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
d62a17ae | 2737 | |
2738 | /* Lookup table. */ | |
c7c0b007 SW |
2739 | table = zebra_vrf_lookup_table_with_table_id(afi, safi, vrf_id, |
2740 | table_id); | |
d62a17ae | 2741 | if (!table) |
2742 | return; | |
2743 | ||
2744 | /* Apply mask. */ | |
2745 | apply_mask(p); | |
2746 | if (src_p) | |
2747 | apply_mask_ipv6(src_p); | |
2748 | ||
2749 | /* Lookup route node. */ | |
2750 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2751 | if (!rn) { | |
2752 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2753 | ||
2754 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2755 | if (src_p && src_p->prefixlen) | |
2756 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2757 | else | |
2758 | src_buf[0] = '\0'; | |
2759 | ||
82d7d0e2 DS |
2760 | if (IS_ZEBRA_DEBUG_RIB) { |
2761 | struct vrf *vrf = vrf_lookup_by_id(vrf_id); | |
2762 | ||
2763 | zlog_debug("%s[%d]:%s%s%s doesn't exist in rib", | |
2764 | vrf->name, table_id, dst_buf, | |
d62a17ae | 2765 | (src_buf[0] != '\0') ? " from " : "", |
2766 | src_buf); | |
82d7d0e2 | 2767 | } |
d62a17ae | 2768 | return; |
2769 | } | |
718e3744 | 2770 | |
5f7a4718 DS |
2771 | dest = rib_dest_from_rnode(rn); |
2772 | fib = dest->selected_fib; | |
2773 | ||
d62a17ae | 2774 | /* Lookup same type route. */ |
a2addae8 | 2775 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2776 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2777 | continue; | |
2778 | ||
d62a17ae | 2779 | if (re->type != type) |
2780 | continue; | |
2781 | if (re->instance != instance) | |
2782 | continue; | |
40ecd8e4 DS |
2783 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && |
2784 | distance != re->distance) | |
2785 | continue; | |
2786 | ||
996c9314 | 2787 | if (re->type == ZEBRA_ROUTE_KERNEL && re->metric != metric) |
f19435a8 | 2788 | continue; |
0eb97b86 | 2789 | if (re->type == ZEBRA_ROUTE_CONNECT && |
c415d895 | 2790 | (rtnh = re->nhe->nhg.nexthop) |
fd36be7e DL |
2791 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { |
2792 | if (rtnh->ifindex != nh->ifindex) | |
d62a17ae | 2793 | continue; |
d62a17ae | 2794 | same = re; |
2795 | break; | |
2796 | } | |
fec211ad | 2797 | |
d62a17ae | 2798 | /* Make sure that the route found has the same gateway. */ |
fec211ad | 2799 | if (nhe_id && re->nhe_id == nhe_id) { |
bc541126 SW |
2800 | same = re; |
2801 | break; | |
fec211ad SW |
2802 | } |
2803 | ||
2804 | if (nh == NULL) { | |
2805 | same = re; | |
2806 | break; | |
2807 | } | |
c415d895 | 2808 | for (ALL_NEXTHOPS(re->nhe->nhg, rtnh)) { |
fec211ad SW |
2809 | /* |
2810 | * No guarantee all kernel send nh with labels | |
2811 | * on delete. | |
2812 | */ | |
2813 | if (nexthop_same_no_labels(rtnh, nh)) { | |
d62a17ae | 2814 | same = re; |
2815 | break; | |
2816 | } | |
d62a17ae | 2817 | } |
fec211ad SW |
2818 | |
2819 | if (same) | |
2820 | break; | |
d62a17ae | 2821 | } |
2822 | /* If same type of route can't be found and this message is from | |
2823 | kernel. */ | |
2824 | if (!same) { | |
5dfeba19 DS |
2825 | /* |
2826 | * In the past(HA!) we could get here because | |
2827 | * we were receiving a route delete from the | |
2828 | * kernel and we're not marking the proto | |
2829 | * as coming from it's appropriate originator. | |
2830 | * Now that we are properly noticing the fact | |
2831 | * that the kernel has deleted our route we | |
2832 | * are not going to get called in this path | |
2833 | * I am going to leave this here because | |
2834 | * this might still work this way on non-linux | |
2835 | * platforms as well as some weird state I have | |
2836 | * not properly thought of yet. | |
2837 | * If we can show that this code path is | |
2838 | * dead then we can remove it. | |
2839 | */ | |
b8faa875 | 2840 | if (fib && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { |
d62a17ae | 2841 | if (IS_ZEBRA_DEBUG_RIB) { |
2da33d6b DS |
2842 | rnode_debug(rn, vrf_id, |
2843 | "rn %p, re %p (%s) was deleted from kernel, adding", | |
2844 | rn, fib, | |
2845 | zebra_route_string(fib->type)); | |
d62a17ae | 2846 | } |
2847 | if (allow_delete) { | |
677c1dd5 | 2848 | UNSET_FLAG(fib->status, ROUTE_ENTRY_INSTALLED); |
d62a17ae | 2849 | /* Unset flags. */ |
c415d895 | 2850 | for (rtnh = fib->nhe->nhg.nexthop; rtnh; |
fd36be7e DL |
2851 | rtnh = rtnh->next) |
2852 | UNSET_FLAG(rtnh->flags, | |
d62a17ae | 2853 | NEXTHOP_FLAG_FIB); |
2854 | ||
ed216282 DS |
2855 | /* |
2856 | * This is a non FRR route | |
2857 | * as such we should mark | |
2858 | * it as deleted | |
2859 | */ | |
5f7a4718 | 2860 | dest->selected_fib = NULL; |
d62a17ae | 2861 | } else { |
2862 | /* This means someone else, other than Zebra, | |
2863 | * has deleted | |
2864 | * a Zebra router from the kernel. We will add | |
2865 | * it back */ | |
2866 | rib_install_kernel(rn, fib, NULL); | |
2867 | } | |
2868 | } else { | |
2869 | if (IS_ZEBRA_DEBUG_RIB) { | |
fd36be7e | 2870 | if (nh) |
d62a17ae | 2871 | rnode_debug( |
2872 | rn, vrf_id, | |
2873 | "via %s ifindex %d type %d " | |
2874 | "doesn't exist in rib", | |
36228974 | 2875 | inet_ntop(afi2family(afi), |
2876 | &nh->gate, buf2, | |
2877 | sizeof(buf2)), | |
2878 | nh->ifindex, type); | |
d62a17ae | 2879 | else |
2880 | rnode_debug( | |
2881 | rn, vrf_id, | |
fd36be7e DL |
2882 | "type %d doesn't exist in rib", |
2883 | type); | |
d62a17ae | 2884 | } |
2885 | route_unlock_node(rn); | |
2886 | return; | |
2887 | } | |
2888 | } | |
718e3744 | 2889 | |
5dfeba19 | 2890 | if (same) { |
996c9314 LB |
2891 | if (fromkernel && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE) |
2892 | && !allow_delete) { | |
5dfeba19 DS |
2893 | rib_install_kernel(rn, same, NULL); |
2894 | route_unlock_node(rn); | |
2895 | ||
2896 | return; | |
2897 | } | |
6134fd82 | 2898 | |
2b83602b | 2899 | /* Special handling for IPv4 or IPv6 routes sourced from |
2900 | * EVPN - the nexthop (and associated MAC) need to be | |
2901 | * uninstalled if no more refs. | |
2902 | */ | |
90264d64 | 2903 | if (CHECK_FLAG(flags, ZEBRA_FLAG_EVPN_ROUTE)) { |
6134fd82 | 2904 | struct nexthop *tmp_nh; |
2905 | ||
c415d895 | 2906 | for (ALL_NEXTHOPS(re->nhe->nhg, tmp_nh)) { |
6134fd82 | 2907 | struct ipaddr vtep_ip; |
2908 | ||
2909 | memset(&vtep_ip, 0, sizeof(struct ipaddr)); | |
1ec31309 | 2910 | if (afi == AFI_IP) { |
2911 | vtep_ip.ipa_type = IPADDR_V4; | |
2912 | memcpy(&(vtep_ip.ipaddr_v4), | |
2913 | &(tmp_nh->gate.ipv4), | |
2914 | sizeof(struct in_addr)); | |
2915 | } else { | |
2916 | vtep_ip.ipa_type = IPADDR_V6; | |
2917 | memcpy(&(vtep_ip.ipaddr_v6), | |
2918 | &(tmp_nh->gate.ipv6), | |
2919 | sizeof(struct in6_addr)); | |
2920 | } | |
a317a9b9 | 2921 | zebra_vxlan_evpn_vrf_route_del(re->vrf_id, |
6134fd82 | 2922 | &vtep_ip, p); |
2923 | } | |
2924 | } | |
cf363e1b MS |
2925 | |
2926 | /* Notify dplane if system route changes */ | |
2927 | if (RIB_SYSTEM_ROUTE(re)) | |
2928 | dplane_sys_route_del(rn, same); | |
2929 | ||
d62a17ae | 2930 | rib_delnode(rn, same); |
5dfeba19 | 2931 | } |
05737783 | 2932 | |
d62a17ae | 2933 | route_unlock_node(rn); |
2934 | return; | |
2935 | } | |
718e3744 | 2936 | |
718e3744 | 2937 | |
d7c0a89a QY |
2938 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, |
2939 | unsigned short instance, int flags, struct prefix *p, | |
2940 | struct prefix_ipv6 *src_p, const struct nexthop *nh, | |
8032b717 SW |
2941 | uint32_t nhe_id, uint32_t table_id, uint32_t metric, uint32_t mtu, |
2942 | uint8_t distance, route_tag_t tag) | |
d62a17ae | 2943 | { |
bbb322f2 SW |
2944 | struct route_entry *re = NULL; |
2945 | struct nexthop *nexthop = NULL; | |
0eb97b86 | 2946 | struct nexthop_group *ng = NULL; |
718e3744 | 2947 | |
66af6845 | 2948 | /* Allocate new route_entry structure. */ |
d62a17ae | 2949 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); |
d62a17ae | 2950 | re->type = type; |
2951 | re->instance = instance; | |
2952 | re->distance = distance; | |
2953 | re->flags = flags; | |
2954 | re->metric = metric; | |
2955 | re->mtu = mtu; | |
2956 | re->table = table_id; | |
2957 | re->vrf_id = vrf_id; | |
98572489 | 2958 | re->uptime = monotime(NULL); |
4e40b6d6 | 2959 | re->tag = tag; |
bbb322f2 SW |
2960 | re->nhe_id = nhe_id; |
2961 | ||
0eb97b86 MS |
2962 | /* If the owner of the route supplies a shared nexthop-group id, |
2963 | * we'll use that. Otherwise, pass the nexthop along directly. | |
2964 | */ | |
e22e8001 | 2965 | if (!nhe_id) { |
0eb97b86 | 2966 | ng = nexthop_group_new(); |
d62a17ae | 2967 | |
e22e8001 SW |
2968 | /* Add nexthop. */ |
2969 | nexthop = nexthop_new(); | |
2970 | *nexthop = *nh; | |
0eb97b86 | 2971 | nexthop_group_add_sorted(ng, nexthop); |
e22e8001 | 2972 | } |
718e3744 | 2973 | |
0eb97b86 | 2974 | return rib_add_multipath(afi, safi, p, src_p, re, ng); |
718e3744 | 2975 | } |
2976 | ||
2a99ab95 SW |
2977 | static const char *rib_update_event2str(rib_update_event_t event) |
2978 | { | |
2979 | const char *ret = "UNKNOWN"; | |
2980 | ||
2981 | switch (event) { | |
2982 | case RIB_UPDATE_KERNEL: | |
2983 | ret = "RIB_UPDATE_KERNEL"; | |
2984 | break; | |
2985 | case RIB_UPDATE_RMAP_CHANGE: | |
2986 | ret = "RIB_UPDATE_RMAP_CHANGE"; | |
2987 | break; | |
2988 | case RIB_UPDATE_OTHER: | |
2989 | ret = "RIB_UPDATE_OTHER"; | |
2990 | break; | |
2991 | case RIB_UPDATE_MAX: | |
2992 | break; | |
2993 | } | |
2994 | ||
2995 | return ret; | |
2996 | } | |
2997 | ||
2998 | ||
2999 | /* Schedule route nodes to be processed if they match the type */ | |
3000 | static void rib_update_route_node(struct route_node *rn, int type) | |
3001 | { | |
3002 | struct route_entry *re, *next; | |
3003 | bool re_changed = false; | |
3004 | ||
3005 | RNODE_FOREACH_RE_SAFE (rn, re, next) { | |
3006 | if (type == ZEBRA_ROUTE_ALL || type == re->type) { | |
3007 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
3008 | re_changed = true; | |
3009 | } | |
3010 | } | |
3011 | ||
3012 | if (re_changed) | |
3013 | rib_queue_add(rn); | |
3014 | } | |
3015 | ||
1c848137 | 3016 | /* Schedule routes of a particular table (address-family) based on event. */ |
d5b8c216 | 3017 | void rib_update_table(struct route_table *table, rib_update_event_t event) |
d62a17ae | 3018 | { |
3019 | struct route_node *rn; | |
2a99ab95 SW |
3020 | |
3021 | if (IS_ZEBRA_DEBUG_EVENT) { | |
3022 | struct zebra_vrf *zvrf; | |
3023 | struct vrf *vrf; | |
3024 | ||
3025 | zvrf = table->info ? ((rib_table_info_t *)table->info)->zvrf | |
3026 | : NULL; | |
3027 | vrf = zvrf ? zvrf->vrf : NULL; | |
3028 | ||
3029 | zlog_debug("%s: %s VRF %s Table %u event %s", __func__, | |
3030 | table->info ? afi2str( | |
3031 | ((rib_table_info_t *)table->info)->afi) | |
3032 | : "Unknown", | |
3033 | vrf ? vrf->name : "Unknown", | |
3034 | zvrf ? zvrf->table_id : 0, | |
3035 | rib_update_event2str(event)); | |
3036 | } | |
d62a17ae | 3037 | |
3038 | /* Walk all routes and queue for processing, if appropriate for | |
3039 | * the trigger event. | |
3040 | */ | |
3041 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
1ca60f2c DS |
3042 | /* |
3043 | * If we are looking at a route node and the node | |
3044 | * has already been queued we don't | |
3045 | * need to queue it up again | |
3046 | */ | |
2a99ab95 SW |
3047 | if (rn->info |
3048 | && CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
3049 | RIB_ROUTE_ANY_QUEUED)) | |
1ca60f2c | 3050 | continue; |
2a99ab95 | 3051 | |
d62a17ae | 3052 | switch (event) { |
2a99ab95 SW |
3053 | case RIB_UPDATE_KERNEL: |
3054 | rib_update_route_node(rn, ZEBRA_ROUTE_KERNEL); | |
3055 | break; | |
d62a17ae | 3056 | case RIB_UPDATE_RMAP_CHANGE: |
3057 | case RIB_UPDATE_OTHER: | |
2a99ab95 | 3058 | rib_update_route_node(rn, ZEBRA_ROUTE_ALL); |
d62a17ae | 3059 | break; |
d62a17ae | 3060 | default: |
3061 | break; | |
3062 | } | |
3063 | } | |
b84c7253 | 3064 | } |
3065 | ||
2a99ab95 | 3066 | static void rib_update_handle_vrf(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 3067 | { |
d62a17ae | 3068 | struct route_table *table; |
1c848137 | 3069 | |
2a99ab95 SW |
3070 | if (IS_ZEBRA_DEBUG_EVENT) |
3071 | zlog_debug("%s: Handling VRF %s event %s", __func__, | |
3072 | vrf_id_to_name(vrf_id), rib_update_event2str(event)); | |
3073 | ||
d62a17ae | 3074 | /* Process routes of interested address-families. */ |
3075 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2a99ab95 | 3076 | if (table) |
d62a17ae | 3077 | rib_update_table(table, event); |
718e3744 | 3078 | |
d62a17ae | 3079 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
2a99ab95 | 3080 | if (table) |
d62a17ae | 3081 | rib_update_table(table, event); |
2a99ab95 SW |
3082 | } |
3083 | ||
3084 | static void rib_update_handle_vrf_all(rib_update_event_t event) | |
3085 | { | |
3086 | struct zebra_router_table *zrt; | |
3087 | ||
3088 | if (IS_ZEBRA_DEBUG_EVENT) | |
3089 | zlog_debug("%s: Handling VRF (ALL) event %s", __func__, | |
3090 | rib_update_event2str(event)); | |
3091 | ||
3092 | /* Just iterate over all the route tables, rather than vrf lookups */ | |
3093 | RB_FOREACH (zrt, zebra_router_table_head, &zrouter.tables) | |
3094 | rib_update_table(zrt->table, event); | |
3095 | } | |
3096 | ||
3097 | struct rib_update_ctx { | |
3098 | rib_update_event_t event; | |
3099 | bool vrf_all; | |
3100 | vrf_id_t vrf_id; | |
3101 | }; | |
3102 | ||
3103 | static struct rib_update_ctx *rib_update_ctx_init(vrf_id_t vrf_id, | |
3104 | rib_update_event_t event) | |
3105 | { | |
3106 | struct rib_update_ctx *ctx; | |
3107 | ||
3108 | ctx = XCALLOC(MTYPE_RIB_UPDATE_CTX, sizeof(struct rib_update_ctx)); | |
3109 | ||
3110 | ctx->event = event; | |
3111 | ctx->vrf_id = vrf_id; | |
3112 | ||
3113 | return ctx; | |
3114 | } | |
3115 | ||
3116 | static void rib_update_ctx_fini(struct rib_update_ctx **ctx) | |
3117 | { | |
3118 | XFREE(MTYPE_RIB_UPDATE_CTX, *ctx); | |
2a99ab95 SW |
3119 | } |
3120 | ||
3121 | static int rib_update_handler(struct thread *thread) | |
3122 | { | |
3123 | struct rib_update_ctx *ctx; | |
3124 | ||
3125 | ctx = THREAD_ARG(thread); | |
3126 | ||
3127 | if (ctx->vrf_all) | |
3128 | rib_update_handle_vrf_all(ctx->event); | |
3129 | else | |
3130 | rib_update_handle_vrf(ctx->vrf_id, ctx->event); | |
3131 | ||
3132 | rib_update_ctx_fini(&ctx); | |
3133 | ||
3134 | return 0; | |
3135 | } | |
3136 | ||
3137 | /* | |
3138 | * Thread list to ensure we don't schedule a ton of events | |
3139 | * if interfaces are flapping for instance. | |
3140 | */ | |
3141 | static struct thread *t_rib_update_threads[RIB_UPDATE_MAX]; | |
3142 | ||
3143 | /* Schedule a RIB update event for specific vrf */ | |
3144 | void rib_update_vrf(vrf_id_t vrf_id, rib_update_event_t event) | |
3145 | { | |
3146 | struct rib_update_ctx *ctx; | |
3147 | ||
3148 | ctx = rib_update_ctx_init(vrf_id, event); | |
3149 | ||
3150 | /* Don't worry about making sure multiple rib updates for specific vrf | |
3151 | * are scheduled at once for now. If it becomes a problem, we can use a | |
3152 | * lookup of some sort to keep track of running threads via t_vrf_id | |
3153 | * like how we are doing it in t_rib_update_threads[]. | |
3154 | */ | |
3155 | thread_add_event(zrouter.master, rib_update_handler, ctx, 0, NULL); | |
3156 | ||
3157 | if (IS_ZEBRA_DEBUG_EVENT) | |
3158 | zlog_debug("%s: Scheduled VRF %s, event %s", __func__, | |
3159 | vrf_id_to_name(ctx->vrf_id), | |
3160 | rib_update_event2str(event)); | |
3161 | } | |
3162 | ||
3163 | /* Schedule a RIB update event for all vrfs */ | |
3164 | void rib_update(rib_update_event_t event) | |
3165 | { | |
3166 | struct rib_update_ctx *ctx; | |
3167 | ||
3168 | ctx = rib_update_ctx_init(0, event); | |
3169 | ||
3170 | ctx->vrf_all = true; | |
3171 | ||
3172 | if (!thread_add_event(zrouter.master, rib_update_handler, ctx, 0, | |
3173 | &t_rib_update_threads[event])) | |
3174 | rib_update_ctx_fini(&ctx); /* Already scheduled */ | |
3175 | else if (IS_ZEBRA_DEBUG_EVENT) | |
05267678 | 3176 | zlog_debug("%s: Scheduled VRF (ALL), event %s", __func__, |
2a99ab95 | 3177 | rib_update_event2str(event)); |
718e3744 | 3178 | } |
3179 | ||
718e3744 | 3180 | /* Delete self installed routes after zebra is relaunched. */ |
95a29032 | 3181 | void rib_sweep_table(struct route_table *table) |
d62a17ae | 3182 | { |
3183 | struct route_node *rn; | |
3184 | struct route_entry *re; | |
3185 | struct route_entry *next; | |
915902cb | 3186 | struct nexthop *nexthop; |
d62a17ae | 3187 | |
915902cb DS |
3188 | if (!table) |
3189 | return; | |
d62a17ae | 3190 | |
915902cb | 3191 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
a2addae8 | 3192 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d4644d41 | 3193 | |
915902cb DS |
3194 | if (IS_ZEBRA_DEBUG_RIB) |
3195 | route_entry_dump(&rn->p, NULL, re); | |
3196 | ||
3197 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
3198 | continue; | |
3199 | ||
3200 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
3201 | continue; | |
3202 | ||
d4644d41 DS |
3203 | /* |
3204 | * If routes are older than startup_time then | |
3205 | * we know we read them in from the kernel. | |
3206 | * As such we can safely remove them. | |
3207 | */ | |
3208 | if (zrouter.startup_time < re->uptime) | |
3209 | continue; | |
3210 | ||
915902cb DS |
3211 | /* |
3212 | * So we are starting up and have received | |
3213 | * routes from the kernel that we have installed | |
3214 | * from a previous run of zebra but not cleaned | |
3215 | * up ( say a kill -9 ) | |
3216 | * But since we haven't actually installed | |
3217 | * them yet( we received them from the kernel ) | |
3218 | * we don't think they are active. | |
3219 | * So let's pretend they are active to actually | |
3220 | * remove them. | |
3221 | * In all honesty I'm not sure if we should | |
3222 | * mark them as active when we receive them | |
3223 | * This is startup only so probably ok. | |
3224 | * | |
3225 | * If we ever decide to move rib_sweep_table | |
3226 | * to a different spot (ie startup ) | |
3227 | * this decision needs to be revisited | |
3228 | */ | |
677c1dd5 | 3229 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
c415d895 | 3230 | for (ALL_NEXTHOPS(re->nhe->nhg, nexthop)) |
915902cb DS |
3231 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
3232 | ||
0c555cc6 DS |
3233 | rib_uninstall_kernel(rn, re); |
3234 | rib_delnode(rn, re); | |
915902cb DS |
3235 | } |
3236 | } | |
718e3744 | 3237 | } |
3238 | ||
3239 | /* Sweep all RIB tables. */ | |
d4644d41 | 3240 | int rib_sweep_route(struct thread *t) |
718e3744 | 3241 | { |
d62a17ae | 3242 | struct vrf *vrf; |
3243 | struct zebra_vrf *zvrf; | |
78104b9b | 3244 | |
a2addae8 | 3245 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
915902cb DS |
3246 | if ((zvrf = vrf->info) == NULL) |
3247 | continue; | |
3248 | ||
d62a17ae | 3249 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); |
3250 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3251 | } | |
95a29032 | 3252 | |
89272910 | 3253 | zebra_router_sweep_route(); |
38e40db1 | 3254 | zebra_router_sweep_nhgs(); |
d4644d41 DS |
3255 | |
3256 | return 0; | |
718e3744 | 3257 | } |
2ea1ab1c VT |
3258 | |
3259 | /* Remove specific by protocol routes from 'table'. */ | |
d7c0a89a | 3260 | unsigned long rib_score_proto_table(uint8_t proto, unsigned short instance, |
47a08aa9 | 3261 | struct route_table *table) |
d62a17ae | 3262 | { |
3263 | struct route_node *rn; | |
3264 | struct route_entry *re; | |
3265 | struct route_entry *next; | |
3266 | unsigned long n = 0; | |
3267 | ||
3268 | if (table) | |
3269 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
a2addae8 | 3270 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 3271 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
3272 | continue; | |
3273 | if (re->type == proto | |
3274 | && re->instance == instance) { | |
3275 | rib_delnode(rn, re); | |
3276 | n++; | |
3277 | } | |
3278 | } | |
3279 | return n; | |
2ea1ab1c VT |
3280 | } |
3281 | ||
3282 | /* Remove specific by protocol routes. */ | |
d7c0a89a | 3283 | unsigned long rib_score_proto(uint8_t proto, unsigned short instance) |
2ea1ab1c | 3284 | { |
d62a17ae | 3285 | struct vrf *vrf; |
3286 | struct zebra_vrf *zvrf; | |
d8612e65 | 3287 | struct other_route_table *ort; |
d62a17ae | 3288 | unsigned long cnt = 0; |
78104b9b | 3289 | |
d8612e65 DS |
3290 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
3291 | zvrf = vrf->info; | |
3292 | if (!zvrf) | |
3293 | continue; | |
78104b9b | 3294 | |
d8612e65 DS |
3295 | cnt += rib_score_proto_table(proto, instance, |
3296 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
3297 | + rib_score_proto_table( | |
3298 | proto, instance, | |
3299 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3300 | ||
81fddbe7 | 3301 | frr_each(otable, &zvrf->other_tables, ort) cnt += |
d8612e65 DS |
3302 | rib_score_proto_table(proto, instance, ort->table); |
3303 | } | |
47a08aa9 | 3304 | |
d62a17ae | 3305 | return cnt; |
2ea1ab1c VT |
3306 | } |
3307 | ||
718e3744 | 3308 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 3309 | void rib_close_table(struct route_table *table) |
718e3744 | 3310 | { |
d62a17ae | 3311 | struct route_node *rn; |
1e9f448f | 3312 | rib_table_info_t *info; |
5f7a4718 | 3313 | rib_dest_t *dest; |
718e3744 | 3314 | |
1e9f448f DS |
3315 | if (!table) |
3316 | return; | |
9fd92e3c | 3317 | |
6ca30e9e | 3318 | info = route_table_get_info(table); |
5adc2528 | 3319 | |
5f7a4718 DS |
3320 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
3321 | dest = rib_dest_from_rnode(rn); | |
1e9f448f | 3322 | |
5f7a4718 | 3323 | if (dest && dest->selected_fib) { |
1e9f448f DS |
3324 | if (info->safi == SAFI_UNICAST) |
3325 | hook_call(rib_update, rn, NULL); | |
3326 | ||
3cdba47a DS |
3327 | rib_uninstall_kernel(rn, dest->selected_fib); |
3328 | dest->selected_fib = NULL; | |
1e9f448f | 3329 | } |
5f7a4718 | 3330 | } |
718e3744 | 3331 | } |
3332 | ||
9bd9717b MS |
3333 | /* |
3334 | * Handler for async dataplane results after a pseudowire installation | |
3335 | */ | |
3336 | static int handle_pw_result(struct zebra_dplane_ctx *ctx) | |
3337 | { | |
9bd9717b MS |
3338 | struct zebra_pw *pw; |
3339 | struct zebra_vrf *vrf; | |
3340 | ||
3341 | /* The pseudowire code assumes success - we act on an error | |
3342 | * result for installation attempts here. | |
3343 | */ | |
3344 | if (dplane_ctx_get_op(ctx) != DPLANE_OP_PW_INSTALL) | |
3345 | goto done; | |
3346 | ||
3347 | if (dplane_ctx_get_status(ctx) != ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
3348 | vrf = zebra_vrf_lookup_by_id(dplane_ctx_get_vrf(ctx)); | |
7c7ef4a8 | 3349 | pw = zebra_pw_find(vrf, dplane_ctx_get_ifname(ctx)); |
9bd9717b MS |
3350 | if (pw) |
3351 | zebra_pw_install_failure(pw); | |
3352 | } | |
3353 | ||
3354 | done: | |
3355 | ||
9f2d0354 | 3356 | return 0; |
9bd9717b MS |
3357 | } |
3358 | ||
3359 | ||
7cdb1a84 | 3360 | /* |
d37f4d6c MS |
3361 | * Handle results from the dataplane system. Dequeue update context |
3362 | * structs, dispatch to appropriate internal handlers. | |
7cdb1a84 MS |
3363 | */ |
3364 | static int rib_process_dplane_results(struct thread *thread) | |
3365 | { | |
25779064 | 3366 | struct zebra_dplane_ctx *ctx; |
4c206c8f | 3367 | struct dplane_ctx_q ctxlist; |
2fc69f03 | 3368 | bool shut_p = false; |
4c206c8f MS |
3369 | |
3370 | /* Dequeue a list of completed updates with one lock/unlock cycle */ | |
7cdb1a84 MS |
3371 | |
3372 | do { | |
4c206c8f MS |
3373 | TAILQ_INIT(&ctxlist); |
3374 | ||
7cdb1a84 | 3375 | /* Take lock controlling queue of results */ |
00dffa8c | 3376 | frr_with_mutex(&dplane_mutex) { |
d37f4d6c | 3377 | /* Dequeue list of context structs */ |
4c206c8f | 3378 | dplane_ctx_list_append(&ctxlist, &rib_dplane_q); |
7cdb1a84 | 3379 | } |
7cdb1a84 | 3380 | |
4c206c8f MS |
3381 | /* Dequeue context block */ |
3382 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3383 | ||
3384 | /* If we've emptied the results queue, we're done */ | |
3385 | if (ctx == NULL) | |
7cdb1a84 | 3386 | break; |
7cdb1a84 | 3387 | |
2fc69f03 MS |
3388 | /* If zebra is shutting down, avoid processing results, |
3389 | * just drain the results queue. | |
3390 | */ | |
3391 | shut_p = atomic_load_explicit(&zrouter.in_shutdown, | |
3392 | memory_order_relaxed); | |
3393 | if (shut_p) { | |
3394 | while (ctx) { | |
3395 | dplane_ctx_fini(&ctx); | |
3396 | ||
3397 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3398 | } | |
3399 | ||
3400 | continue; | |
3401 | } | |
3402 | ||
4c206c8f | 3403 | while (ctx) { |
d37f4d6c MS |
3404 | switch (dplane_ctx_get_op(ctx)) { |
3405 | case DPLANE_OP_ROUTE_INSTALL: | |
3406 | case DPLANE_OP_ROUTE_UPDATE: | |
3407 | case DPLANE_OP_ROUTE_DELETE: | |
188a00e0 MS |
3408 | { |
3409 | /* Bit of special case for route updates | |
3410 | * that were generated by async notifications: | |
3411 | * we don't want to continue processing these | |
3412 | * in the rib. | |
3413 | */ | |
3414 | if (dplane_ctx_get_notif_provider(ctx) == 0) | |
3415 | rib_process_result(ctx); | |
3416 | else | |
3417 | dplane_ctx_fini(&ctx); | |
3418 | } | |
3419 | break; | |
d37f4d6c | 3420 | |
54818e3b MS |
3421 | case DPLANE_OP_ROUTE_NOTIFY: |
3422 | rib_process_dplane_notify(ctx); | |
d37f4d6c MS |
3423 | break; |
3424 | ||
9865e1c3 SW |
3425 | case DPLANE_OP_NH_INSTALL: |
3426 | case DPLANE_OP_NH_UPDATE: | |
3427 | case DPLANE_OP_NH_DELETE: | |
3428 | zebra_nhg_dplane_result(ctx); | |
3429 | break; | |
3430 | ||
d37f4d6c MS |
3431 | case DPLANE_OP_LSP_INSTALL: |
3432 | case DPLANE_OP_LSP_UPDATE: | |
3433 | case DPLANE_OP_LSP_DELETE: | |
188a00e0 MS |
3434 | { |
3435 | /* Bit of special case for LSP updates | |
3436 | * that were generated by async notifications: | |
3437 | * we don't want to continue processing these. | |
3438 | */ | |
3439 | if (dplane_ctx_get_notif_provider(ctx) == 0) | |
3440 | zebra_mpls_lsp_dplane_result(ctx); | |
3441 | else | |
3442 | dplane_ctx_fini(&ctx); | |
3443 | } | |
3444 | break; | |
d37f4d6c | 3445 | |
104e3ad9 MS |
3446 | case DPLANE_OP_LSP_NOTIFY: |
3447 | zebra_mpls_process_dplane_notify(ctx); | |
d37f4d6c MS |
3448 | break; |
3449 | ||
9bd9717b MS |
3450 | case DPLANE_OP_PW_INSTALL: |
3451 | case DPLANE_OP_PW_UNINSTALL: | |
3452 | handle_pw_result(ctx); | |
3453 | break; | |
3454 | ||
cf363e1b MS |
3455 | case DPLANE_OP_SYS_ROUTE_ADD: |
3456 | case DPLANE_OP_SYS_ROUTE_DELETE: | |
3457 | /* No further processing in zebra for these. */ | |
3458 | dplane_ctx_fini(&ctx); | |
3459 | break; | |
3460 | ||
036d93c0 MS |
3461 | case DPLANE_OP_MAC_INSTALL: |
3462 | case DPLANE_OP_MAC_DELETE: | |
3463 | zebra_vxlan_handle_result(ctx); | |
3464 | break; | |
3465 | ||
931fa60c MS |
3466 | /* Some op codes not handled here */ |
3467 | case DPLANE_OP_ADDR_INSTALL: | |
3468 | case DPLANE_OP_ADDR_UNINSTALL: | |
3469 | case DPLANE_OP_NEIGH_INSTALL: | |
3470 | case DPLANE_OP_NEIGH_UPDATE: | |
3471 | case DPLANE_OP_NEIGH_DELETE: | |
0bbd4ff4 MS |
3472 | case DPLANE_OP_VTEP_ADD: |
3473 | case DPLANE_OP_VTEP_DELETE: | |
931fa60c | 3474 | case DPLANE_OP_NONE: |
d37f4d6c MS |
3475 | /* Don't expect this: just return the struct? */ |
3476 | dplane_ctx_fini(&ctx); | |
3477 | break; | |
931fa60c | 3478 | |
d37f4d6c | 3479 | } /* Dispatch by op code */ |
4c206c8f MS |
3480 | |
3481 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3482 | } | |
3483 | ||
5709131c | 3484 | } while (1); |
7cdb1a84 | 3485 | |
5709131c | 3486 | return 0; |
7cdb1a84 MS |
3487 | } |
3488 | ||
3489 | /* | |
3490 | * Results are returned from the dataplane subsystem, in the context of | |
1bcea841 | 3491 | * the dataplane pthread. We enqueue the results here for processing by |
7cdb1a84 MS |
3492 | * the main thread later. |
3493 | */ | |
4c206c8f | 3494 | static int rib_dplane_results(struct dplane_ctx_q *ctxlist) |
7cdb1a84 MS |
3495 | { |
3496 | /* Take lock controlling queue of results */ | |
00dffa8c | 3497 | frr_with_mutex(&dplane_mutex) { |
4c206c8f MS |
3498 | /* Enqueue context blocks */ |
3499 | dplane_ctx_list_append(&rib_dplane_q, ctxlist); | |
7cdb1a84 | 3500 | } |
7cdb1a84 | 3501 | |
4c206c8f | 3502 | /* Ensure event is signalled to zebra main pthread */ |
3801e764 | 3503 | thread_add_event(zrouter.master, rib_process_dplane_results, NULL, 0, |
7cdb1a84 MS |
3504 | &t_dplane); |
3505 | ||
5709131c | 3506 | return 0; |
7cdb1a84 MS |
3507 | } |
3508 | ||
eaa2716d SW |
3509 | /* |
3510 | * Ensure there are no empty slots in the route_info array. | |
3511 | * Every route type in zebra should be present there. | |
3512 | */ | |
3513 | static void check_route_info(void) | |
3514 | { | |
3515 | int len = array_size(route_info); | |
3516 | ||
3517 | /* | |
3518 | * ZEBRA_ROUTE_SYSTEM is special cased since | |
3519 | * its key is 0 anyway. | |
3520 | * | |
3521 | * ZEBRA_ROUTE_ALL is also ignored. | |
3522 | */ | |
3523 | for (int i = 0; i < len; i++) { | |
3524 | if (i == ZEBRA_ROUTE_SYSTEM || i == ZEBRA_ROUTE_ALL) | |
3525 | continue; | |
3526 | assert(route_info[i].key); | |
8dc7a759 | 3527 | assert(route_info[i].meta_q_map < MQ_SIZE); |
eaa2716d SW |
3528 | } |
3529 | } | |
3530 | ||
718e3744 | 3531 | /* Routing information base initialize. */ |
d62a17ae | 3532 | void rib_init(void) |
718e3744 | 3533 | { |
eaa2716d SW |
3534 | check_route_info(); |
3535 | ||
2561d12e | 3536 | rib_queue_init(); |
7cdb1a84 MS |
3537 | |
3538 | /* Init dataplane, and register for results */ | |
3539 | pthread_mutex_init(&dplane_mutex, NULL); | |
3540 | TAILQ_INIT(&rib_dplane_q); | |
4c206c8f | 3541 | zebra_dplane_init(rib_dplane_results); |
718e3744 | 3542 | } |
0915bb0c AS |
3543 | |
3544 | /* | |
3545 | * vrf_id_get_next | |
3546 | * | |
3547 | * Get the first vrf id that is greater than the given vrf id if any. | |
3548 | * | |
2951a7a4 | 3549 | * Returns true if a vrf id was found, false otherwise. |
0915bb0c | 3550 | */ |
d62a17ae | 3551 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 3552 | { |
d62a17ae | 3553 | struct vrf *vrf; |
b72ede27 | 3554 | |
d62a17ae | 3555 | vrf = vrf_lookup_by_id(vrf_id); |
3556 | if (vrf) { | |
3557 | vrf = RB_NEXT(vrf_id_head, vrf); | |
3558 | if (vrf) { | |
3559 | *next_id_p = vrf->vrf_id; | |
3560 | return 1; | |
3561 | } | |
3562 | } | |
0915bb0c | 3563 | |
d62a17ae | 3564 | return 0; |
0915bb0c AS |
3565 | } |
3566 | ||
3567 | /* | |
3568 | * rib_tables_iter_next | |
3569 | * | |
3570 | * Returns the next table in the iteration. | |
3571 | */ | |
d62a17ae | 3572 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
3573 | { | |
3574 | struct route_table *table; | |
3575 | ||
3576 | /* | |
3577 | * Array that helps us go over all AFI/SAFI combinations via one | |
3578 | * index. | |
3579 | */ | |
2b64873d | 3580 | static const struct { |
d62a17ae | 3581 | afi_t afi; |
3582 | safi_t safi; | |
3583 | } afi_safis[] = { | |
3584 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
3585 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
3586 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
3587 | }; | |
3588 | ||
3589 | table = NULL; | |
3590 | ||
3591 | switch (iter->state) { | |
3592 | ||
3593 | case RIB_TABLES_ITER_S_INIT: | |
3594 | iter->vrf_id = VRF_DEFAULT; | |
3595 | iter->afi_safi_ix = -1; | |
3596 | ||
3597 | /* Fall through */ | |
3598 | ||
3599 | case RIB_TABLES_ITER_S_ITERATING: | |
3600 | iter->afi_safi_ix++; | |
3601 | while (1) { | |
3602 | ||
3603 | while (iter->afi_safi_ix | |
7e3a1ec7 | 3604 | < (int)array_size(afi_safis)) { |
d62a17ae | 3605 | table = zebra_vrf_table( |
3606 | afi_safis[iter->afi_safi_ix].afi, | |
3607 | afi_safis[iter->afi_safi_ix].safi, | |
3608 | iter->vrf_id); | |
3609 | if (table) | |
3610 | break; | |
3611 | ||
3612 | iter->afi_safi_ix++; | |
3613 | } | |
3614 | ||
3615 | /* | |
3616 | * Found another table in this vrf. | |
3617 | */ | |
3618 | if (table) | |
3619 | break; | |
3620 | ||
3621 | /* | |
3622 | * Done with all tables in the current vrf, go to the | |
3623 | * next | |
3624 | * one. | |
3625 | */ | |
3626 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
3627 | break; | |
3628 | ||
3629 | iter->afi_safi_ix = 0; | |
3630 | } | |
0915bb0c | 3631 | |
0915bb0c AS |
3632 | break; |
3633 | ||
d62a17ae | 3634 | case RIB_TABLES_ITER_S_DONE: |
3635 | return NULL; | |
0915bb0c AS |
3636 | } |
3637 | ||
d62a17ae | 3638 | if (table) |
3639 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
3640 | else | |
3641 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 3642 | |
d62a17ae | 3643 | return table; |
0915bb0c | 3644 | } |