]>
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" | |
ad28e79a | 59 | #include "zebra/zebra_nhg.h" |
7cdb1a84 | 60 | |
2a99ab95 SW |
61 | DEFINE_MTYPE_STATIC(ZEBRA, RIB_UPDATE_CTX, "Rib update context object"); |
62 | ||
7cdb1a84 MS |
63 | /* |
64 | * Event, list, and mutex for delivery of dataplane results | |
65 | */ | |
66 | static pthread_mutex_t dplane_mutex; | |
67 | static struct thread *t_dplane; | |
25779064 | 68 | static struct dplane_ctx_q rib_dplane_q; |
718e3744 | 69 | |
d62a17ae | 70 | DEFINE_HOOK(rib_update, (struct route_node * rn, const char *reason), |
71 | (rn, reason)) | |
4f8ea50c | 72 | |
6baf7bb8 DS |
73 | /* Should we allow non Quagga processes to delete our routes */ |
74 | extern int allow_delete; | |
75 | ||
718e3744 | 76 | /* Each route type's string and default distance value. */ |
d62a17ae | 77 | static const struct { |
78 | int key; | |
8dc7a759 | 79 | uint8_t distance; |
73547a75 | 80 | uint8_t meta_q_map; |
d62a17ae | 81 | } route_info[ZEBRA_ROUTE_MAX] = { |
73547a75 DS |
82 | [ZEBRA_ROUTE_SYSTEM] = {ZEBRA_ROUTE_SYSTEM, 0, 4}, |
83 | [ZEBRA_ROUTE_KERNEL] = {ZEBRA_ROUTE_KERNEL, 0, 0}, | |
84 | [ZEBRA_ROUTE_CONNECT] = {ZEBRA_ROUTE_CONNECT, 0, 0}, | |
85 | [ZEBRA_ROUTE_STATIC] = {ZEBRA_ROUTE_STATIC, 1, 1}, | |
86 | [ZEBRA_ROUTE_RIP] = {ZEBRA_ROUTE_RIP, 120, 2}, | |
87 | [ZEBRA_ROUTE_RIPNG] = {ZEBRA_ROUTE_RIPNG, 120, 2}, | |
88 | [ZEBRA_ROUTE_OSPF] = {ZEBRA_ROUTE_OSPF, 110, 2}, | |
89 | [ZEBRA_ROUTE_OSPF6] = {ZEBRA_ROUTE_OSPF6, 110, 2}, | |
90 | [ZEBRA_ROUTE_ISIS] = {ZEBRA_ROUTE_ISIS, 115, 2}, | |
91 | [ZEBRA_ROUTE_BGP] = {ZEBRA_ROUTE_BGP, 20 /* IBGP is 200. */, 3}, | |
92 | [ZEBRA_ROUTE_PIM] = {ZEBRA_ROUTE_PIM, 255, 4}, | |
93 | [ZEBRA_ROUTE_EIGRP] = {ZEBRA_ROUTE_EIGRP, 90, 2}, | |
94 | [ZEBRA_ROUTE_NHRP] = {ZEBRA_ROUTE_NHRP, 10, 2}, | |
95 | [ZEBRA_ROUTE_HSLS] = {ZEBRA_ROUTE_HSLS, 255, 4}, | |
96 | [ZEBRA_ROUTE_OLSR] = {ZEBRA_ROUTE_OLSR, 255, 4}, | |
97 | [ZEBRA_ROUTE_TABLE] = {ZEBRA_ROUTE_TABLE, 150, 1}, | |
98 | [ZEBRA_ROUTE_LDP] = {ZEBRA_ROUTE_LDP, 150, 4}, | |
99 | [ZEBRA_ROUTE_VNC] = {ZEBRA_ROUTE_VNC, 20, 3}, | |
100 | [ZEBRA_ROUTE_VNC_DIRECT] = {ZEBRA_ROUTE_VNC_DIRECT, 20, 3}, | |
101 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = {ZEBRA_ROUTE_VNC_DIRECT_RH, 20, 3}, | |
102 | [ZEBRA_ROUTE_BGP_DIRECT] = {ZEBRA_ROUTE_BGP_DIRECT, 20, 3}, | |
103 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = {ZEBRA_ROUTE_BGP_DIRECT_EXT, 20, 3}, | |
104 | [ZEBRA_ROUTE_BABEL] = {ZEBRA_ROUTE_BABEL, 100, 2}, | |
105 | [ZEBRA_ROUTE_SHARP] = {ZEBRA_ROUTE_SHARP, 150, 4}, | |
98156652 | 106 | [ZEBRA_ROUTE_PBR] = {ZEBRA_ROUTE_PBR, 200, 4}, |
42d96b73 | 107 | [ZEBRA_ROUTE_BFD] = {ZEBRA_ROUTE_BFD, 255, 4}, |
eab7b6e3 | 108 | [ZEBRA_ROUTE_OPENFABRIC] = {ZEBRA_ROUTE_OPENFABRIC, 115, 2}, |
303b93cd | 109 | [ZEBRA_ROUTE_VRRP] = {ZEBRA_ROUTE_VRRP, 255, 4} |
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 | ||
fa713d9e | 196 | /* Add nexthop to the end of a rib node's nexthop list */ |
d62a17ae | 197 | void route_entry_nexthop_add(struct route_entry *re, struct nexthop *nexthop) |
fa713d9e | 198 | { |
6b468511 | 199 | _nexthop_group_add_sorted(re->ng, nexthop); |
718e3744 | 200 | } |
201 | ||
6e26278c | 202 | |
6e26278c DS |
203 | /** |
204 | * copy_nexthop - copy a nexthop to the rib structure. | |
205 | */ | |
d62a17ae | 206 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) |
6e26278c | 207 | { |
6b468511 DS |
208 | assert(!re->ng->nexthop); |
209 | copy_nexthops(&re->ng->nexthop, nh, NULL); | |
6e26278c DS |
210 | } |
211 | ||
718e3744 | 212 | /* Delete specified nexthop from the list. */ |
d62a17ae | 213 | void route_entry_nexthop_delete(struct route_entry *re, struct nexthop *nexthop) |
718e3744 | 214 | { |
d62a17ae | 215 | if (nexthop->next) |
216 | nexthop->next->prev = nexthop->prev; | |
217 | if (nexthop->prev) | |
218 | nexthop->prev->next = nexthop->next; | |
219 | else | |
6b468511 | 220 | re->ng->nexthop = nexthop->next; |
718e3744 | 221 | } |
222 | ||
fa713d9e | 223 | |
d62a17ae | 224 | struct nexthop *route_entry_nexthop_ifindex_add(struct route_entry *re, |
4a7371e9 DS |
225 | ifindex_t ifindex, |
226 | vrf_id_t nh_vrf_id) | |
718e3744 | 227 | { |
d62a17ae | 228 | struct nexthop *nexthop; |
718e3744 | 229 | |
d62a17ae | 230 | nexthop = nexthop_new(); |
231 | nexthop->type = NEXTHOP_TYPE_IFINDEX; | |
232 | nexthop->ifindex = ifindex; | |
4a7371e9 | 233 | nexthop->vrf_id = nh_vrf_id; |
718e3744 | 234 | |
d62a17ae | 235 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 236 | |
d62a17ae | 237 | return nexthop; |
718e3744 | 238 | } |
239 | ||
d62a17ae | 240 | struct nexthop *route_entry_nexthop_ipv4_add(struct route_entry *re, |
241 | struct in_addr *ipv4, | |
4a7371e9 DS |
242 | struct in_addr *src, |
243 | vrf_id_t nh_vrf_id) | |
718e3744 | 244 | { |
d62a17ae | 245 | struct nexthop *nexthop; |
718e3744 | 246 | |
d62a17ae | 247 | nexthop = nexthop_new(); |
248 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
4a7371e9 | 249 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 250 | nexthop->gate.ipv4 = *ipv4; |
251 | if (src) | |
252 | nexthop->src.ipv4 = *src; | |
718e3744 | 253 | |
d62a17ae | 254 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 255 | |
d62a17ae | 256 | return nexthop; |
718e3744 | 257 | } |
258 | ||
d62a17ae | 259 | struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re, |
260 | struct in_addr *ipv4, | |
261 | struct in_addr *src, | |
4a7371e9 DS |
262 | ifindex_t ifindex, |
263 | vrf_id_t nh_vrf_id) | |
718e3744 | 264 | { |
d62a17ae | 265 | struct nexthop *nexthop; |
266 | struct interface *ifp; | |
718e3744 | 267 | |
d62a17ae | 268 | nexthop = nexthop_new(); |
4a7371e9 | 269 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 270 | nexthop->type = NEXTHOP_TYPE_IPV4_IFINDEX; |
271 | nexthop->gate.ipv4 = *ipv4; | |
272 | if (src) | |
273 | nexthop->src.ipv4 = *src; | |
274 | nexthop->ifindex = ifindex; | |
4a7371e9 | 275 | ifp = if_lookup_by_index(nexthop->ifindex, nh_vrf_id); |
d62a17ae | 276 | /*Pending: need to think if null ifp here is ok during bootup? |
277 | There was a crash because ifp here was coming to be NULL */ | |
278 | if (ifp) | |
2b83602b | 279 | if (connected_is_unnumbered(ifp)) |
d62a17ae | 280 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); |
718e3744 | 281 | |
d62a17ae | 282 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 283 | |
d62a17ae | 284 | return nexthop; |
718e3744 | 285 | } |
286 | ||
d62a17ae | 287 | struct nexthop *route_entry_nexthop_ipv6_add(struct route_entry *re, |
4a7371e9 DS |
288 | struct in6_addr *ipv6, |
289 | vrf_id_t nh_vrf_id) | |
718e3744 | 290 | { |
d62a17ae | 291 | struct nexthop *nexthop; |
718e3744 | 292 | |
d62a17ae | 293 | nexthop = nexthop_new(); |
4a7371e9 | 294 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 295 | nexthop->type = NEXTHOP_TYPE_IPV6; |
296 | nexthop->gate.ipv6 = *ipv6; | |
718e3744 | 297 | |
d62a17ae | 298 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 299 | |
d62a17ae | 300 | return nexthop; |
718e3744 | 301 | } |
302 | ||
d62a17ae | 303 | struct nexthop *route_entry_nexthop_ipv6_ifindex_add(struct route_entry *re, |
304 | struct in6_addr *ipv6, | |
4a7371e9 DS |
305 | ifindex_t ifindex, |
306 | vrf_id_t nh_vrf_id) | |
718e3744 | 307 | { |
d62a17ae | 308 | struct nexthop *nexthop; |
718e3744 | 309 | |
d62a17ae | 310 | nexthop = nexthop_new(); |
4a7371e9 | 311 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 312 | nexthop->type = NEXTHOP_TYPE_IPV6_IFINDEX; |
313 | nexthop->gate.ipv6 = *ipv6; | |
314 | nexthop->ifindex = ifindex; | |
718e3744 | 315 | |
d62a17ae | 316 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 317 | |
d62a17ae | 318 | return nexthop; |
718e3744 | 319 | } |
718e3744 | 320 | |
a8309422 | 321 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re, |
60466a63 | 322 | enum blackhole_type bh_type) |
595db7f1 | 323 | { |
d62a17ae | 324 | struct nexthop *nexthop; |
595db7f1 | 325 | |
d62a17ae | 326 | nexthop = nexthop_new(); |
4a7371e9 | 327 | nexthop->vrf_id = VRF_DEFAULT; |
d62a17ae | 328 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; |
a8309422 | 329 | nexthop->bh_type = bh_type; |
595db7f1 | 330 | |
d62a17ae | 331 | route_entry_nexthop_add(re, nexthop); |
595db7f1 | 332 | |
d62a17ae | 333 | return nexthop; |
595db7f1 | 334 | } |
335 | ||
d62a17ae | 336 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
337 | union g_addr *addr, struct route_node **rn_out) | |
338 | { | |
339 | struct prefix p; | |
340 | struct route_table *table; | |
341 | struct route_node *rn; | |
5f7a4718 | 342 | struct route_entry *match = NULL; |
d62a17ae | 343 | |
344 | /* Lookup table. */ | |
345 | table = zebra_vrf_table(afi, safi, vrf_id); | |
346 | if (!table) | |
347 | return 0; | |
348 | ||
349 | memset(&p, 0, sizeof(struct prefix)); | |
350 | p.family = afi; | |
351 | if (afi == AFI_IP) { | |
352 | p.u.prefix4 = addr->ipv4; | |
353 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
354 | } else { | |
355 | p.u.prefix6 = addr->ipv6; | |
356 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
357 | } | |
718e3744 | 358 | |
d62a17ae | 359 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 360 | |
d62a17ae | 361 | while (rn) { |
5f7a4718 DS |
362 | rib_dest_t *dest; |
363 | ||
d62a17ae | 364 | route_unlock_node(rn); |
718e3744 | 365 | |
5f7a4718 | 366 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
367 | if (dest && dest->selected_fib |
368 | && !CHECK_FLAG(dest->selected_fib->status, | |
369 | ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 370 | match = dest->selected_fib; |
718e3744 | 371 | |
d62a17ae | 372 | /* If there is no selected route or matched route is EGP, go up |
373 | tree. */ | |
374 | if (!match) { | |
375 | do { | |
376 | rn = rn->parent; | |
377 | } while (rn && rn->info == NULL); | |
378 | if (rn) | |
379 | route_lock_node(rn); | |
380 | } else { | |
381 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
677c1dd5 DS |
382 | if (!CHECK_FLAG(match->status, |
383 | ROUTE_ENTRY_INSTALLED)) | |
d62a17ae | 384 | return NULL; |
385 | } | |
386 | ||
387 | if (rn_out) | |
388 | *rn_out = rn; | |
389 | return match; | |
390 | } | |
16814f96 | 391 | } |
d62a17ae | 392 | return NULL; |
393 | } | |
394 | ||
395 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
396 | struct in_addr addr, | |
397 | struct route_node **rn_out) | |
398 | { | |
399 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
400 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
401 | union g_addr gaddr = {.ipv4 = addr}; | |
402 | ||
526052fb | 403 | switch (zrouter.ipv4_multicast_mode) { |
d62a17ae | 404 | case MCAST_MRIB_ONLY: |
405 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
406 | rn_out); | |
407 | case MCAST_URIB_ONLY: | |
408 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
409 | case MCAST_NO_CONFIG: | |
410 | case MCAST_MIX_MRIB_FIRST: | |
411 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
412 | &m_rn); | |
413 | if (!mre) | |
414 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
415 | &gaddr, &u_rn); | |
416 | break; | |
417 | case MCAST_MIX_DISTANCE: | |
418 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
419 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
420 | if (mre && ure) | |
421 | re = ure->distance < mre->distance ? ure : mre; | |
422 | else if (mre) | |
423 | re = mre; | |
424 | else if (ure) | |
425 | re = ure; | |
426 | break; | |
427 | case MCAST_MIX_PFXLEN: | |
428 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
429 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
430 | if (mre && ure) | |
431 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
432 | else if (mre) | |
433 | re = mre; | |
434 | else if (ure) | |
435 | re = ure; | |
436 | break; | |
718e3744 | 437 | } |
33550aa8 | 438 | |
d62a17ae | 439 | if (rn_out) |
440 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 441 | |
d62a17ae | 442 | if (IS_ZEBRA_DEBUG_RIB) { |
443 | char buf[BUFSIZ]; | |
444 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 445 | |
32391aff DS |
446 | zlog_debug("%s: %s: vrf: %u found %s, using %s", |
447 | __func__, buf, vrf_id, | |
d62a17ae | 448 | mre ? (ure ? "MRIB+URIB" : "MRIB") |
449 | : ure ? "URIB" : "nothing", | |
450 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
451 | } | |
452 | return re; | |
4623d897 DL |
453 | } |
454 | ||
d62a17ae | 455 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 456 | { |
d62a17ae | 457 | struct route_table *table; |
458 | struct route_node *rn; | |
5f7a4718 | 459 | struct route_entry *match = NULL; |
5f7a4718 | 460 | rib_dest_t *dest; |
718e3744 | 461 | |
d62a17ae | 462 | /* Lookup table. */ |
463 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
464 | if (!table) | |
465 | return 0; | |
718e3744 | 466 | |
d62a17ae | 467 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 468 | |
d62a17ae | 469 | /* No route for this prefix. */ |
470 | if (!rn) | |
471 | return NULL; | |
718e3744 | 472 | |
d62a17ae | 473 | /* Unlock node. */ |
474 | route_unlock_node(rn); | |
5f7a4718 | 475 | dest = rib_dest_from_rnode(rn); |
718e3744 | 476 | |
996c9314 LB |
477 | if (dest && dest->selected_fib |
478 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 479 | match = dest->selected_fib; |
718e3744 | 480 | |
d62a17ae | 481 | if (!match) |
482 | return NULL; | |
718e3744 | 483 | |
d62a17ae | 484 | if (match->type == ZEBRA_ROUTE_CONNECT) |
485 | return match; | |
f9e1b38e | 486 | |
677c1dd5 DS |
487 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_INSTALLED)) |
488 | return match; | |
718e3744 | 489 | |
d62a17ae | 490 | return NULL; |
718e3744 | 491 | } |
492 | ||
a64448ba DS |
493 | /* |
494 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
495 | * (nexthops) must have a label. | |
496 | */ | |
d62a17ae | 497 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 498 | { |
d62a17ae | 499 | struct nexthop *nexthop = NULL; |
a64448ba | 500 | |
d62a17ae | 501 | if (re->type != ZEBRA_ROUTE_BGP) |
502 | return 0; | |
a64448ba | 503 | |
6b468511 | 504 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
d62a17ae | 505 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) |
506 | return 0; | |
6b0655a2 | 507 | |
d62a17ae | 508 | return 1; |
a64448ba | 509 | } |
718e3744 | 510 | |
6ae24471 DS |
511 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
512 | * is only used for IPv4. | |
513 | */ | |
0c555cc6 DS |
514 | void rib_install_kernel(struct route_node *rn, struct route_entry *re, |
515 | struct route_entry *old) | |
718e3744 | 516 | { |
d62a17ae | 517 | struct nexthop *nexthop; |
518 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 519 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
97f5b441 MS |
520 | const struct prefix *p, *src_p; |
521 | enum zebra_dplane_result ret; | |
522 | ||
523 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
718e3744 | 524 | |
d62a17ae | 525 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 526 | |
d62a17ae | 527 | if (info->safi != SAFI_UNICAST) { |
6b468511 | 528 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
d62a17ae | 529 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 530 | return; |
25b9cb0c DL |
531 | } else { |
532 | struct nexthop *prev; | |
533 | ||
6b468511 | 534 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) { |
25b9cb0c | 535 | UNSET_FLAG (nexthop->flags, NEXTHOP_FLAG_DUPLICATE); |
6b468511 | 536 | for (ALL_NEXTHOPS_PTR(re->ng, prev)) { |
25b9cb0c DL |
537 | if (prev == nexthop) |
538 | break; | |
996c9314 LB |
539 | if (nexthop_same_firsthop(nexthop, prev)) { |
540 | SET_FLAG(nexthop->flags, | |
541 | NEXTHOP_FLAG_DUPLICATE); | |
25b9cb0c DL |
542 | break; |
543 | } | |
544 | } | |
545 | } | |
d62a17ae | 546 | } |
718e3744 | 547 | |
2063a814 DS |
548 | /* |
549 | * If this is a replace to a new RE let the originator of the RE | |
550 | * know that they've lost | |
551 | */ | |
9a9f8926 | 552 | if (old && (old != re) && (old->type != re->type)) |
28610f7e | 553 | zsend_route_notify_owner(old, p, ZAPI_ROUTE_BETTER_ADMIN_WON); |
25b9cb0c | 554 | |
97f5b441 MS |
555 | /* Update fib selection */ |
556 | dest->selected_fib = re; | |
557 | ||
d62a17ae | 558 | /* |
559 | * Make sure we update the FPM any time we send new information to | |
560 | * the kernel. | |
561 | */ | |
562 | hook_call(rib_update, rn, "installing in kernel"); | |
97f5b441 MS |
563 | |
564 | /* Send add or update */ | |
8263d1d0 | 565 | if (old) |
97f5b441 | 566 | ret = dplane_route_update(rn, re, old); |
5709131c | 567 | else |
97f5b441 | 568 | ret = dplane_route_add(rn, re); |
97f5b441 MS |
569 | |
570 | switch (ret) { | |
ea1c14f6 | 571 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
60f98b23 | 572 | SET_FLAG(re->status, ROUTE_ENTRY_QUEUED); |
efe6c026 MS |
573 | |
574 | if (old) { | |
60f98b23 | 575 | SET_FLAG(old->status, ROUTE_ENTRY_QUEUED); |
efe6c026 MS |
576 | |
577 | /* Free old FIB nexthop group */ | |
578 | if (old->fib_ng.nexthop) { | |
579 | nexthops_free(old->fib_ng.nexthop); | |
580 | old->fib_ng.nexthop = NULL; | |
581 | } | |
582 | ||
583 | if (!RIB_SYSTEM_ROUTE(old)) { | |
584 | /* Clear old route's FIB flags */ | |
6b468511 | 585 | for (ALL_NEXTHOPS_PTR(old->ng, nexthop)) { |
efe6c026 MS |
586 | UNSET_FLAG(nexthop->flags, |
587 | NEXTHOP_FLAG_FIB); | |
588 | } | |
589 | } | |
590 | } | |
591 | ||
97f5b441 MS |
592 | if (zvrf) |
593 | zvrf->installs_queued++; | |
1e885672 | 594 | break; |
ea1c14f6 | 595 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
596 | { |
597 | char str[SRCDEST2STR_BUFFER]; | |
598 | ||
599 | srcdest_rnode2str(rn, str, sizeof(str)); | |
600 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
601 | "%u:%s: Failed to enqueue dataplane install", | |
602 | re->vrf_id, str); | |
1e885672 | 603 | break; |
97f5b441 | 604 | } |
ea1c14f6 | 605 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
97f5b441 MS |
606 | if (zvrf) |
607 | zvrf->installs++; | |
1e885672 DS |
608 | break; |
609 | } | |
d62a17ae | 610 | |
0c555cc6 | 611 | return; |
718e3744 | 612 | } |
613 | ||
614 | /* Uninstall the route from kernel. */ | |
0c555cc6 | 615 | void rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 616 | { |
d62a17ae | 617 | struct nexthop *nexthop; |
618 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 619 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
05737783 | 620 | |
d62a17ae | 621 | if (info->safi != SAFI_UNICAST) { |
677c1dd5 | 622 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
6b468511 | 623 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
d6792f9d | 624 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 625 | return; |
d62a17ae | 626 | } |
416ec78d | 627 | |
d62a17ae | 628 | /* |
629 | * Make sure we update the FPM any time we send new information to | |
97f5b441 | 630 | * the dataplane. |
d62a17ae | 631 | */ |
632 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
97f5b441 MS |
633 | |
634 | switch (dplane_route_delete(rn, re)) { | |
ea1c14f6 | 635 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
97f5b441 MS |
636 | if (zvrf) |
637 | zvrf->removals_queued++; | |
1e885672 | 638 | break; |
ea1c14f6 | 639 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
640 | { |
641 | char str[SRCDEST2STR_BUFFER]; | |
642 | ||
643 | srcdest_rnode2str(rn, str, sizeof(str)); | |
644 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
645 | "%u:%s: Failed to enqueue dataplane uninstall", | |
646 | re->vrf_id, str); | |
1e885672 | 647 | break; |
97f5b441 | 648 | } |
ea1c14f6 | 649 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
1e885672 DS |
650 | if (zvrf) |
651 | zvrf->removals++; | |
652 | break; | |
653 | } | |
718e3744 | 654 | |
0c555cc6 | 655 | return; |
718e3744 | 656 | } |
657 | ||
658 | /* Uninstall the route from kernel. */ | |
d62a17ae | 659 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 660 | { |
d62a17ae | 661 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
5f7a4718 | 662 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
97f5b441 | 663 | struct nexthop *nexthop; |
416ec78d | 664 | |
5f7a4718 | 665 | if (dest && dest->selected_fib == re) { |
d62a17ae | 666 | if (info->safi == SAFI_UNICAST) |
667 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 668 | |
d62a17ae | 669 | /* If labeled-unicast route, uninstall transit LSP. */ |
670 | if (zebra_rib_labeled_unicast(re)) | |
671 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
7cdb1a84 | 672 | |
3cdba47a | 673 | rib_uninstall_kernel(rn, re); |
97f5b441 MS |
674 | |
675 | dest->selected_fib = NULL; | |
676 | ||
efe6c026 MS |
677 | /* Free FIB nexthop group, if present */ |
678 | if (re->fib_ng.nexthop) { | |
679 | nexthops_free(re->fib_ng.nexthop); | |
680 | re->fib_ng.nexthop = NULL; | |
681 | } | |
682 | ||
6b468511 | 683 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
97f5b441 | 684 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
d62a17ae | 685 | } |
446bb95e | 686 | |
d62a17ae | 687 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
86391e56 MS |
688 | const struct prefix *p, *src_p; |
689 | ||
d62a17ae | 690 | srcdest_rnode_prefixes(rn, &p, &src_p); |
05737783 | 691 | |
40f321c0 | 692 | redistribute_delete(p, src_p, re, NULL); |
d62a17ae | 693 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); |
694 | } | |
718e3744 | 695 | } |
696 | ||
9fd92e3c AS |
697 | /* |
698 | * rib_can_delete_dest | |
699 | * | |
2951a7a4 | 700 | * Returns true if the given dest can be deleted from the table. |
9fd92e3c | 701 | */ |
d62a17ae | 702 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 703 | { |
045207e2 | 704 | if (re_list_first(&dest->routes)) { |
d62a17ae | 705 | return 0; |
706 | } | |
9fd92e3c | 707 | |
c86ba6c2 DS |
708 | /* |
709 | * Unresolved rnh's are stored on the default route's list | |
710 | * | |
711 | * dest->rnode can also be the source prefix node in an | |
712 | * ipv6 sourcedest table. Fortunately the prefix of a | |
713 | * source prefix node can never be the default prefix. | |
714 | */ | |
715 | if (is_default_prefix(&dest->rnode->p)) | |
716 | return 0; | |
717 | ||
d62a17ae | 718 | /* |
719 | * Don't delete the dest if we have to update the FPM about this | |
720 | * prefix. | |
721 | */ | |
722 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
723 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
724 | return 0; | |
5adc2528 | 725 | |
d62a17ae | 726 | return 1; |
9fd92e3c AS |
727 | } |
728 | ||
699dae23 DS |
729 | void zebra_rib_evaluate_rn_nexthops(struct route_node *rn, uint32_t seq) |
730 | { | |
731 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
699dae23 DS |
732 | struct rnh *rnh; |
733 | ||
734 | /* | |
50872b08 | 735 | * We are storing the rnh's associated withb |
699dae23 DS |
736 | * the tracked nexthop as a list of the rn's. |
737 | * Unresolved rnh's are placed at the top | |
738 | * of the tree list.( 0.0.0.0/0 for v4 and 0::0/0 for v6 ) | |
739 | * As such for each rn we need to walk up the tree | |
740 | * and see if any rnh's need to see if they | |
741 | * would match a more specific route | |
742 | */ | |
743 | while (rn) { | |
50872b08 DS |
744 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
745 | char buf[PREFIX_STRLEN]; | |
746 | ||
0a7be328 | 747 | zlog_debug("%s: %s Being examined for Nexthop Tracking Count: %zd", |
50872b08 | 748 | __PRETTY_FUNCTION__, |
0a7be328 DS |
749 | srcdest_rnode2str(rn, buf, sizeof(buf)), |
750 | dest ? rnh_list_count(&dest->nht) : 0); | |
50872b08 | 751 | } |
699dae23 DS |
752 | if (!dest) { |
753 | rn = rn->parent; | |
754 | if (rn) | |
755 | dest = rib_dest_from_rnode(rn); | |
756 | continue; | |
757 | } | |
758 | /* | |
759 | * If we have any rnh's stored in the nht list | |
760 | * then we know that this route node was used for | |
761 | * nht resolution and as such we need to call the | |
762 | * nexthop tracking evaluation code | |
763 | */ | |
739c9c90 | 764 | frr_each_safe(rnh_list, &dest->nht, rnh) { |
699dae23 DS |
765 | struct zebra_vrf *zvrf = |
766 | zebra_vrf_lookup_by_id(rnh->vrf_id); | |
767 | struct prefix *p = &rnh->node->p; | |
768 | ||
50872b08 | 769 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
699dae23 DS |
770 | char buf1[PREFIX_STRLEN]; |
771 | char buf2[PREFIX_STRLEN]; | |
772 | ||
0a7be328 | 773 | zlog_debug("%u:%s has Nexthop(%s) Type: %s depending on it, evaluating %u:%u", |
699dae23 | 774 | zvrf->vrf->vrf_id, |
50872b08 | 775 | srcdest_rnode2str(rn, buf1, |
699dae23 DS |
776 | sizeof(buf1)), |
777 | prefix2str(p, buf2, sizeof(buf2)), | |
0a7be328 | 778 | rnh_type2str(rnh->type), |
699dae23 DS |
779 | seq, rnh->seqno); |
780 | } | |
781 | ||
782 | /* | |
783 | * If we have evaluated this node on this pass | |
784 | * already, due to following the tree up | |
785 | * then we know that we can move onto the next | |
786 | * rnh to process. | |
787 | * | |
788 | * Additionally we call zebra_evaluate_rnh | |
789 | * when we gc the dest. In this case we know | |
790 | * that there must be no other re's where | |
791 | * we were originally as such we know that | |
792 | * that sequence number is ok to respect. | |
793 | */ | |
50872b08 DS |
794 | if (rnh->seqno == seq) { |
795 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) | |
796 | zlog_debug( | |
797 | "\tNode processed and moved already"); | |
699dae23 | 798 | continue; |
50872b08 | 799 | } |
699dae23 DS |
800 | |
801 | rnh->seqno = seq; | |
802 | zebra_evaluate_rnh(zvrf, family2afi(p->family), 0, | |
803 | rnh->type, p); | |
804 | } | |
805 | ||
806 | rn = rn->parent; | |
807 | if (rn) | |
808 | dest = rib_dest_from_rnode(rn); | |
809 | } | |
810 | } | |
811 | ||
9fd92e3c AS |
812 | /* |
813 | * rib_gc_dest | |
814 | * | |
815 | * Garbage collect the rib dest corresponding to the given route node | |
816 | * if appropriate. | |
817 | * | |
2951a7a4 | 818 | * Returns true if the dest was deleted, false otherwise. |
9fd92e3c | 819 | */ |
d62a17ae | 820 | int rib_gc_dest(struct route_node *rn) |
9fd92e3c | 821 | { |
d62a17ae | 822 | rib_dest_t *dest; |
9fd92e3c | 823 | |
d62a17ae | 824 | dest = rib_dest_from_rnode(rn); |
825 | if (!dest) | |
826 | return 0; | |
9fd92e3c | 827 | |
d62a17ae | 828 | if (!rib_can_delete_dest(dest)) |
829 | return 0; | |
9fd92e3c | 830 | |
c9abf558 DS |
831 | if (IS_ZEBRA_DEBUG_RIB) { |
832 | struct zebra_vrf *zvrf; | |
833 | ||
834 | zvrf = rib_dest_vrf(dest); | |
d62a17ae | 835 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); |
c9abf558 | 836 | } |
9fd92e3c | 837 | |
699dae23 DS |
838 | zebra_rib_evaluate_rn_nexthops(rn, zebra_router_get_next_sequence()); |
839 | ||
d62a17ae | 840 | dest->rnode = NULL; |
aa57abfb | 841 | rnh_list_fini(&dest->nht); |
d62a17ae | 842 | XFREE(MTYPE_RIB_DEST, dest); |
843 | rn->info = NULL; | |
9fd92e3c | 844 | |
d62a17ae | 845 | /* |
846 | * Release the one reference that we keep on the route node. | |
847 | */ | |
848 | route_unlock_node(rn); | |
849 | return 1; | |
9fd92e3c AS |
850 | } |
851 | ||
d62a17ae | 852 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
853 | struct route_entry *new) | |
3e5c6e00 | 854 | { |
d62a17ae | 855 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 856 | |
d62a17ae | 857 | /* Update real nexthop. This may actually determine if nexthop is active |
858 | * or not. */ | |
6b468511 | 859 | if (!nexthop_group_active_nexthop_num(new->ng)) { |
d62a17ae | 860 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
861 | return; | |
862 | } | |
3e5c6e00 | 863 | |
d62a17ae | 864 | if (IS_ZEBRA_DEBUG_RIB) { |
865 | char buf[SRCDEST2STR_BUFFER]; | |
866 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
867 | zlog_debug("%u:%s: Adding route rn %p, re %p (%s)", |
868 | zvrf_id(zvrf), buf, rn, new, | |
869 | zebra_route_string(new->type)); | |
d62a17ae | 870 | } |
3e5c6e00 | 871 | |
d62a17ae | 872 | /* If labeled-unicast route, install transit LSP. */ |
873 | if (zebra_rib_labeled_unicast(new)) | |
874 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 875 | |
3cdba47a | 876 | rib_install_kernel(rn, new, NULL); |
3e5c6e00 | 877 | |
d62a17ae | 878 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 879 | } |
880 | ||
d62a17ae | 881 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
882 | struct route_entry *old) | |
3e5c6e00 | 883 | { |
d62a17ae | 884 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 885 | |
d62a17ae | 886 | /* Uninstall from kernel. */ |
887 | if (IS_ZEBRA_DEBUG_RIB) { | |
888 | char buf[SRCDEST2STR_BUFFER]; | |
889 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
890 | zlog_debug("%u:%s: Deleting route rn %p, re %p (%s)", |
891 | zvrf_id(zvrf), buf, rn, old, | |
892 | zebra_route_string(old->type)); | |
d62a17ae | 893 | } |
3e5c6e00 | 894 | |
d62a17ae | 895 | /* If labeled-unicast route, uninstall transit LSP. */ |
896 | if (zebra_rib_labeled_unicast(old)) | |
897 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
898 | ||
3cdba47a | 899 | rib_uninstall_kernel(rn, old); |
d62a17ae | 900 | |
901 | /* Update nexthop for route, reset changed flag. */ | |
ef57f35f DL |
902 | /* Note: this code also handles the Linux case when an interface goes |
903 | * down, causing the kernel to delete routes without sending DELROUTE | |
904 | * notifications | |
905 | */ | |
6883bf8d | 906 | if (RIB_KERNEL_ROUTE(old)) |
212df1de PG |
907 | SET_FLAG(old->status, ROUTE_ENTRY_REMOVED); |
908 | else | |
909 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 910 | } |
911 | ||
912 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
913 | struct route_node *rn, | |
914 | struct route_entry *old, | |
915 | struct route_entry *new) | |
916 | { | |
d62a17ae | 917 | int nh_active = 0; |
d62a17ae | 918 | |
919 | /* | |
920 | * We have to install or update if a new route has been selected or | |
921 | * something has changed. | |
922 | */ | |
923 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
924 | hook_call(rib_update, rn, "updating existing route"); | |
925 | ||
926 | /* Update the nexthop; we could determine here that nexthop is | |
927 | * inactive. */ | |
6b468511 | 928 | if (nexthop_group_active_nexthop_num(new->ng)) |
d62a17ae | 929 | nh_active = 1; |
930 | ||
931 | /* If nexthop is active, install the selected route, if | |
932 | * appropriate. If | |
933 | * the install succeeds, cleanup flags for prior route, if | |
934 | * different from | |
935 | * newly selected. | |
936 | */ | |
937 | if (nh_active) { | |
938 | if (IS_ZEBRA_DEBUG_RIB) { | |
939 | char buf[SRCDEST2STR_BUFFER]; | |
940 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
941 | if (new != old) | |
942 | zlog_debug( | |
2da33d6b | 943 | "%u:%s: Updating route rn %p, re %p (%s) old %p (%s)", |
d62a17ae | 944 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
945 | zebra_route_string(new->type), |
946 | old, | |
947 | zebra_route_string(old->type)); | |
d62a17ae | 948 | else |
949 | zlog_debug( | |
2da33d6b | 950 | "%u:%s: Updating route rn %p, re %p (%s)", |
d62a17ae | 951 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 952 | zebra_route_string(new->type)); |
d62a17ae | 953 | } |
954 | ||
955 | /* If labeled-unicast route, uninstall transit LSP. */ | |
956 | if (zebra_rib_labeled_unicast(old)) | |
957 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
958 | ||
3cdba47a DS |
959 | /* |
960 | * Non-system route should be installed. | |
961 | * If labeled-unicast route, install transit | |
962 | * LSP. | |
963 | */ | |
964 | if (zebra_rib_labeled_unicast(new)) | |
965 | zebra_mpls_lsp_install(zvrf, rn, new); | |
d62a17ae | 966 | |
3cdba47a | 967 | rib_install_kernel(rn, new, old); |
d62a17ae | 968 | } |
a64448ba | 969 | |
d62a17ae | 970 | /* |
971 | * If nexthop for selected route is not active or install | |
972 | * failed, we | |
973 | * may need to uninstall and delete for redistribution. | |
974 | */ | |
ed216282 | 975 | if (!nh_active) { |
d62a17ae | 976 | if (IS_ZEBRA_DEBUG_RIB) { |
977 | char buf[SRCDEST2STR_BUFFER]; | |
978 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
979 | if (new != old) | |
980 | zlog_debug( | |
2da33d6b | 981 | "%u:%s: Deleting route rn %p, re %p (%s) old %p (%s) - nexthop inactive", |
d62a17ae | 982 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
983 | zebra_route_string(new->type), |
984 | old, | |
985 | zebra_route_string(old->type)); | |
d62a17ae | 986 | else |
987 | zlog_debug( | |
2da33d6b | 988 | "%u:%s: Deleting route rn %p, re %p (%s) - nexthop inactive", |
d62a17ae | 989 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 990 | zebra_route_string(new->type)); |
d62a17ae | 991 | } |
992 | ||
993 | /* If labeled-unicast route, uninstall transit LSP. */ | |
994 | if (zebra_rib_labeled_unicast(old)) | |
995 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
996 | ||
3cdba47a | 997 | rib_uninstall_kernel(rn, old); |
d62a17ae | 998 | } |
999 | } else { | |
1000 | /* | |
1001 | * Same route selected; check if in the FIB and if not, | |
3cdba47a DS |
1002 | * re-install. This is housekeeping code to deal with |
1003 | * race conditions in kernel with linux netlink reporting | |
1004 | * interface up before IPv4 or IPv6 protocol is ready | |
d62a17ae | 1005 | * to add routes. |
1006 | */ | |
3cdba47a DS |
1007 | if (!CHECK_FLAG(new->status, ROUTE_ENTRY_INSTALLED) || |
1008 | RIB_SYSTEM_ROUTE(new)) | |
677c1dd5 | 1009 | rib_install_kernel(rn, new, NULL); |
d62a17ae | 1010 | } |
1011 | ||
1012 | /* Update prior route. */ | |
6883bf8d | 1013 | if (new != old) |
d62a17ae | 1014 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1015 | |
d62a17ae | 1016 | /* Clear changed flag. */ |
1017 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 1018 | } |
1019 | ||
d62a17ae | 1020 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
1021 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1022 | struct route_entry *alternate) | |
1023 | { | |
1024 | if (current == NULL) | |
1025 | return alternate; | |
1026 | ||
1027 | /* filter route selection in following order: | |
1028 | * - connected beats other types | |
fec4ca19 | 1029 | * - if both connected, loopback or vrf wins |
d62a17ae | 1030 | * - lower distance beats higher |
1031 | * - lower metric beats higher for equal distance | |
1032 | * - last, hence oldest, route wins tie break. | |
1033 | */ | |
1034 | ||
fec4ca19 DS |
1035 | /* Connected routes. Check to see if either are a vrf |
1036 | * or loopback interface. If not, pick the last connected | |
d62a17ae | 1037 | * route of the set of lowest metric connected routes. |
1038 | */ | |
1039 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
fec4ca19 DS |
1040 | if (current->type != ZEBRA_ROUTE_CONNECT) |
1041 | return alternate; | |
1042 | ||
1043 | /* both are connected. are either loop or vrf? */ | |
1044 | struct nexthop *nexthop = NULL; | |
1045 | ||
6b468511 | 1046 | for (ALL_NEXTHOPS_PTR(alternate->ng, nexthop)) { |
873fde8c | 1047 | struct interface *ifp = if_lookup_by_index( |
1048 | nexthop->ifindex, alternate->vrf_id); | |
1049 | ||
1050 | if (ifp && if_is_loopback_or_vrf(ifp)) | |
fec4ca19 DS |
1051 | return alternate; |
1052 | } | |
1053 | ||
6b468511 | 1054 | for (ALL_NEXTHOPS_PTR(current->ng, nexthop)) { |
873fde8c | 1055 | struct interface *ifp = if_lookup_by_index( |
1056 | nexthop->ifindex, current->vrf_id); | |
1057 | ||
1058 | if (ifp && if_is_loopback_or_vrf(ifp)) | |
fec4ca19 DS |
1059 | return current; |
1060 | } | |
1061 | ||
1062 | /* Neither are loop or vrf so pick best metric */ | |
1063 | if (alternate->metric <= current->metric) | |
d62a17ae | 1064 | return alternate; |
1065 | ||
1066 | return current; | |
1067 | } | |
3e5c6e00 | 1068 | |
d62a17ae | 1069 | if (current->type == ZEBRA_ROUTE_CONNECT) |
1070 | return current; | |
3e5c6e00 | 1071 | |
d62a17ae | 1072 | /* higher distance loses */ |
1073 | if (alternate->distance < current->distance) | |
1074 | return alternate; | |
1075 | if (current->distance < alternate->distance) | |
1076 | return current; | |
3e5c6e00 | 1077 | |
d62a17ae | 1078 | /* metric tie-breaks equal distance */ |
1079 | if (alternate->metric <= current->metric) | |
1080 | return alternate; | |
3e5c6e00 | 1081 | |
d62a17ae | 1082 | return current; |
3e5c6e00 | 1083 | } |
1084 | ||
d62a17ae | 1085 | /* Core function for processing routing information base. */ |
1086 | static void rib_process(struct route_node *rn) | |
1087 | { | |
1088 | struct route_entry *re; | |
1089 | struct route_entry *next; | |
1090 | struct route_entry *old_selected = NULL; | |
1091 | struct route_entry *new_selected = NULL; | |
1092 | struct route_entry *old_fib = NULL; | |
1093 | struct route_entry *new_fib = NULL; | |
1094 | struct route_entry *best = NULL; | |
1095 | char buf[SRCDEST2STR_BUFFER]; | |
1096 | rib_dest_t *dest; | |
1097 | struct zebra_vrf *zvrf = NULL; | |
86391e56 MS |
1098 | const struct prefix *p, *src_p; |
1099 | ||
d62a17ae | 1100 | srcdest_rnode_prefixes(rn, &p, &src_p); |
1101 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1102 | ||
1103 | assert(rn); | |
1104 | ||
1105 | dest = rib_dest_from_rnode(rn); | |
1106 | if (dest) { | |
1107 | zvrf = rib_dest_vrf(dest); | |
1108 | vrf_id = zvrf_id(zvrf); | |
1109 | } | |
bab85d4f | 1110 | |
d62a17ae | 1111 | if (IS_ZEBRA_DEBUG_RIB) |
1112 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1113 | |
d62a17ae | 1114 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1115 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1116 | |
607425e5 DS |
1117 | /* |
1118 | * we can have rn's that have a NULL info pointer | |
1119 | * (dest). As such let's not let the deref happen | |
1120 | * additionally we know RNODE_FOREACH_RE_SAFE | |
1121 | * will not iterate so we are ok. | |
1122 | */ | |
1123 | if (dest) | |
1124 | old_fib = dest->selected_fib; | |
5f7a4718 | 1125 | |
a2addae8 | 1126 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1127 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1128 | zlog_debug( | |
2da33d6b DS |
1129 | "%u:%s: Examine re %p (%s) status %x flags %x dist %d metric %d", |
1130 | vrf_id, buf, re, zebra_route_string(re->type), | |
1131 | re->status, re->flags, re->distance, | |
1132 | re->metric); | |
d62a17ae | 1133 | |
d62a17ae | 1134 | /* Currently selected re. */ |
1135 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1136 | assert(old_selected == NULL); | |
1137 | old_selected = re; | |
1138 | } | |
bab85d4f | 1139 | |
d62a17ae | 1140 | /* Skip deleted entries from selection */ |
1141 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1142 | continue; | |
1143 | ||
1144 | /* Skip unreachable nexthop. */ | |
1145 | /* This first call to nexthop_active_update is merely to | |
6883bf8d DS |
1146 | * determine if there's any change to nexthops associated |
1147 | * with this RIB entry. Now, rib_process() can be invoked due | |
1148 | * to an external event such as link down or due to | |
1149 | * next-hop-tracking evaluation. In the latter case, | |
d62a17ae | 1150 | * a decision has already been made that the NHs have changed. |
6883bf8d DS |
1151 | * So, no need to invoke a potentially expensive call again. |
1152 | * Further, since the change might be in a recursive NH which | |
1153 | * is not caught in the nexthop_active_update() code. Thus, we | |
1154 | * might miss changes to recursive NHs. | |
d62a17ae | 1155 | */ |
6883bf8d | 1156 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) |
99eabcec | 1157 | && !nexthop_active_update(rn, re)) { |
d62a17ae | 1158 | if (re->type == ZEBRA_ROUTE_TABLE) { |
1159 | /* XXX: HERE BE DRAGONS!!!!! | |
1160 | * In all honesty, I have not yet figured out | |
6883bf8d DS |
1161 | * what this part does or why the |
1162 | * ROUTE_ENTRY_CHANGED test above is correct | |
d62a17ae | 1163 | * or why we need to delete a route here, and |
6883bf8d DS |
1164 | * also not whether this concerns both selected |
1165 | * and fib route, or only selected | |
1166 | * or only fib | |
1167 | * | |
1168 | * This entry was denied by the 'ip protocol | |
1169 | * table' route-map, we need to delete it */ | |
d62a17ae | 1170 | if (re != old_selected) { |
1171 | if (IS_ZEBRA_DEBUG_RIB) | |
1172 | zlog_debug( | |
32391aff | 1173 | "%s: %u:%s: imported via import-table but denied " |
d62a17ae | 1174 | "by the ip protocol table route-map", |
32391aff | 1175 | __func__, vrf_id, buf); |
d62a17ae | 1176 | rib_unlink(rn, re); |
1177 | } else | |
1178 | SET_FLAG(re->status, | |
1179 | ROUTE_ENTRY_REMOVED); | |
1180 | } | |
1181 | ||
1182 | continue; | |
1183 | } | |
bab85d4f | 1184 | |
d62a17ae | 1185 | /* Infinite distance. */ |
1186 | if (re->distance == DISTANCE_INFINITY) { | |
1187 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1188 | continue; | |
1189 | } | |
bab85d4f | 1190 | |
d62a17ae | 1191 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1192 | best = rib_choose_best(new_fib, re); | |
1193 | if (new_fib && best != new_fib) | |
1194 | UNSET_FLAG(new_fib->status, | |
1195 | ROUTE_ENTRY_CHANGED); | |
1196 | new_fib = best; | |
1197 | } else { | |
1198 | best = rib_choose_best(new_selected, re); | |
1199 | if (new_selected && best != new_selected) | |
1200 | UNSET_FLAG(new_selected->status, | |
1201 | ROUTE_ENTRY_CHANGED); | |
1202 | new_selected = best; | |
1203 | } | |
1204 | if (best != re) | |
1205 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1206 | } /* RNODE_FOREACH_RE */ | |
1207 | ||
1208 | /* If no FIB override route, use the selected route also for FIB */ | |
1209 | if (new_fib == NULL) | |
1210 | new_fib = new_selected; | |
1211 | ||
1212 | /* After the cycle is finished, the following pointers will be set: | |
1213 | * old_selected --- RE entry currently having SELECTED | |
1214 | * new_selected --- RE entry that is newly SELECTED | |
1215 | * old_fib --- RE entry currently in kernel FIB | |
1216 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1217 | * | |
1218 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1219 | * the zclients. new_fib (which can be new_selected) will be installed | |
1220 | * in kernel. | |
1221 | */ | |
1222 | ||
1223 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1224 | zlog_debug( | |
1225 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1226 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1227 | (void *)old_fib, (void *)new_fib); | |
1228 | } | |
446bb95e | 1229 | |
d62a17ae | 1230 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1231 | * fib == selected */ | |
9d303b37 DL |
1232 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, |
1233 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1234 | |
1235 | /* Update fib according to selection results */ | |
1236 | if (new_fib && old_fib) | |
1237 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1238 | else if (new_fib) | |
1239 | rib_process_add_fib(zvrf, rn, new_fib); | |
1240 | else if (old_fib) | |
1241 | rib_process_del_fib(zvrf, rn, old_fib); | |
1242 | ||
8cb41cd6 | 1243 | /* Update SELECTED entry */ |
d62a17ae | 1244 | if (old_selected != new_selected || selected_changed) { |
93bdadae | 1245 | |
6883bf8d | 1246 | if (new_selected && new_selected != new_fib) |
d62a17ae | 1247 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); |
41ec9222 | 1248 | |
3cdba47a | 1249 | if (new_selected) |
5af4b346 MS |
1250 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); |
1251 | ||
d62a17ae | 1252 | if (old_selected) { |
40f321c0 MS |
1253 | /* |
1254 | * If we're removing the old entry, we should tell | |
1255 | * redist subscribers about that *if* they aren't | |
1256 | * going to see a redist for the new entry. | |
1257 | */ | |
1258 | if (!new_selected || CHECK_FLAG(old_selected->status, | |
1259 | ROUTE_ENTRY_REMOVED)) | |
1260 | redistribute_delete(p, src_p, | |
1261 | old_selected, | |
1262 | new_selected); | |
1263 | ||
d62a17ae | 1264 | if (old_selected != new_selected) |
1265 | UNSET_FLAG(old_selected->flags, | |
1266 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1267 | } |
d62a17ae | 1268 | } |
3e5c6e00 | 1269 | |
d62a17ae | 1270 | /* Remove all RE entries queued for removal */ |
a2addae8 | 1271 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1272 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { |
1273 | if (IS_ZEBRA_DEBUG_RIB) { | |
1274 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1275 | (void *)rn, (void *)re); | |
1276 | } | |
1277 | rib_unlink(rn, re); | |
1278 | } | |
1279 | } | |
4d38fdb4 | 1280 | |
d62a17ae | 1281 | /* |
1282 | * Check if the dest can be deleted now. | |
1283 | */ | |
1284 | rib_gc_dest(rn); | |
e96f9203 DO |
1285 | } |
1286 | ||
a1494c25 DS |
1287 | static void zebra_rib_evaluate_mpls(struct route_node *rn) |
1288 | { | |
1289 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
1290 | struct zebra_vrf *zvrf = vrf_info_lookup(VRF_DEFAULT); | |
1291 | ||
1292 | if (!dest) | |
1293 | return; | |
1294 | ||
1295 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_LSPS)) { | |
1296 | if (IS_ZEBRA_DEBUG_MPLS) | |
1297 | zlog_debug( | |
1298 | "%u: Scheduling all LSPs upon RIB completion", | |
1299 | zvrf_id(zvrf)); | |
1300 | zebra_mpls_lsp_schedule(zvrf); | |
1301 | mpls_unmark_lsps_for_processing(rn); | |
1302 | } | |
1303 | } | |
1304 | ||
e5ac2adf MS |
1305 | /* |
1306 | * Utility to match route with dplane context data | |
1307 | */ | |
1308 | static bool rib_route_match_ctx(const struct route_entry *re, | |
25779064 MS |
1309 | const struct zebra_dplane_ctx *ctx, |
1310 | bool is_update) | |
e5ac2adf MS |
1311 | { |
1312 | bool result = false; | |
1313 | ||
1314 | if (is_update) { | |
1315 | /* | |
1316 | * In 'update' case, we test info about the 'previous' or | |
1317 | * 'old' route | |
1318 | */ | |
1319 | if ((re->type == dplane_ctx_get_old_type(ctx)) && | |
1320 | (re->instance == dplane_ctx_get_old_instance(ctx))) { | |
1321 | result = true; | |
1322 | ||
1323 | /* TODO -- we're using this extra test, but it's not | |
1324 | * exactly clear why. | |
1325 | */ | |
1326 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1327 | (re->distance != dplane_ctx_get_old_distance(ctx) || | |
1328 | re->tag != dplane_ctx_get_old_tag(ctx))) { | |
1329 | result = false; | |
1330 | } | |
1331 | } | |
1332 | ||
1333 | } else { | |
1334 | /* | |
1335 | * Ordinary, single-route case using primary context info | |
1336 | */ | |
1337 | if ((dplane_ctx_get_op(ctx) != DPLANE_OP_ROUTE_DELETE) && | |
1338 | CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1339 | /* Skip route that's been deleted */ | |
1340 | goto done; | |
1341 | } | |
1342 | ||
1343 | if ((re->type == dplane_ctx_get_type(ctx)) && | |
1344 | (re->instance == dplane_ctx_get_instance(ctx))) { | |
1345 | result = true; | |
1346 | ||
1347 | /* TODO -- we're using this extra test, but it's not | |
1348 | * exactly clear why. | |
1349 | */ | |
1350 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1351 | (re->distance != dplane_ctx_get_distance(ctx) || | |
1352 | re->tag != dplane_ctx_get_tag(ctx))) { | |
1353 | result = false; | |
1354 | } | |
1355 | } | |
1356 | } | |
1357 | ||
1358 | done: | |
1359 | ||
1360 | return (result); | |
1361 | } | |
1362 | ||
3cdba47a DS |
1363 | static void zebra_rib_fixup_system(struct route_node *rn) |
1364 | { | |
1365 | struct route_entry *re; | |
1366 | ||
1367 | RNODE_FOREACH_RE(rn, re) { | |
1368 | struct nexthop *nhop; | |
1369 | ||
1370 | if (!RIB_SYSTEM_ROUTE(re)) | |
1371 | continue; | |
1372 | ||
1373 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1374 | continue; | |
1375 | ||
1376 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
6d0ee6a0 | 1377 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); |
3cdba47a | 1378 | |
6b468511 | 1379 | for (ALL_NEXTHOPS_PTR(re->ng, nhop)) { |
3cdba47a DS |
1380 | if (CHECK_FLAG(nhop->flags, NEXTHOP_FLAG_RECURSIVE)) |
1381 | continue; | |
1382 | ||
1383 | SET_FLAG(nhop->flags, NEXTHOP_FLAG_FIB); | |
1384 | } | |
1385 | } | |
1386 | } | |
1387 | ||
e5ac2adf | 1388 | /* |
78bf56b0 MS |
1389 | * Update a route from a dplane context. This consolidates common code |
1390 | * that can be used in processing of results from FIB updates, and in | |
1391 | * async notification processing. | |
efe6c026 | 1392 | * The return is 'true' if the installed nexthops changed; 'false' otherwise. |
e5ac2adf | 1393 | */ |
efe6c026 MS |
1394 | static bool rib_update_re_from_ctx(struct route_entry *re, |
1395 | struct route_node *rn, | |
1396 | struct zebra_dplane_ctx *ctx) | |
78bf56b0 | 1397 | { |
941e261c | 1398 | char dest_str[PREFIX_STRLEN] = ""; |
efe6c026 | 1399 | char nh_str[NEXTHOP_STRLEN]; |
78bf56b0 | 1400 | struct nexthop *nexthop, *ctx_nexthop; |
941e261c MS |
1401 | bool matched; |
1402 | const struct nexthop_group *ctxnhg; | |
efe6c026 MS |
1403 | bool is_selected = false; /* Is 're' currently the selected re? */ |
1404 | bool changed_p = false; /* Change to nexthops? */ | |
1405 | rib_dest_t *dest; | |
78bf56b0 | 1406 | |
941e261c MS |
1407 | /* Note well: only capturing the prefix string if debug is enabled here; |
1408 | * unconditional log messages will have to generate the string. | |
1409 | */ | |
1410 | if (IS_ZEBRA_DEBUG_RIB) | |
1411 | prefix2str(&(rn->p), dest_str, sizeof(dest_str)); | |
78bf56b0 | 1412 | |
efe6c026 MS |
1413 | dest = rib_dest_from_rnode(rn); |
1414 | if (dest) | |
1415 | is_selected = (re == dest->selected_fib); | |
1416 | ||
1417 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1418 | zlog_debug("update_from_ctx: %u:%s: %sSELECTED", | |
1419 | re->vrf_id, dest_str, (is_selected ? "" : "NOT ")); | |
1420 | ||
1421 | /* Update zebra's nexthop FIB flag for each nexthop that was installed. | |
941e261c | 1422 | * If the installed set differs from the set requested by the rib/owner, |
efe6c026 MS |
1423 | * we use the fib-specific nexthop-group to record the actual FIB |
1424 | * status. | |
78bf56b0 | 1425 | */ |
941e261c MS |
1426 | |
1427 | /* | |
1428 | * First check the fib nexthop-group, if it's present. The comparison | |
efe6c026 | 1429 | * here is quite strict: we require that the fib sets match exactly. |
941e261c MS |
1430 | */ |
1431 | matched = false; | |
1432 | do { | |
1433 | if (re->fib_ng.nexthop == NULL) | |
1434 | break; | |
1435 | ||
1436 | matched = true; | |
1437 | ||
1438 | /* First check the route's fib nexthops */ | |
1439 | for (ALL_NEXTHOPS(re->fib_ng, nexthop)) { | |
1440 | ||
1441 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1442 | continue; | |
1443 | ||
efe6c026 | 1444 | ctx_nexthop = NULL; |
941e261c MS |
1445 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), |
1446 | ctx_nexthop)) { | |
1447 | if (nexthop_same(ctx_nexthop, nexthop)) | |
1448 | break; | |
1449 | } | |
1450 | ||
1451 | if (ctx_nexthop == NULL) { | |
1452 | /* Nexthop not in the new installed set */ | |
efe6c026 MS |
1453 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1454 | nexthop2str(nexthop, nh_str, | |
1455 | sizeof(nh_str)); | |
1456 | zlog_debug("update_from_ctx: no match for fib nh %s", | |
1457 | nh_str); | |
1458 | } | |
1459 | ||
941e261c MS |
1460 | matched = false; |
1461 | break; | |
1462 | } | |
1463 | } | |
1464 | ||
1465 | if (!matched) | |
1466 | break; | |
1467 | ||
1468 | /* Check the new installed set */ | |
efe6c026 | 1469 | ctx_nexthop = NULL; |
941e261c MS |
1470 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), ctx_nexthop)) { |
1471 | ||
1472 | if (CHECK_FLAG(ctx_nexthop->flags, | |
1473 | NEXTHOP_FLAG_RECURSIVE)) | |
1474 | continue; | |
1475 | ||
1476 | /* Compare with the current group's nexthops */ | |
efe6c026 MS |
1477 | nexthop = NULL; |
1478 | for (ALL_NEXTHOPS(re->fib_ng, nexthop)) { | |
941e261c MS |
1479 | if (nexthop_same(nexthop, ctx_nexthop)) |
1480 | break; | |
1481 | } | |
1482 | ||
1483 | if (nexthop == NULL) { | |
1484 | /* Nexthop not in the old installed set */ | |
efe6c026 MS |
1485 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1486 | nexthop2str(ctx_nexthop, nh_str, | |
1487 | sizeof(nh_str)); | |
1488 | zlog_debug("update_from_ctx: no fib match for notif nh %s", | |
1489 | nh_str); | |
1490 | } | |
941e261c MS |
1491 | matched = false; |
1492 | break; | |
1493 | } | |
1494 | } | |
1495 | ||
1496 | } while (0); | |
1497 | ||
1498 | /* If the new FIB set matches the existing FIB set, we're done. */ | |
1499 | if (matched) { | |
1500 | if (IS_ZEBRA_DEBUG_RIB) | |
1501 | zlog_debug("%u:%s update_from_ctx(): existing fib nhg, no change", | |
1502 | re->vrf_id, dest_str); | |
1503 | goto done; | |
1504 | ||
1505 | } else if (re->fib_ng.nexthop) { | |
1506 | /* | |
1507 | * Free stale fib list and move on to check the rib nhg. | |
1508 | */ | |
1509 | if (IS_ZEBRA_DEBUG_RIB) | |
1510 | zlog_debug("%u:%s update_from_ctx(): replacing fib nhg", | |
1511 | re->vrf_id, dest_str); | |
1512 | nexthops_free(re->fib_ng.nexthop); | |
1513 | re->fib_ng.nexthop = NULL; | |
efe6c026 MS |
1514 | |
1515 | /* Note that the installed nexthops have changed */ | |
1516 | changed_p = true; | |
941e261c MS |
1517 | } else { |
1518 | if (IS_ZEBRA_DEBUG_RIB) | |
1519 | zlog_debug("%u:%s update_from_ctx(): no fib nhg", | |
1520 | re->vrf_id, dest_str); | |
1521 | } | |
1522 | ||
1523 | /* | |
1524 | * Compare with the rib nexthop group. The comparison here is different: | |
1525 | * the RIB group may be a superset of the list installed in the FIB. We | |
1526 | * walk the RIB group, looking for the 'installable' candidate | |
1527 | * nexthops, and then check those against the set | |
1528 | * that is actually installed. | |
1529 | */ | |
1530 | matched = true; | |
6b468511 | 1531 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) { |
78bf56b0 | 1532 | |
5695d9ac | 1533 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) |
78bf56b0 MS |
1534 | continue; |
1535 | ||
941e261c MS |
1536 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) |
1537 | continue; | |
5695d9ac | 1538 | |
941e261c | 1539 | /* Check for a FIB nexthop corresponding to the RIB nexthop */ |
efe6c026 | 1540 | ctx_nexthop = NULL; |
941e261c | 1541 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), ctx_nexthop)) { |
78bf56b0 MS |
1542 | if (nexthop_same(ctx_nexthop, nexthop)) |
1543 | break; | |
1544 | } | |
1545 | ||
5695d9ac MS |
1546 | /* If the FIB doesn't know about the nexthop, |
1547 | * it's not installed | |
1548 | */ | |
1549 | if (ctx_nexthop == NULL) { | |
efe6c026 MS |
1550 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1551 | nexthop2str(nexthop, nh_str, sizeof(nh_str)); | |
1552 | zlog_debug("update_from_ctx: no notif match for rib nh %s", | |
1553 | nh_str); | |
1554 | } | |
941e261c | 1555 | matched = false; |
efe6c026 MS |
1556 | |
1557 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1558 | changed_p = true; | |
1559 | ||
1560 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
11260e70 MS |
1561 | |
1562 | /* Keep checking nexthops */ | |
1563 | continue; | |
5695d9ac | 1564 | } |
78bf56b0 | 1565 | |
efe6c026 MS |
1566 | if (CHECK_FLAG(ctx_nexthop->flags, NEXTHOP_FLAG_FIB)) { |
1567 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1568 | changed_p = true; | |
1569 | ||
5695d9ac | 1570 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
efe6c026 MS |
1571 | } else { |
1572 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1573 | changed_p = true; | |
1574 | ||
5695d9ac | 1575 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
efe6c026 | 1576 | } |
78bf56b0 MS |
1577 | } |
1578 | ||
941e261c MS |
1579 | /* If all nexthops were processed, we're done */ |
1580 | if (matched) { | |
1581 | if (IS_ZEBRA_DEBUG_RIB) | |
efe6c026 MS |
1582 | zlog_debug("%u:%s update_from_ctx(): rib nhg matched, changed '%s'", |
1583 | re->vrf_id, dest_str, | |
1584 | (changed_p ? "true" : "false")); | |
941e261c MS |
1585 | goto done; |
1586 | } | |
78bf56b0 | 1587 | |
941e261c MS |
1588 | /* FIB nexthop set differs from the RIB set: |
1589 | * create a fib-specific nexthop-group | |
1590 | */ | |
1591 | if (IS_ZEBRA_DEBUG_RIB) | |
efe6c026 MS |
1592 | zlog_debug("%u:%s update_from_ctx(): changed %s, adding new fib nhg", |
1593 | re->vrf_id, dest_str, | |
1594 | (changed_p ? "true" : "false")); | |
941e261c MS |
1595 | |
1596 | ctxnhg = dplane_ctx_get_ng(ctx); | |
188a00e0 MS |
1597 | |
1598 | if (ctxnhg->nexthop) | |
1599 | copy_nexthops(&(re->fib_ng.nexthop), ctxnhg->nexthop, NULL); | |
1600 | else { | |
1601 | /* Bit of a special case when the fib has _no_ installed | |
1602 | * nexthops. | |
1603 | */ | |
1604 | nexthop = nexthop_new(); | |
1605 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
50d89650 | 1606 | _nexthop_add(&(re->fib_ng.nexthop), nexthop); |
188a00e0 | 1607 | } |
941e261c MS |
1608 | |
1609 | done: | |
efe6c026 | 1610 | return changed_p; |
78bf56b0 MS |
1611 | } |
1612 | ||
e5ac2adf | 1613 | /* |
fad4d69c MS |
1614 | * Helper to locate a zebra route-node from a dplane context. This is used |
1615 | * when processing dplane results, e.g. Note well: the route-node is returned | |
1616 | * with a ref held - route_unlock_node() must be called eventually. | |
e5ac2adf | 1617 | */ |
fad4d69c MS |
1618 | static struct route_node * |
1619 | rib_find_rn_from_ctx(const struct zebra_dplane_ctx *ctx) | |
e5ac2adf MS |
1620 | { |
1621 | struct route_table *table = NULL; | |
1622 | struct route_node *rn = NULL; | |
e5ac2adf MS |
1623 | const struct prefix *dest_pfx, *src_pfx; |
1624 | ||
1625 | /* Locate rn and re(s) from ctx */ | |
1626 | ||
1627 | table = zebra_vrf_table_with_table_id(dplane_ctx_get_afi(ctx), | |
1628 | dplane_ctx_get_safi(ctx), | |
1629 | dplane_ctx_get_vrf(ctx), | |
1630 | dplane_ctx_get_table(ctx)); | |
1631 | if (table == NULL) { | |
1632 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fad4d69c | 1633 | zlog_debug("Failed to find route for ctx: no table for afi %d, safi %d, vrf %u", |
e5ac2adf MS |
1634 | dplane_ctx_get_afi(ctx), |
1635 | dplane_ctx_get_safi(ctx), | |
1636 | dplane_ctx_get_vrf(ctx)); | |
1637 | } | |
1638 | goto done; | |
1639 | } | |
1640 | ||
fad4d69c MS |
1641 | dest_pfx = dplane_ctx_get_dest(ctx); |
1642 | src_pfx = dplane_ctx_get_src(ctx); | |
1643 | ||
1644 | rn = srcdest_rnode_get(table, dest_pfx, | |
1645 | src_pfx ? (struct prefix_ipv6 *)src_pfx : NULL); | |
1646 | ||
1647 | done: | |
1648 | return rn; | |
1649 | } | |
1650 | ||
1651 | ||
97f5b441 | 1652 | |
fad4d69c MS |
1653 | /* |
1654 | * Route-update results processing after async dataplane update. | |
1655 | */ | |
1656 | static void rib_process_result(struct zebra_dplane_ctx *ctx) | |
1657 | { | |
1658 | struct zebra_vrf *zvrf = NULL; | |
1659 | struct route_node *rn = NULL; | |
1660 | struct route_entry *re = NULL, *old_re = NULL, *rib; | |
1661 | bool is_update = false; | |
1662 | char dest_str[PREFIX_STRLEN] = ""; | |
1663 | enum dplane_op_e op; | |
1664 | enum zebra_dplane_result status; | |
1665 | const struct prefix *dest_pfx, *src_pfx; | |
1666 | uint32_t seq; | |
efe6c026 | 1667 | bool fib_changed = false; |
97f5b441 | 1668 | |
fad4d69c | 1669 | zvrf = vrf_info_lookup(dplane_ctx_get_vrf(ctx)); |
e5ac2adf MS |
1670 | dest_pfx = dplane_ctx_get_dest(ctx); |
1671 | ||
1672 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1673 | * unconditional log messages will have to generate the string. | |
1674 | */ | |
5709131c | 1675 | if (IS_ZEBRA_DEBUG_DPLANE) |
e5ac2adf | 1676 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
e5ac2adf | 1677 | |
fad4d69c MS |
1678 | /* Locate rn and re(s) from ctx */ |
1679 | rn = rib_find_rn_from_ctx(ctx); | |
e5ac2adf MS |
1680 | if (rn == NULL) { |
1681 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fe2c53d4 | 1682 | zlog_debug("Failed to process dplane results: no route for %u:%s", |
e5ac2adf MS |
1683 | dplane_ctx_get_vrf(ctx), dest_str); |
1684 | } | |
1685 | goto done; | |
1686 | } | |
1687 | ||
1688 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); | |
1689 | ||
1690 | op = dplane_ctx_get_op(ctx); | |
1691 | status = dplane_ctx_get_status(ctx); | |
1692 | ||
c831033f | 1693 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
f183e380 | 1694 | zlog_debug("%u:%s Processing dplane ctx %p, op %s result %s", |
e5ac2adf | 1695 | dplane_ctx_get_vrf(ctx), dest_str, ctx, |
f183e380 | 1696 | dplane_op2str(op), dplane_res2str(status)); |
e5ac2adf | 1697 | |
e5ac2adf MS |
1698 | /* |
1699 | * Update is a bit of a special case, where we may have both old and new | |
1700 | * routes to post-process. | |
1701 | */ | |
1702 | is_update = dplane_ctx_is_update(ctx); | |
1703 | ||
1704 | /* | |
1705 | * Take a pass through the routes, look for matches with the context | |
1706 | * info. | |
1707 | */ | |
1708 | RNODE_FOREACH_RE(rn, rib) { | |
1709 | ||
1710 | if (re == NULL) { | |
5709131c | 1711 | if (rib_route_match_ctx(rib, ctx, false)) |
e5ac2adf | 1712 | re = rib; |
e5ac2adf MS |
1713 | } |
1714 | ||
1715 | /* Check for old route match */ | |
1716 | if (is_update && (old_re == NULL)) { | |
5709131c | 1717 | if (rib_route_match_ctx(rib, ctx, true /*is_update*/)) |
e5ac2adf | 1718 | old_re = rib; |
e5ac2adf MS |
1719 | } |
1720 | ||
1721 | /* Have we found the routes we need to work on? */ | |
5709131c | 1722 | if (re && ((!is_update || old_re))) |
e5ac2adf | 1723 | break; |
e5ac2adf MS |
1724 | } |
1725 | ||
699dae23 DS |
1726 | seq = dplane_ctx_get_seq(ctx); |
1727 | ||
e5ac2adf MS |
1728 | /* |
1729 | * Check sequence number(s) to detect stale results before continuing | |
1730 | */ | |
60f98b23 | 1731 | if (re) { |
699dae23 | 1732 | if (re->dplane_sequence != seq) { |
60f98b23 DS |
1733 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
1734 | zlog_debug("%u:%s Stale dplane result for re %p", | |
1735 | dplane_ctx_get_vrf(ctx), | |
1736 | dest_str, re); | |
1737 | } else | |
1738 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); | |
e5ac2adf MS |
1739 | } |
1740 | ||
60f98b23 DS |
1741 | if (old_re) { |
1742 | if (old_re->dplane_sequence != dplane_ctx_get_old_seq(ctx)) { | |
1743 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
1744 | zlog_debug("%u:%s Stale dplane result for old_re %p", | |
1745 | dplane_ctx_get_vrf(ctx), | |
1746 | dest_str, old_re); | |
1747 | } else | |
9ef0c6ba | 1748 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_QUEUED); |
e5ac2adf MS |
1749 | } |
1750 | ||
12e7fe3a | 1751 | switch (op) { |
12e7fe3a DS |
1752 | case DPLANE_OP_ROUTE_INSTALL: |
1753 | case DPLANE_OP_ROUTE_UPDATE: | |
1754 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
677c1dd5 DS |
1755 | if (re) { |
1756 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
1757 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1758 | } | |
b9f0e5ee DS |
1759 | /* |
1760 | * On an update operation from the same route type | |
1761 | * context retrieval currently has no way to know | |
1762 | * which was the old and which was the new. | |
1763 | * So don't unset our flags that we just set. | |
1764 | * We know redistribution is ok because the | |
1765 | * old_re in this case is used for nothing | |
1766 | * more than knowing whom to contact if necessary. | |
1767 | */ | |
1768 | if (old_re && old_re != re) { | |
677c1dd5 DS |
1769 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
1770 | UNSET_FLAG(old_re->status, | |
1771 | ROUTE_ENTRY_INSTALLED); | |
1772 | } | |
f183e380 | 1773 | |
78bf56b0 | 1774 | /* Update zebra route based on the results in |
941e261c | 1775 | * the context struct. |
78bf56b0 | 1776 | */ |
941e261c | 1777 | if (re) { |
efe6c026 MS |
1778 | fib_changed = |
1779 | rib_update_re_from_ctx(re, rn, ctx); | |
1780 | ||
1781 | if (!fib_changed) { | |
1782 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
1783 | zlog_debug("%u:%s no fib change for re", | |
1784 | dplane_ctx_get_vrf( | |
1785 | ctx), | |
1786 | dest_str); | |
12e7fe3a DS |
1787 | } |
1788 | ||
941e261c MS |
1789 | /* Redistribute */ |
1790 | redistribute_update(dest_pfx, src_pfx, | |
228a811a | 1791 | re, old_re); |
f183e380 MS |
1792 | } |
1793 | ||
3cdba47a DS |
1794 | /* |
1795 | * System routes are weird in that they | |
1796 | * allow multiple to be installed that match | |
1797 | * to the same prefix, so after we get the | |
1798 | * result we need to clean them up so that | |
1799 | * we can actually use them. | |
1800 | */ | |
1801 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
1802 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
1803 | zebra_rib_fixup_system(rn); | |
1804 | ||
f4c6e2a8 | 1805 | if (zvrf) |
12e7fe3a | 1806 | zvrf->installs++; |
f183e380 | 1807 | |
12e7fe3a | 1808 | /* Notify route owner */ |
677c1dd5 | 1809 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_INSTALLED); |
e5ac2adf | 1810 | |
12e7fe3a | 1811 | } else { |
7a230a9d | 1812 | if (re) { |
677c1dd5 | 1813 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); |
7a230a9d DS |
1814 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
1815 | } if (old_re) | |
677c1dd5 DS |
1816 | SET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
1817 | if (re) | |
1818 | zsend_route_notify_owner(re, dest_pfx, | |
1819 | ZAPI_ROUTE_FAIL_INSTALL); | |
97f5b441 | 1820 | |
12e7fe3a DS |
1821 | zlog_warn("%u:%s: Route install failed", |
1822 | dplane_ctx_get_vrf(ctx), | |
1823 | prefix2str(dest_pfx, | |
1824 | dest_str, sizeof(dest_str))); | |
1825 | } | |
1826 | break; | |
1827 | case DPLANE_OP_ROUTE_DELETE: | |
677c1dd5 DS |
1828 | if (re) |
1829 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
1830 | /* |
1831 | * In the delete case, the zebra core datastructs were | |
1832 | * updated (or removed) at the time the delete was issued, | |
1833 | * so we're just notifying the route owner. | |
e5ac2adf | 1834 | */ |
12e7fe3a | 1835 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { |
677c1dd5 DS |
1836 | if (re) { |
1837 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1838 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
1839 | } | |
12e7fe3a | 1840 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_REMOVED); |
e5ac2adf | 1841 | |
12e7fe3a DS |
1842 | if (zvrf) |
1843 | zvrf->removals++; | |
1844 | } else { | |
677c1dd5 DS |
1845 | if (re) |
1846 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
1847 | zsend_route_notify_owner_ctx(ctx, |
1848 | ZAPI_ROUTE_REMOVE_FAIL); | |
e5ac2adf | 1849 | |
12e7fe3a DS |
1850 | zlog_warn("%u:%s: Route Deletion failure", |
1851 | dplane_ctx_get_vrf(ctx), | |
1852 | prefix2str(dest_pfx, | |
1853 | dest_str, sizeof(dest_str))); | |
1854 | } | |
3cdba47a DS |
1855 | |
1856 | /* | |
1857 | * System routes are weird in that they | |
1858 | * allow multiple to be installed that match | |
1859 | * to the same prefix, so after we get the | |
1860 | * result we need to clean them up so that | |
1861 | * we can actually use them. | |
1862 | */ | |
1863 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
1864 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
1865 | zebra_rib_fixup_system(rn); | |
12e7fe3a | 1866 | break; |
d37f4d6c MS |
1867 | default: |
1868 | break; | |
e5ac2adf | 1869 | } |
699dae23 DS |
1870 | |
1871 | zebra_rib_evaluate_rn_nexthops(rn, seq); | |
a1494c25 | 1872 | zebra_rib_evaluate_mpls(rn); |
e5ac2adf MS |
1873 | done: |
1874 | ||
5f27bcba DS |
1875 | if (rn) |
1876 | route_unlock_node(rn); | |
1877 | ||
e5ac2adf MS |
1878 | /* Return context to dataplane module */ |
1879 | dplane_ctx_fini(&ctx); | |
1880 | } | |
1881 | ||
54818e3b MS |
1882 | /* |
1883 | * Handle notification from async dataplane: the dataplane has detected | |
1884 | * some change to a route, and notifies zebra so that the control plane | |
1885 | * can reflect that change. | |
1886 | */ | |
1887 | static void rib_process_dplane_notify(struct zebra_dplane_ctx *ctx) | |
1888 | { | |
1889 | struct route_node *rn = NULL; | |
efe6c026 MS |
1890 | struct route_entry *re = NULL; |
1891 | struct nexthop *nexthop; | |
54818e3b MS |
1892 | char dest_str[PREFIX_STRLEN] = ""; |
1893 | const struct prefix *dest_pfx, *src_pfx; | |
941e261c | 1894 | rib_dest_t *dest; |
efe6c026 | 1895 | bool fib_changed = false; |
941e261c | 1896 | bool debug_p = IS_ZEBRA_DEBUG_DPLANE | IS_ZEBRA_DEBUG_RIB; |
efe6c026 | 1897 | int start_count, end_count; |
54818e3b MS |
1898 | dest_pfx = dplane_ctx_get_dest(ctx); |
1899 | ||
1900 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1901 | * unconditional log messages will have to generate the string. | |
1902 | */ | |
941e261c | 1903 | if (debug_p) |
54818e3b MS |
1904 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
1905 | ||
1906 | /* Locate rn and re(s) from ctx */ | |
1907 | rn = rib_find_rn_from_ctx(ctx); | |
1908 | if (rn == NULL) { | |
941e261c MS |
1909 | if (debug_p) { |
1910 | zlog_debug("Failed to process dplane notification: no routes for %u:%s", | |
54818e3b MS |
1911 | dplane_ctx_get_vrf(ctx), dest_str); |
1912 | } | |
1913 | goto done; | |
1914 | } | |
1915 | ||
941e261c | 1916 | dest = rib_dest_from_rnode(rn); |
54818e3b MS |
1917 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); |
1918 | ||
efe6c026 | 1919 | if (debug_p) |
54818e3b MS |
1920 | zlog_debug("%u:%s Processing dplane notif ctx %p", |
1921 | dplane_ctx_get_vrf(ctx), dest_str, ctx); | |
1922 | ||
1923 | /* | |
1924 | * Take a pass through the routes, look for matches with the context | |
1925 | * info. | |
1926 | */ | |
efe6c026 MS |
1927 | RNODE_FOREACH_RE(rn, re) { |
1928 | if (rib_route_match_ctx(re, ctx, false /*!update*/)) | |
54818e3b MS |
1929 | break; |
1930 | } | |
1931 | ||
1932 | /* No match? Nothing we can do */ | |
1933 | if (re == NULL) { | |
941e261c MS |
1934 | if (debug_p) |
1935 | zlog_debug("%u:%s Unable to process dplane notification: no entry for type %s", | |
54818e3b | 1936 | dplane_ctx_get_vrf(ctx), dest_str, |
941e261c MS |
1937 | zebra_route_string( |
1938 | dplane_ctx_get_type(ctx))); | |
54818e3b MS |
1939 | |
1940 | goto done; | |
1941 | } | |
1942 | ||
272e8903 MS |
1943 | /* Ensure we clear the QUEUED flag */ |
1944 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); | |
1945 | ||
941e261c MS |
1946 | /* Is this a notification that ... matters? We only really care about |
1947 | * the route that is currently selected for installation. | |
1948 | */ | |
1949 | if (re != dest->selected_fib) { | |
1950 | /* TODO -- don't skip processing entirely? We might like to | |
1951 | * at least report on the event. | |
1952 | */ | |
1953 | if (debug_p) | |
efe6c026 | 1954 | zlog_debug("%u:%s dplane notif, but type %s not selected_fib", |
941e261c MS |
1955 | dplane_ctx_get_vrf(ctx), dest_str, |
1956 | zebra_route_string( | |
1957 | dplane_ctx_get_type(ctx))); | |
1958 | goto done; | |
1959 | } | |
1960 | ||
efe6c026 MS |
1961 | /* We'll want to determine whether the installation status of the |
1962 | * route has changed: we'll check the status before processing, | |
1963 | * and then again if there's been a change. | |
1964 | */ | |
1965 | start_count = 0; | |
1966 | for (ALL_NEXTHOPS_PTR(rib_active_nhg(re), nexthop)) { | |
1967 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1968 | start_count++; | |
1969 | } | |
1970 | ||
54818e3b MS |
1971 | /* Update zebra's nexthop FIB flags based on the context struct's |
1972 | * nexthops. | |
1973 | */ | |
efe6c026 | 1974 | fib_changed = rib_update_re_from_ctx(re, rn, ctx); |
54818e3b | 1975 | |
efe6c026 MS |
1976 | if (!fib_changed) { |
1977 | if (debug_p) | |
54818e3b MS |
1978 | zlog_debug("%u:%s No change from dplane notification", |
1979 | dplane_ctx_get_vrf(ctx), dest_str); | |
efe6c026 MS |
1980 | |
1981 | goto done; | |
54818e3b MS |
1982 | } |
1983 | ||
188a00e0 MS |
1984 | /* |
1985 | * Perform follow-up work if the actual status of the prefix | |
efe6c026 | 1986 | * changed. |
941e261c MS |
1987 | */ |
1988 | ||
efe6c026 MS |
1989 | end_count = 0; |
1990 | for (ALL_NEXTHOPS_PTR(rib_active_nhg(re), nexthop)) { | |
1991 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1992 | end_count++; | |
1993 | } | |
941e261c | 1994 | |
188a00e0 MS |
1995 | /* Various fib transitions: changed nexthops; from installed to |
1996 | * not-installed; or not-installed to installed. | |
efe6c026 | 1997 | */ |
188a00e0 | 1998 | if (start_count > 0 && end_count > 0) { |
11260e70 MS |
1999 | if (debug_p) |
2000 | zlog_debug("%u:%s applied nexthop changes from dplane notification", | |
2001 | dplane_ctx_get_vrf(ctx), dest_str); | |
188a00e0 MS |
2002 | |
2003 | /* Changed nexthops - update kernel/others */ | |
2004 | dplane_route_notif_update(rn, re, | |
2005 | DPLANE_OP_ROUTE_UPDATE, ctx); | |
2006 | ||
2007 | } else if (start_count == 0 && end_count > 0) { | |
efe6c026 MS |
2008 | if (debug_p) |
2009 | zlog_debug("%u:%s installed transition from dplane notification", | |
2010 | dplane_ctx_get_vrf(ctx), dest_str); | |
2011 | ||
2012 | /* We expect this to be the selected route, so we want | |
272e8903 | 2013 | * to tell others about this transition. |
efe6c026 MS |
2014 | */ |
2015 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
2016 | ||
188a00e0 MS |
2017 | /* Changed nexthops - update kernel/others */ |
2018 | dplane_route_notif_update(rn, re, DPLANE_OP_ROUTE_INSTALL, ctx); | |
2019 | ||
efe6c026 MS |
2020 | /* Redistribute, lsp, and nht update */ |
2021 | redistribute_update(dest_pfx, src_pfx, re, NULL); | |
2022 | ||
2023 | zebra_rib_evaluate_rn_nexthops( | |
2024 | rn, zebra_router_get_next_sequence()); | |
2025 | ||
2026 | zebra_rib_evaluate_mpls(rn); | |
2027 | ||
2028 | } else if (start_count > 0 && end_count == 0) { | |
2029 | if (debug_p) | |
2030 | zlog_debug("%u:%s un-installed transition from dplane notification", | |
2031 | dplane_ctx_get_vrf(ctx), dest_str); | |
2032 | ||
2033 | /* Transition from _something_ installed to _nothing_ | |
2034 | * installed. | |
2035 | */ | |
2036 | /* We expect this to be the selected route, so we want | |
2037 | * to tell others about this transistion. | |
2038 | */ | |
2039 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
2040 | ||
188a00e0 MS |
2041 | /* Changed nexthops - update kernel/others */ |
2042 | dplane_route_notif_update(rn, re, DPLANE_OP_ROUTE_DELETE, ctx); | |
2043 | ||
efe6c026 | 2044 | /* Redistribute, lsp, and nht update */ |
40f321c0 | 2045 | redistribute_delete(dest_pfx, src_pfx, re, NULL); |
efe6c026 MS |
2046 | |
2047 | zebra_rib_evaluate_rn_nexthops( | |
2048 | rn, zebra_router_get_next_sequence()); | |
2049 | ||
2050 | zebra_rib_evaluate_mpls(rn); | |
2051 | } | |
941e261c | 2052 | |
54818e3b | 2053 | done: |
941e261c MS |
2054 | if (rn) |
2055 | route_unlock_node(rn); | |
2056 | ||
54818e3b MS |
2057 | /* Return context to dataplane module */ |
2058 | dplane_ctx_fini(&ctx); | |
2059 | } | |
2060 | ||
5110a0c6 | 2061 | /* Take a list of route_node structs and return 1, if there was a record |
d62a17ae | 2062 | * picked from it and processed by rib_process(). Don't process more, |
5110a0c6 | 2063 | * than one RN record; operate only in the specified sub-queue. |
e96f9203 | 2064 | */ |
d7c0a89a | 2065 | static unsigned int process_subq(struct list *subq, uint8_t qindex) |
e96f9203 | 2066 | { |
d62a17ae | 2067 | struct listnode *lnode = listhead(subq); |
2068 | struct route_node *rnode; | |
2069 | rib_dest_t *dest; | |
2070 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 2071 | |
d62a17ae | 2072 | if (!lnode) |
2073 | return 0; | |
5110a0c6 | 2074 | |
d62a17ae | 2075 | rnode = listgetdata(lnode); |
2076 | dest = rib_dest_from_rnode(rnode); | |
2077 | if (dest) | |
2078 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 2079 | |
d62a17ae | 2080 | rib_process(rnode); |
5110a0c6 | 2081 | |
d62a17ae | 2082 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
2083 | char buf[SRCDEST2STR_BUFFER]; | |
efe6c026 | 2084 | |
d62a17ae | 2085 | srcdest_rnode2str(rnode, buf, sizeof(buf)); |
2086 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
2087 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
2088 | } | |
41ec9222 | 2089 | |
d62a17ae | 2090 | if (rnode->info) |
2091 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
2092 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 2093 | |
67b9467f | 2094 | #if 0 |
5110a0c6 SH |
2095 | else |
2096 | { | |
2097 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
2098 | __func__, rnode, rnode->lock); | |
2099 | zlog_backtrace(LOG_DEBUG); | |
2100 | } | |
67b9467f | 2101 | #endif |
d62a17ae | 2102 | route_unlock_node(rnode); |
2103 | list_delete_node(subq, lnode); | |
2104 | return 1; | |
e96f9203 DO |
2105 | } |
2106 | ||
2107 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from | |
d62a17ae | 2108 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
2109 | * data | |
e96f9203 DO |
2110 | * is pointed to the meta queue structure. |
2111 | */ | |
d62a17ae | 2112 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 2113 | { |
d62a17ae | 2114 | struct meta_queue *mq = data; |
2115 | unsigned i; | |
91f16812 MS |
2116 | uint32_t queue_len, queue_limit; |
2117 | ||
2118 | /* Ensure there's room for more dataplane updates */ | |
2119 | queue_limit = dplane_get_in_queue_limit(); | |
2120 | queue_len = dplane_get_in_queue_len(); | |
2121 | if (queue_len > queue_limit) { | |
2122 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
8b962e77 MS |
2123 | zlog_debug("rib queue: dplane queue len %u, limit %u, retrying", |
2124 | queue_len, queue_limit); | |
91f16812 MS |
2125 | |
2126 | /* Ensure that the meta-queue is actually enqueued */ | |
489a9614 | 2127 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2128 | work_queue_add(zrouter.ribq, zrouter.mq); |
91f16812 MS |
2129 | |
2130 | return WQ_QUEUE_BLOCKED; | |
2131 | } | |
5110a0c6 | 2132 | |
d62a17ae | 2133 | for (i = 0; i < MQ_SIZE; i++) |
2134 | if (process_subq(mq->subq[i], i)) { | |
2135 | mq->size--; | |
2136 | break; | |
2137 | } | |
2138 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
2139 | } |
2140 | ||
f52ed677 DS |
2141 | |
2142 | /* | |
2143 | * Look into the RN and queue it into the highest priority queue | |
2144 | * at this point in time for processing. | |
2145 | * | |
2146 | * We will enqueue a route node only once per invocation. | |
2147 | * | |
2148 | * There are two possibilities here that should be kept in mind. | |
2149 | * If the original invocation has not been pulled off for processing | |
2150 | * yet, A subsuquent invocation can have a route entry with a better | |
2151 | * meta queue index value and we can have a situation where | |
2152 | * we might have the same node enqueued 2 times. Not necessarily | |
2153 | * an optimal situation but it should be ok. | |
2154 | * | |
2155 | * The other possibility is that the original invocation has not | |
2156 | * been pulled off for processing yet, A subsusquent invocation | |
2157 | * doesn't have a route_entry with a better meta-queue and the | |
2158 | * original metaqueue index value will win and we'll end up with | |
2159 | * the route node enqueued once. | |
e96f9203 | 2160 | */ |
d62a17ae | 2161 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) |
e96f9203 | 2162 | { |
f52ed677 DS |
2163 | struct route_entry *re = NULL, *curr_re = NULL; |
2164 | uint8_t qindex = MQ_SIZE, curr_qindex = MQ_SIZE; | |
5110a0c6 | 2165 | |
f52ed677 DS |
2166 | RNODE_FOREACH_RE (rn, curr_re) { |
2167 | curr_qindex = route_info[curr_re->type].meta_q_map; | |
d62a17ae | 2168 | |
f52ed677 DS |
2169 | if (curr_qindex <= qindex) { |
2170 | re = curr_re; | |
2171 | qindex = curr_qindex; | |
d62a17ae | 2172 | } |
f52ed677 | 2173 | } |
5110a0c6 | 2174 | |
f52ed677 DS |
2175 | if (!re) |
2176 | return; | |
5110a0c6 | 2177 | |
f52ed677 DS |
2178 | /* Invariant: at this point we always have rn->info set. */ |
2179 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
2180 | RIB_ROUTE_QUEUED(qindex))) { | |
d62a17ae | 2181 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2182 | rnode_debug(rn, re->vrf_id, | |
f52ed677 | 2183 | "rn %p is already queued in sub-queue %u", |
d62a17ae | 2184 | (void *)rn, qindex); |
f52ed677 | 2185 | return; |
d62a17ae | 2186 | } |
f52ed677 DS |
2187 | |
2188 | SET_FLAG(rib_dest_from_rnode(rn)->flags, RIB_ROUTE_QUEUED(qindex)); | |
2189 | listnode_add(mq->subq[qindex], rn); | |
2190 | route_lock_node(rn); | |
2191 | mq->size++; | |
2192 | ||
2193 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2194 | rnode_debug(rn, re->vrf_id, "queued rn %p into sub-queue %u", | |
2195 | (void *)rn, qindex); | |
4d38fdb4 | 2196 | } |
2197 | ||
6d691129 | 2198 | /* Add route_node to work queue and schedule processing */ |
d62a17ae | 2199 | void rib_queue_add(struct route_node *rn) |
4d38fdb4 | 2200 | { |
d62a17ae | 2201 | assert(rn); |
fc328ac9 | 2202 | |
d62a17ae | 2203 | /* Pointless to queue a route_node with no RIB entries to add or remove |
2204 | */ | |
2205 | if (!rnode_to_ribs(rn)) { | |
2206 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
2207 | __func__, (void *)rn, rn->lock); | |
2208 | zlog_backtrace(LOG_DEBUG); | |
2209 | return; | |
2210 | } | |
4d38fdb4 | 2211 | |
489a9614 | 2212 | if (zrouter.ribq == NULL) { |
e914ccbe | 2213 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2214 | "%s: work_queue does not exist!", __func__); |
d62a17ae | 2215 | return; |
2216 | } | |
2217 | ||
2218 | /* | |
2219 | * The RIB queue should normally be either empty or holding the only | |
2220 | * work_queue_item element. In the latter case this element would | |
2221 | * hold a pointer to the meta queue structure, which must be used to | |
2222 | * actually queue the route nodes to process. So create the MQ | |
2223 | * holder, if necessary, then push the work into it in any case. | |
2224 | * This semantics was introduced after 0.99.9 release. | |
2225 | */ | |
489a9614 | 2226 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2227 | work_queue_add(zrouter.ribq, zrouter.mq); |
e96f9203 | 2228 | |
ea45a4e7 | 2229 | rib_meta_queue_add(zrouter.mq, rn); |
fc328ac9 | 2230 | |
d62a17ae | 2231 | return; |
4d38fdb4 | 2232 | } |
2233 | ||
5110a0c6 SH |
2234 | /* Create new meta queue. |
2235 | A destructor function doesn't seem to be necessary here. | |
2236 | */ | |
d62a17ae | 2237 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 2238 | { |
d62a17ae | 2239 | struct meta_queue *new; |
2240 | unsigned i; | |
5110a0c6 | 2241 | |
d62a17ae | 2242 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
e96f9203 | 2243 | |
d62a17ae | 2244 | for (i = 0; i < MQ_SIZE; i++) { |
2245 | new->subq[i] = list_new(); | |
2246 | assert(new->subq[i]); | |
2247 | } | |
5110a0c6 | 2248 | |
d62a17ae | 2249 | return new; |
e96f9203 DO |
2250 | } |
2251 | ||
d62a17ae | 2252 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 2253 | { |
d62a17ae | 2254 | unsigned i; |
5a8dfcd8 | 2255 | |
d62a17ae | 2256 | for (i = 0; i < MQ_SIZE; i++) |
6a154c88 | 2257 | list_delete(&mq->subq[i]); |
5a8dfcd8 | 2258 | |
d62a17ae | 2259 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
2260 | } |
2261 | ||
4d38fdb4 | 2262 | /* initialise zebra rib work queue */ |
2561d12e | 2263 | static void rib_queue_init(void) |
4d38fdb4 | 2264 | { |
489a9614 DS |
2265 | if (!(zrouter.ribq = work_queue_new(zrouter.master, |
2266 | "route_node processing"))) { | |
e914ccbe | 2267 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2268 | "%s: could not initialise work queue!", __func__); |
d62a17ae | 2269 | return; |
2270 | } | |
4d38fdb4 | 2271 | |
d62a17ae | 2272 | /* fill in the work queue spec */ |
489a9614 DS |
2273 | zrouter.ribq->spec.workfunc = &meta_queue_process; |
2274 | zrouter.ribq->spec.errorfunc = NULL; | |
46a4e345 | 2275 | zrouter.ribq->spec.completion_func = NULL; |
d62a17ae | 2276 | /* XXX: TODO: These should be runtime configurable via vty */ |
489a9614 DS |
2277 | zrouter.ribq->spec.max_retries = 3; |
2278 | zrouter.ribq->spec.hold = ZEBRA_RIB_PROCESS_HOLD_TIME; | |
2279 | zrouter.ribq->spec.retry = ZEBRA_RIB_PROCESS_RETRY_TIME; | |
d62a17ae | 2280 | |
ea45a4e7 | 2281 | if (!(zrouter.mq = meta_queue_new())) { |
e914ccbe | 2282 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2283 | "%s: could not initialise meta queue!", __func__); |
d62a17ae | 2284 | return; |
2285 | } | |
2286 | return; | |
718e3744 | 2287 | } |
2288 | ||
434434f7 DS |
2289 | rib_dest_t *zebra_rib_create_dest(struct route_node *rn) |
2290 | { | |
2291 | rib_dest_t *dest; | |
2292 | ||
2293 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); | |
aa57abfb | 2294 | rnh_list_init(&dest->nht); |
434434f7 DS |
2295 | route_lock_node(rn); /* rn route table reference */ |
2296 | rn->info = dest; | |
2297 | dest->rnode = rn; | |
2298 | ||
2299 | return dest; | |
2300 | } | |
2301 | ||
6d691129 PJ |
2302 | /* RIB updates are processed via a queue of pointers to route_nodes. |
2303 | * | |
2304 | * The queue length is bounded by the maximal size of the routing table, | |
2305 | * as a route_node will not be requeued, if already queued. | |
2306 | * | |
f0f77c9a DS |
2307 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
2308 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 2309 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 2310 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 2311 | * |
f0f77c9a | 2312 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
2313 | * |
2314 | * rib_addnode | |
f0f77c9a DS |
2315 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
2316 | * |-------->| | best RE, if required | |
3c0755dc PJ |
2317 | * | | |
2318 | * static_install->|->rib_addqueue...... -> rib_process | |
2319 | * | | | |
2320 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
2321 | * |-> set ROUTE_ENTRY_REMOVE | |
2322 | * rib_delnode (RE freed) | |
6d691129 | 2323 | * |
9fd92e3c AS |
2324 | * The 'info' pointer of a route_node points to a rib_dest_t |
2325 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
2326 | * dest is created on-demand by rib_link() and is kept around at least | |
2327 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 2328 | * |
6d691129 PJ |
2329 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
2330 | * | |
2331 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
2332 | * - dest attached to route_node: |
2333 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
2334 | * - route_node processing queue |
2335 | * - managed by: rib_addqueue, rib_process. | |
2336 | * | |
2337 | */ | |
d62a17ae | 2338 | |
f0f77c9a | 2339 | /* Add RE to head of the route node. */ |
d62a17ae | 2340 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
2341 | { | |
d62a17ae | 2342 | rib_dest_t *dest; |
2343 | afi_t afi; | |
2344 | const char *rmap_name; | |
9fd92e3c | 2345 | |
d62a17ae | 2346 | assert(re && rn); |
9fd92e3c | 2347 | |
d62a17ae | 2348 | dest = rib_dest_from_rnode(rn); |
2349 | if (!dest) { | |
2350 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2351 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 2352 | |
434434f7 | 2353 | dest = zebra_rib_create_dest(rn); |
d62a17ae | 2354 | } |
2263a412 | 2355 | |
045207e2 | 2356 | re_list_add_head(&dest->routes, re); |
d62a17ae | 2357 | |
2358 | afi = (rn->p.family == AF_INET) | |
2359 | ? AFI_IP | |
2360 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
fe257ae7 DS |
2361 | if (is_zebra_import_table_enabled(afi, re->vrf_id, re->table)) { |
2362 | struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(re->vrf_id); | |
2363 | ||
d62a17ae | 2364 | rmap_name = zebra_get_import_table_route_map(afi, re->table); |
fe257ae7 | 2365 | zebra_add_import_table_entry(zvrf, rn, re, rmap_name); |
d62a17ae | 2366 | } else if (process) |
2367 | rib_queue_add(rn); | |
2368 | } | |
2369 | ||
7e24fdf3 DS |
2370 | static void rib_addnode(struct route_node *rn, |
2371 | struct route_entry *re, int process) | |
d62a17ae | 2372 | { |
2373 | /* RE node has been un-removed before route-node is processed. | |
2374 | * route_node must hence already be on the queue for processing.. | |
2375 | */ | |
2376 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
2377 | if (IS_ZEBRA_DEBUG_RIB) | |
2378 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
2379 | (void *)rn, (void *)re); | |
2380 | ||
2381 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2382 | return; | |
2383 | } | |
2384 | rib_link(rn, re, process); | |
6d691129 PJ |
2385 | } |
2386 | ||
9fd92e3c AS |
2387 | /* |
2388 | * rib_unlink | |
2389 | * | |
2390 | * Detach a rib structure from a route_node. | |
2391 | * | |
2392 | * Note that a call to rib_unlink() should be followed by a call to | |
2393 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
2394 | * longer required to be deleted. | |
2395 | */ | |
d62a17ae | 2396 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 2397 | { |
d62a17ae | 2398 | rib_dest_t *dest; |
22bcedb2 | 2399 | rib_table_info_t *info; |
9fd92e3c | 2400 | |
d62a17ae | 2401 | assert(rn && re); |
6d691129 | 2402 | |
d62a17ae | 2403 | if (IS_ZEBRA_DEBUG_RIB) |
2404 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
2405 | (void *)re); | |
6d691129 | 2406 | |
d62a17ae | 2407 | dest = rib_dest_from_rnode(rn); |
6d691129 | 2408 | |
045207e2 | 2409 | re_list_del(&dest->routes, re); |
7a4bb9c5 | 2410 | |
2eb07de3 DS |
2411 | if (dest->selected_fib == re) |
2412 | dest->selected_fib = NULL; | |
2413 | ||
22bcedb2 DS |
2414 | info = srcdest_rnode_table_info(rn); |
2415 | zebra_nhg_release(info->afi, re); | |
2416 | ||
6b468511 DS |
2417 | nexthops_free(re->ng->nexthop); |
2418 | nexthop_group_delete(&re->ng); | |
ee5e8a48 MS |
2419 | nexthops_free(re->fib_ng.nexthop); |
2420 | ||
d62a17ae | 2421 | XFREE(MTYPE_RE, re); |
2422 | } | |
2423 | ||
2424 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2425 | { | |
2426 | afi_t afi; | |
2427 | ||
2428 | if (IS_ZEBRA_DEBUG_RIB) | |
2429 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2430 | (void *)rn, (void *)re); | |
2431 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2432 | ||
2433 | afi = (rn->p.family == AF_INET) | |
2434 | ? AFI_IP | |
2435 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
fe257ae7 DS |
2436 | if (is_zebra_import_table_enabled(afi, re->vrf_id, re->table)) { |
2437 | struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(re->vrf_id); | |
2438 | ||
2439 | zebra_del_import_table_entry(zvrf, rn, re); | |
d62a17ae | 2440 | /* Just clean up if non main table */ |
2441 | if (IS_ZEBRA_DEBUG_RIB) { | |
2442 | char buf[SRCDEST2STR_BUFFER]; | |
2443 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
2444 | zlog_debug("%u:%s: Freeing route rn %p, re %p (%s)", |
2445 | re->vrf_id, buf, rn, re, | |
2446 | zebra_route_string(re->type)); | |
d62a17ae | 2447 | } |
7a4bb9c5 | 2448 | |
d62a17ae | 2449 | rib_unlink(rn, re); |
2450 | } else { | |
2451 | rib_queue_add(rn); | |
2452 | } | |
718e3744 | 2453 | } |
2454 | ||
f0f77c9a | 2455 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2456 | * standard debug log. Calling function name and IP prefix in |
2457 | * question are passed as 1st and 2nd arguments. | |
2458 | */ | |
2459 | ||
d62a17ae | 2460 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2461 | union prefixconstptr src_pp, | |
2462 | const struct route_entry *re) | |
2463 | { | |
d62a17ae | 2464 | const struct prefix *src_p = src_pp.p; |
2465 | bool is_srcdst = src_p && src_p->prefixlen; | |
2466 | char straddr[PREFIX_STRLEN]; | |
2467 | char srcaddr[PREFIX_STRLEN]; | |
53c16fbe | 2468 | char nhname[PREFIX_STRLEN]; |
d62a17ae | 2469 | struct nexthop *nexthop; |
2470 | ||
2471 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2472 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2473 | is_srcdst ? " from " : "", | |
2474 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2475 | : "", | |
2476 | re->vrf_id); | |
cc54cfee | 2477 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", |
53c16fbe | 2478 | straddr, (unsigned long)re->uptime, re->type, re->instance, |
cc54cfee | 2479 | re->table); |
d62a17ae | 2480 | zlog_debug( |
2481 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
53c16fbe DS |
2482 | straddr, re->metric, re->mtu, re->distance, re->flags, re->status); |
2483 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", straddr, | |
eecacedc | 2484 | nexthop_group_nexthop_num(re->ng), re->nexthop_active_num); |
d62a17ae | 2485 | |
6b468511 | 2486 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) { |
2d68a0f2 DS |
2487 | struct interface *ifp; |
2488 | struct vrf *vrf = vrf_lookup_by_id(nexthop->vrf_id); | |
2489 | ||
2490 | switch (nexthop->type) { | |
2491 | case NEXTHOP_TYPE_BLACKHOLE: | |
53c16fbe | 2492 | sprintf(nhname, "Blackhole"); |
2d68a0f2 DS |
2493 | break; |
2494 | case NEXTHOP_TYPE_IFINDEX: | |
2495 | ifp = if_lookup_by_index(nexthop->ifindex, | |
2496 | nexthop->vrf_id); | |
53c16fbe | 2497 | sprintf(nhname, "%s", ifp ? ifp->name : "Unknown"); |
2d68a0f2 DS |
2498 | break; |
2499 | case NEXTHOP_TYPE_IPV4: | |
2500 | /* fallthrough */ | |
2501 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
53c16fbe | 2502 | inet_ntop(AF_INET, &nexthop->gate, nhname, |
2d68a0f2 DS |
2503 | INET6_ADDRSTRLEN); |
2504 | break; | |
2505 | case NEXTHOP_TYPE_IPV6: | |
2506 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
53c16fbe | 2507 | inet_ntop(AF_INET6, &nexthop->gate, nhname, |
2d68a0f2 DS |
2508 | INET6_ADDRSTRLEN); |
2509 | break; | |
2510 | } | |
df38b099 | 2511 | zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s%s%s%s", |
53c16fbe | 2512 | straddr, (nexthop->rparent ? " NH" : "NH"), nhname, |
2d68a0f2 DS |
2513 | nexthop->ifindex, vrf ? vrf->name : "Unknown", |
2514 | nexthop->vrf_id, | |
d62a17ae | 2515 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) |
2516 | ? "ACTIVE " | |
2517 | : ""), | |
677c1dd5 | 2518 | (CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED) |
d62a17ae | 2519 | ? "FIB " |
2520 | : ""), | |
2521 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
df38b099 DS |
2522 | ? "RECURSIVE " |
2523 | : ""), | |
2524 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK) | |
2525 | ? "ONLINK " | |
2526 | : ""), | |
2527 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_MATCHED) | |
2528 | ? "MATCHED " | |
2529 | : ""), | |
2530 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_DUPLICATE) | |
2531 | ? "DUPLICATE " | |
d62a17ae | 2532 | : "")); |
2533 | } | |
53c16fbe | 2534 | zlog_debug("%s: dump complete", straddr); |
dc95824a DO |
2535 | } |
2536 | ||
2537 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2538 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2539 | */ |
2540 | ||
d62a17ae | 2541 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2542 | { | |
2543 | struct route_table *table; | |
2544 | struct route_node *rn; | |
2545 | struct route_entry *re; | |
2546 | char prefix_buf[INET_ADDRSTRLEN]; | |
2547 | ||
2548 | /* Lookup table. */ | |
2549 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2550 | if (!table) { | |
e914ccbe | 2551 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2552 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2553 | vrf_id); | |
d62a17ae | 2554 | return; |
2555 | } | |
2556 | ||
2557 | /* Scan the RIB table for exactly matching RE entry. */ | |
2558 | rn = route_node_lookup(table, (struct prefix *)p); | |
2559 | ||
2560 | /* No route for this prefix. */ | |
2561 | if (!rn) { | |
32391aff | 2562 | zlog_debug("%s:%u lookup failed for %s", __func__, vrf_id, |
d62a17ae | 2563 | prefix2str((struct prefix *)p, prefix_buf, |
2564 | sizeof(prefix_buf))); | |
2565 | return; | |
2566 | } | |
2567 | ||
2568 | /* Unlock node. */ | |
2569 | route_unlock_node(rn); | |
2570 | ||
2571 | /* let's go */ | |
a2addae8 | 2572 | RNODE_FOREACH_RE (rn, re) { |
32391aff DS |
2573 | zlog_debug("%s:%u rn %p, re %p: %s, %s", |
2574 | __func__, vrf_id, | |
2575 | (void *)rn, (void *)re, | |
d62a17ae | 2576 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) |
2577 | ? "removed" | |
2578 | : "NOT removed"), | |
2579 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2580 | ? "selected" | |
2581 | : "NOT selected")); | |
2582 | route_entry_dump(p, NULL, re); | |
2583 | } | |
dc95824a DO |
2584 | } |
2585 | ||
20e5ff0a DO |
2586 | /* Check if requested address assignment will fail due to another |
2587 | * route being installed by zebra in FIB already. Take necessary | |
2588 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2589 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2590 | */ |
d62a17ae | 2591 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2592 | { | |
2593 | struct route_table *table; | |
2594 | struct route_node *rn; | |
5f7a4718 | 2595 | rib_dest_t *dest; |
d62a17ae | 2596 | |
2597 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
e914ccbe | 2598 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2599 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2600 | vrf_id); | |
d62a17ae | 2601 | return; |
2602 | } | |
2603 | ||
2604 | /* No matches would be the simplest case. */ | |
2605 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2606 | return; | |
2607 | ||
2608 | /* Unlock node. */ | |
2609 | route_unlock_node(rn); | |
2610 | ||
5f7a4718 | 2611 | dest = rib_dest_from_rnode(rn); |
d62a17ae | 2612 | /* Check all RE entries. In case any changes have to be done, requeue |
2613 | * the RN into RIBQ head. If the routing message about the new connected | |
2614 | * route (generated by the IP address we are going to assign very soon) | |
2615 | * comes before the RIBQ is processed, the new RE entry will join | |
2616 | * RIBQ record already on head. This is necessary for proper | |
2617 | * revalidation | |
2618 | * of the rest of the RE. | |
2619 | */ | |
3cdba47a | 2620 | if (dest->selected_fib) { |
5f7a4718 DS |
2621 | if (IS_ZEBRA_DEBUG_RIB) { |
2622 | char buf[PREFIX_STRLEN]; | |
2623 | ||
2624 | zlog_debug("%u:%s: freeing way for connected prefix", | |
2625 | dest->selected_fib->vrf_id, | |
2626 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2627 | route_entry_dump(&rn->p, NULL, dest->selected_fib); | |
d62a17ae | 2628 | } |
5f7a4718 | 2629 | rib_uninstall(rn, dest->selected_fib); |
d62a17ae | 2630 | rib_queue_add(rn); |
ade4a886 | 2631 | } |
20e5ff0a DO |
2632 | } |
2633 | ||
d62a17ae | 2634 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
2635 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
718e3744 | 2636 | { |
d62a17ae | 2637 | struct route_table *table; |
2638 | struct route_node *rn; | |
40ecd8e4 | 2639 | struct route_entry *same = NULL; |
d62a17ae | 2640 | int ret = 0; |
b4c034b0 | 2641 | |
d62a17ae | 2642 | if (!re) |
2643 | return 0; | |
b4c034b0 | 2644 | |
1f610a1f | 2645 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
05737783 | 2646 | |
d62a17ae | 2647 | /* Lookup table. */ |
7865c65d RW |
2648 | table = zebra_vrf_table_with_table_id(afi, safi, re->vrf_id, re->table); |
2649 | if (!table) { | |
6b468511 DS |
2650 | nexthops_free(re->ng->nexthop); |
2651 | nexthop_group_delete(&re->ng); | |
7865c65d | 2652 | XFREE(MTYPE_RE, re); |
d62a17ae | 2653 | return 0; |
7865c65d | 2654 | } |
cddf391b | 2655 | |
d62a17ae | 2656 | /* Make it sure prefixlen is applied to the prefix. */ |
2657 | apply_mask(p); | |
2658 | if (src_p) | |
2659 | apply_mask_ipv6(src_p); | |
718e3744 | 2660 | |
6b468511 | 2661 | zebra_nhg_find(afi, re->ng, re); |
d62a17ae | 2662 | /* Set default distance by route type. */ |
eea2ef07 | 2663 | if (re->distance == 0) |
0492eea0 | 2664 | re->distance = route_distance(re->type); |
718e3744 | 2665 | |
d62a17ae | 2666 | /* Lookup route node.*/ |
2667 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2668 | |
40ecd8e4 DS |
2669 | /* |
2670 | * If same type of route are installed, treat it as a implicit | |
2671 | * withdraw. | |
2672 | * If the user has specified the No route replace semantics | |
2673 | * for the install don't do a route replace. | |
2674 | */ | |
a2addae8 | 2675 | RNODE_FOREACH_RE (rn, same) { |
d62a17ae | 2676 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) |
2677 | continue; | |
41ec9222 | 2678 | |
eb327fa5 RW |
2679 | if (same->type != re->type) |
2680 | continue; | |
2681 | if (same->instance != re->instance) | |
2682 | continue; | |
996c9314 LB |
2683 | if (same->type == ZEBRA_ROUTE_KERNEL |
2684 | && same->metric != re->metric) | |
eb327fa5 | 2685 | continue; |
40ecd8e4 DS |
2686 | |
2687 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && | |
2688 | same->distance != re->distance) | |
2689 | continue; | |
2690 | ||
844b3a87 | 2691 | /* |
40ecd8e4 DS |
2692 | * We should allow duplicate connected routes |
2693 | * because of IPv6 link-local routes and unnumbered | |
2694 | * interfaces on Linux. | |
844b3a87 RW |
2695 | */ |
2696 | if (same->type != ZEBRA_ROUTE_CONNECT) | |
d62a17ae | 2697 | break; |
2698 | } | |
718e3744 | 2699 | |
cf363e1b MS |
2700 | /* If this route is kernel/connected route, notify the dataplane. */ |
2701 | if (RIB_SYSTEM_ROUTE(re)) { | |
2702 | /* Notify dataplane */ | |
2703 | dplane_sys_route_add(rn, re); | |
2704 | } | |
2705 | ||
d62a17ae | 2706 | /* Link new re to node.*/ |
2707 | if (IS_ZEBRA_DEBUG_RIB) { | |
2da33d6b DS |
2708 | rnode_debug(rn, re->vrf_id, |
2709 | "Inserting route rn %p, re %p (%s) existing %p", | |
2710 | rn, re, zebra_route_string(re->type), same); | |
718e3744 | 2711 | |
d62a17ae | 2712 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2713 | route_entry_dump(p, src_p, re); | |
718e3744 | 2714 | } |
6883bf8d DS |
2715 | |
2716 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 2717 | rib_addnode(rn, re, 1); |
2718 | ret = 1; | |
6b0655a2 | 2719 | |
d62a17ae | 2720 | /* Free implicit route.*/ |
2721 | if (same) { | |
2722 | rib_delnode(rn, same); | |
2723 | ret = -1; | |
2724 | } | |
718e3744 | 2725 | |
d62a17ae | 2726 | route_unlock_node(rn); |
2727 | return ret; | |
2728 | } | |
2729 | ||
2730 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
d7c0a89a | 2731 | unsigned short instance, int flags, struct prefix *p, |
fd36be7e | 2732 | struct prefix_ipv6 *src_p, const struct nexthop *nh, |
40ecd8e4 DS |
2733 | uint32_t table_id, uint32_t metric, uint8_t distance, |
2734 | bool fromkernel) | |
d62a17ae | 2735 | { |
2736 | struct route_table *table; | |
2737 | struct route_node *rn; | |
2738 | struct route_entry *re; | |
2739 | struct route_entry *fib = NULL; | |
2740 | struct route_entry *same = NULL; | |
fd36be7e | 2741 | struct nexthop *rtnh; |
d62a17ae | 2742 | char buf2[INET6_ADDRSTRLEN]; |
5f7a4718 | 2743 | rib_dest_t *dest; |
d62a17ae | 2744 | |
1f610a1f | 2745 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
d62a17ae | 2746 | |
2747 | /* Lookup table. */ | |
2748 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2749 | if (!table) | |
2750 | return; | |
2751 | ||
2752 | /* Apply mask. */ | |
2753 | apply_mask(p); | |
2754 | if (src_p) | |
2755 | apply_mask_ipv6(src_p); | |
2756 | ||
2757 | /* Lookup route node. */ | |
2758 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2759 | if (!rn) { | |
2760 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2761 | ||
2762 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2763 | if (src_p && src_p->prefixlen) | |
2764 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2765 | else | |
2766 | src_buf[0] = '\0'; | |
2767 | ||
82d7d0e2 DS |
2768 | if (IS_ZEBRA_DEBUG_RIB) { |
2769 | struct vrf *vrf = vrf_lookup_by_id(vrf_id); | |
2770 | ||
2771 | zlog_debug("%s[%d]:%s%s%s doesn't exist in rib", | |
2772 | vrf->name, table_id, dst_buf, | |
d62a17ae | 2773 | (src_buf[0] != '\0') ? " from " : "", |
2774 | src_buf); | |
82d7d0e2 | 2775 | } |
d62a17ae | 2776 | return; |
2777 | } | |
718e3744 | 2778 | |
5f7a4718 DS |
2779 | dest = rib_dest_from_rnode(rn); |
2780 | fib = dest->selected_fib; | |
2781 | ||
d62a17ae | 2782 | /* Lookup same type route. */ |
a2addae8 | 2783 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2784 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2785 | continue; | |
2786 | ||
d62a17ae | 2787 | if (re->type != type) |
2788 | continue; | |
2789 | if (re->instance != instance) | |
2790 | continue; | |
40ecd8e4 DS |
2791 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && |
2792 | distance != re->distance) | |
2793 | continue; | |
2794 | ||
996c9314 | 2795 | if (re->type == ZEBRA_ROUTE_KERNEL && re->metric != metric) |
f19435a8 | 2796 | continue; |
6b468511 | 2797 | if (re->type == ZEBRA_ROUTE_CONNECT && (rtnh = re->ng->nexthop) |
fd36be7e DL |
2798 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { |
2799 | if (rtnh->ifindex != nh->ifindex) | |
d62a17ae | 2800 | continue; |
d62a17ae | 2801 | same = re; |
2802 | break; | |
2803 | } | |
2804 | /* Make sure that the route found has the same gateway. */ | |
2805 | else { | |
fd36be7e | 2806 | if (nh == NULL) { |
d62a17ae | 2807 | same = re; |
2808 | break; | |
2809 | } | |
6b468511 | 2810 | for (ALL_NEXTHOPS_PTR(re->ng, rtnh)) |
a5a2d802 SW |
2811 | /* |
2812 | * No guarantee all kernel send nh with labels | |
2813 | * on delete. | |
2814 | */ | |
2815 | if (nexthop_same_no_labels(rtnh, nh)) { | |
d62a17ae | 2816 | same = re; |
2817 | break; | |
2818 | } | |
2819 | if (same) | |
2820 | break; | |
2821 | } | |
2822 | } | |
2823 | /* If same type of route can't be found and this message is from | |
2824 | kernel. */ | |
2825 | if (!same) { | |
5dfeba19 DS |
2826 | /* |
2827 | * In the past(HA!) we could get here because | |
2828 | * we were receiving a route delete from the | |
2829 | * kernel and we're not marking the proto | |
2830 | * as coming from it's appropriate originator. | |
2831 | * Now that we are properly noticing the fact | |
2832 | * that the kernel has deleted our route we | |
2833 | * are not going to get called in this path | |
2834 | * I am going to leave this here because | |
2835 | * this might still work this way on non-linux | |
2836 | * platforms as well as some weird state I have | |
2837 | * not properly thought of yet. | |
2838 | * If we can show that this code path is | |
2839 | * dead then we can remove it. | |
2840 | */ | |
b8faa875 | 2841 | if (fib && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { |
d62a17ae | 2842 | if (IS_ZEBRA_DEBUG_RIB) { |
2da33d6b DS |
2843 | rnode_debug(rn, vrf_id, |
2844 | "rn %p, re %p (%s) was deleted from kernel, adding", | |
2845 | rn, fib, | |
2846 | zebra_route_string(fib->type)); | |
d62a17ae | 2847 | } |
2848 | if (allow_delete) { | |
677c1dd5 | 2849 | UNSET_FLAG(fib->status, ROUTE_ENTRY_INSTALLED); |
d62a17ae | 2850 | /* Unset flags. */ |
6b468511 | 2851 | for (rtnh = fib->ng->nexthop; rtnh; |
fd36be7e DL |
2852 | rtnh = rtnh->next) |
2853 | UNSET_FLAG(rtnh->flags, | |
d62a17ae | 2854 | NEXTHOP_FLAG_FIB); |
2855 | ||
ed216282 DS |
2856 | /* |
2857 | * This is a non FRR route | |
2858 | * as such we should mark | |
2859 | * it as deleted | |
2860 | */ | |
5f7a4718 | 2861 | dest->selected_fib = NULL; |
d62a17ae | 2862 | } else { |
2863 | /* This means someone else, other than Zebra, | |
2864 | * has deleted | |
2865 | * a Zebra router from the kernel. We will add | |
2866 | * it back */ | |
2867 | rib_install_kernel(rn, fib, NULL); | |
2868 | } | |
2869 | } else { | |
2870 | if (IS_ZEBRA_DEBUG_RIB) { | |
fd36be7e | 2871 | if (nh) |
d62a17ae | 2872 | rnode_debug( |
2873 | rn, vrf_id, | |
2874 | "via %s ifindex %d type %d " | |
2875 | "doesn't exist in rib", | |
36228974 | 2876 | inet_ntop(afi2family(afi), |
2877 | &nh->gate, buf2, | |
2878 | sizeof(buf2)), | |
2879 | nh->ifindex, type); | |
d62a17ae | 2880 | else |
2881 | rnode_debug( | |
2882 | rn, vrf_id, | |
fd36be7e DL |
2883 | "type %d doesn't exist in rib", |
2884 | type); | |
d62a17ae | 2885 | } |
2886 | route_unlock_node(rn); | |
2887 | return; | |
2888 | } | |
2889 | } | |
718e3744 | 2890 | |
5dfeba19 | 2891 | if (same) { |
996c9314 LB |
2892 | if (fromkernel && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE) |
2893 | && !allow_delete) { | |
5dfeba19 DS |
2894 | rib_install_kernel(rn, same, NULL); |
2895 | route_unlock_node(rn); | |
2896 | ||
2897 | return; | |
2898 | } | |
6134fd82 | 2899 | |
2b83602b | 2900 | /* Special handling for IPv4 or IPv6 routes sourced from |
2901 | * EVPN - the nexthop (and associated MAC) need to be | |
2902 | * uninstalled if no more refs. | |
2903 | */ | |
90264d64 | 2904 | if (CHECK_FLAG(flags, ZEBRA_FLAG_EVPN_ROUTE)) { |
6134fd82 | 2905 | struct nexthop *tmp_nh; |
2906 | ||
6b468511 | 2907 | for (ALL_NEXTHOPS_PTR(re->ng, tmp_nh)) { |
6134fd82 | 2908 | struct ipaddr vtep_ip; |
2909 | ||
2910 | memset(&vtep_ip, 0, sizeof(struct ipaddr)); | |
1ec31309 | 2911 | if (afi == AFI_IP) { |
2912 | vtep_ip.ipa_type = IPADDR_V4; | |
2913 | memcpy(&(vtep_ip.ipaddr_v4), | |
2914 | &(tmp_nh->gate.ipv4), | |
2915 | sizeof(struct in_addr)); | |
2916 | } else { | |
2917 | vtep_ip.ipa_type = IPADDR_V6; | |
2918 | memcpy(&(vtep_ip.ipaddr_v6), | |
2919 | &(tmp_nh->gate.ipv6), | |
2920 | sizeof(struct in6_addr)); | |
2921 | } | |
a317a9b9 | 2922 | zebra_vxlan_evpn_vrf_route_del(re->vrf_id, |
6134fd82 | 2923 | &vtep_ip, p); |
2924 | } | |
2925 | } | |
cf363e1b MS |
2926 | |
2927 | /* Notify dplane if system route changes */ | |
2928 | if (RIB_SYSTEM_ROUTE(re)) | |
2929 | dplane_sys_route_del(rn, same); | |
2930 | ||
d62a17ae | 2931 | rib_delnode(rn, same); |
5dfeba19 | 2932 | } |
05737783 | 2933 | |
d62a17ae | 2934 | route_unlock_node(rn); |
2935 | return; | |
2936 | } | |
718e3744 | 2937 | |
718e3744 | 2938 | |
d7c0a89a QY |
2939 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, |
2940 | unsigned short instance, int flags, struct prefix *p, | |
2941 | struct prefix_ipv6 *src_p, const struct nexthop *nh, | |
2942 | uint32_t table_id, uint32_t metric, uint32_t mtu, uint8_t distance, | |
2943 | route_tag_t tag) | |
d62a17ae | 2944 | { |
2945 | struct route_entry *re; | |
66af6845 | 2946 | struct nexthop *nexthop; |
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; |
6b468511 | 2960 | re->ng = nexthop_group_new(); |
d62a17ae | 2961 | |
66af6845 RW |
2962 | /* Add nexthop. */ |
2963 | nexthop = nexthop_new(); | |
2964 | *nexthop = *nh; | |
2965 | route_entry_nexthop_add(re, nexthop); | |
718e3744 | 2966 | |
66af6845 | 2967 | return rib_add_multipath(afi, safi, p, src_p, re); |
718e3744 | 2968 | } |
2969 | ||
2a99ab95 SW |
2970 | static const char *rib_update_event2str(rib_update_event_t event) |
2971 | { | |
2972 | const char *ret = "UNKNOWN"; | |
2973 | ||
2974 | switch (event) { | |
2975 | case RIB_UPDATE_KERNEL: | |
2976 | ret = "RIB_UPDATE_KERNEL"; | |
2977 | break; | |
2978 | case RIB_UPDATE_RMAP_CHANGE: | |
2979 | ret = "RIB_UPDATE_RMAP_CHANGE"; | |
2980 | break; | |
2981 | case RIB_UPDATE_OTHER: | |
2982 | ret = "RIB_UPDATE_OTHER"; | |
2983 | break; | |
2984 | case RIB_UPDATE_MAX: | |
2985 | break; | |
2986 | } | |
2987 | ||
2988 | return ret; | |
2989 | } | |
2990 | ||
2991 | ||
2992 | /* Schedule route nodes to be processed if they match the type */ | |
2993 | static void rib_update_route_node(struct route_node *rn, int type) | |
2994 | { | |
2995 | struct route_entry *re, *next; | |
2996 | bool re_changed = false; | |
2997 | ||
2998 | RNODE_FOREACH_RE_SAFE (rn, re, next) { | |
2999 | if (type == ZEBRA_ROUTE_ALL || type == re->type) { | |
3000 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
3001 | re_changed = true; | |
3002 | } | |
3003 | } | |
3004 | ||
3005 | if (re_changed) | |
3006 | rib_queue_add(rn); | |
3007 | } | |
3008 | ||
1c848137 | 3009 | /* Schedule routes of a particular table (address-family) based on event. */ |
d5b8c216 | 3010 | void rib_update_table(struct route_table *table, rib_update_event_t event) |
d62a17ae | 3011 | { |
3012 | struct route_node *rn; | |
2a99ab95 SW |
3013 | |
3014 | if (IS_ZEBRA_DEBUG_EVENT) { | |
3015 | struct zebra_vrf *zvrf; | |
3016 | struct vrf *vrf; | |
3017 | ||
3018 | zvrf = table->info ? ((rib_table_info_t *)table->info)->zvrf | |
3019 | : NULL; | |
3020 | vrf = zvrf ? zvrf->vrf : NULL; | |
3021 | ||
3022 | zlog_debug("%s: %s VRF %s Table %u event %s", __func__, | |
3023 | table->info ? afi2str( | |
3024 | ((rib_table_info_t *)table->info)->afi) | |
3025 | : "Unknown", | |
3026 | vrf ? vrf->name : "Unknown", | |
3027 | zvrf ? zvrf->table_id : 0, | |
3028 | rib_update_event2str(event)); | |
3029 | } | |
d62a17ae | 3030 | |
3031 | /* Walk all routes and queue for processing, if appropriate for | |
3032 | * the trigger event. | |
3033 | */ | |
3034 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
1ca60f2c DS |
3035 | /* |
3036 | * If we are looking at a route node and the node | |
3037 | * has already been queued we don't | |
3038 | * need to queue it up again | |
3039 | */ | |
2a99ab95 SW |
3040 | if (rn->info |
3041 | && CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
3042 | RIB_ROUTE_ANY_QUEUED)) | |
1ca60f2c | 3043 | continue; |
2a99ab95 | 3044 | |
d62a17ae | 3045 | switch (event) { |
2a99ab95 SW |
3046 | case RIB_UPDATE_KERNEL: |
3047 | rib_update_route_node(rn, ZEBRA_ROUTE_KERNEL); | |
3048 | break; | |
d62a17ae | 3049 | case RIB_UPDATE_RMAP_CHANGE: |
3050 | case RIB_UPDATE_OTHER: | |
2a99ab95 | 3051 | rib_update_route_node(rn, ZEBRA_ROUTE_ALL); |
d62a17ae | 3052 | break; |
d62a17ae | 3053 | default: |
3054 | break; | |
3055 | } | |
3056 | } | |
b84c7253 | 3057 | } |
3058 | ||
2a99ab95 | 3059 | static void rib_update_handle_vrf(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 3060 | { |
d62a17ae | 3061 | struct route_table *table; |
1c848137 | 3062 | |
2a99ab95 SW |
3063 | if (IS_ZEBRA_DEBUG_EVENT) |
3064 | zlog_debug("%s: Handling VRF %s event %s", __func__, | |
3065 | vrf_id_to_name(vrf_id), rib_update_event2str(event)); | |
3066 | ||
d62a17ae | 3067 | /* Process routes of interested address-families. */ |
3068 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2a99ab95 | 3069 | if (table) |
d62a17ae | 3070 | rib_update_table(table, event); |
718e3744 | 3071 | |
d62a17ae | 3072 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
2a99ab95 | 3073 | if (table) |
d62a17ae | 3074 | rib_update_table(table, event); |
2a99ab95 SW |
3075 | } |
3076 | ||
3077 | static void rib_update_handle_vrf_all(rib_update_event_t event) | |
3078 | { | |
3079 | struct zebra_router_table *zrt; | |
3080 | ||
3081 | if (IS_ZEBRA_DEBUG_EVENT) | |
3082 | zlog_debug("%s: Handling VRF (ALL) event %s", __func__, | |
3083 | rib_update_event2str(event)); | |
3084 | ||
3085 | /* Just iterate over all the route tables, rather than vrf lookups */ | |
3086 | RB_FOREACH (zrt, zebra_router_table_head, &zrouter.tables) | |
3087 | rib_update_table(zrt->table, event); | |
3088 | } | |
3089 | ||
3090 | struct rib_update_ctx { | |
3091 | rib_update_event_t event; | |
3092 | bool vrf_all; | |
3093 | vrf_id_t vrf_id; | |
3094 | }; | |
3095 | ||
3096 | static struct rib_update_ctx *rib_update_ctx_init(vrf_id_t vrf_id, | |
3097 | rib_update_event_t event) | |
3098 | { | |
3099 | struct rib_update_ctx *ctx; | |
3100 | ||
3101 | ctx = XCALLOC(MTYPE_RIB_UPDATE_CTX, sizeof(struct rib_update_ctx)); | |
3102 | ||
3103 | ctx->event = event; | |
3104 | ctx->vrf_id = vrf_id; | |
3105 | ||
3106 | return ctx; | |
3107 | } | |
3108 | ||
3109 | static void rib_update_ctx_fini(struct rib_update_ctx **ctx) | |
3110 | { | |
3111 | XFREE(MTYPE_RIB_UPDATE_CTX, *ctx); | |
3112 | ||
3113 | *ctx = NULL; | |
3114 | } | |
3115 | ||
3116 | static int rib_update_handler(struct thread *thread) | |
3117 | { | |
3118 | struct rib_update_ctx *ctx; | |
3119 | ||
3120 | ctx = THREAD_ARG(thread); | |
3121 | ||
3122 | if (ctx->vrf_all) | |
3123 | rib_update_handle_vrf_all(ctx->event); | |
3124 | else | |
3125 | rib_update_handle_vrf(ctx->vrf_id, ctx->event); | |
3126 | ||
3127 | rib_update_ctx_fini(&ctx); | |
3128 | ||
3129 | return 0; | |
3130 | } | |
3131 | ||
3132 | /* | |
3133 | * Thread list to ensure we don't schedule a ton of events | |
3134 | * if interfaces are flapping for instance. | |
3135 | */ | |
3136 | static struct thread *t_rib_update_threads[RIB_UPDATE_MAX]; | |
3137 | ||
3138 | /* Schedule a RIB update event for specific vrf */ | |
3139 | void rib_update_vrf(vrf_id_t vrf_id, rib_update_event_t event) | |
3140 | { | |
3141 | struct rib_update_ctx *ctx; | |
3142 | ||
3143 | ctx = rib_update_ctx_init(vrf_id, event); | |
3144 | ||
3145 | /* Don't worry about making sure multiple rib updates for specific vrf | |
3146 | * are scheduled at once for now. If it becomes a problem, we can use a | |
3147 | * lookup of some sort to keep track of running threads via t_vrf_id | |
3148 | * like how we are doing it in t_rib_update_threads[]. | |
3149 | */ | |
3150 | thread_add_event(zrouter.master, rib_update_handler, ctx, 0, NULL); | |
3151 | ||
3152 | if (IS_ZEBRA_DEBUG_EVENT) | |
3153 | zlog_debug("%s: Scheduled VRF %s, event %s", __func__, | |
3154 | vrf_id_to_name(ctx->vrf_id), | |
3155 | rib_update_event2str(event)); | |
3156 | } | |
3157 | ||
3158 | /* Schedule a RIB update event for all vrfs */ | |
3159 | void rib_update(rib_update_event_t event) | |
3160 | { | |
3161 | struct rib_update_ctx *ctx; | |
3162 | ||
3163 | ctx = rib_update_ctx_init(0, event); | |
3164 | ||
3165 | ctx->vrf_all = true; | |
3166 | ||
3167 | if (!thread_add_event(zrouter.master, rib_update_handler, ctx, 0, | |
3168 | &t_rib_update_threads[event])) | |
3169 | rib_update_ctx_fini(&ctx); /* Already scheduled */ | |
3170 | else if (IS_ZEBRA_DEBUG_EVENT) | |
3171 | zlog_debug("%s: Schedued VRF (ALL), event %s", __func__, | |
3172 | rib_update_event2str(event)); | |
718e3744 | 3173 | } |
3174 | ||
718e3744 | 3175 | /* Delete self installed routes after zebra is relaunched. */ |
95a29032 | 3176 | void rib_sweep_table(struct route_table *table) |
d62a17ae | 3177 | { |
3178 | struct route_node *rn; | |
3179 | struct route_entry *re; | |
3180 | struct route_entry *next; | |
915902cb | 3181 | struct nexthop *nexthop; |
d62a17ae | 3182 | |
915902cb DS |
3183 | if (!table) |
3184 | return; | |
d62a17ae | 3185 | |
915902cb | 3186 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
a2addae8 | 3187 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d4644d41 | 3188 | |
915902cb DS |
3189 | if (IS_ZEBRA_DEBUG_RIB) |
3190 | route_entry_dump(&rn->p, NULL, re); | |
3191 | ||
3192 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
3193 | continue; | |
3194 | ||
3195 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
3196 | continue; | |
3197 | ||
d4644d41 DS |
3198 | /* |
3199 | * If routes are older than startup_time then | |
3200 | * we know we read them in from the kernel. | |
3201 | * As such we can safely remove them. | |
3202 | */ | |
3203 | if (zrouter.startup_time < re->uptime) | |
3204 | continue; | |
3205 | ||
915902cb DS |
3206 | /* |
3207 | * So we are starting up and have received | |
3208 | * routes from the kernel that we have installed | |
3209 | * from a previous run of zebra but not cleaned | |
3210 | * up ( say a kill -9 ) | |
3211 | * But since we haven't actually installed | |
3212 | * them yet( we received them from the kernel ) | |
3213 | * we don't think they are active. | |
3214 | * So let's pretend they are active to actually | |
3215 | * remove them. | |
3216 | * In all honesty I'm not sure if we should | |
3217 | * mark them as active when we receive them | |
3218 | * This is startup only so probably ok. | |
3219 | * | |
3220 | * If we ever decide to move rib_sweep_table | |
3221 | * to a different spot (ie startup ) | |
3222 | * this decision needs to be revisited | |
3223 | */ | |
677c1dd5 | 3224 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
6b468511 | 3225 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
915902cb DS |
3226 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
3227 | ||
0c555cc6 DS |
3228 | rib_uninstall_kernel(rn, re); |
3229 | rib_delnode(rn, re); | |
915902cb DS |
3230 | } |
3231 | } | |
718e3744 | 3232 | } |
3233 | ||
3234 | /* Sweep all RIB tables. */ | |
d4644d41 | 3235 | int rib_sweep_route(struct thread *t) |
718e3744 | 3236 | { |
d62a17ae | 3237 | struct vrf *vrf; |
3238 | struct zebra_vrf *zvrf; | |
78104b9b | 3239 | |
a2addae8 | 3240 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
915902cb DS |
3241 | if ((zvrf = vrf->info) == NULL) |
3242 | continue; | |
3243 | ||
d62a17ae | 3244 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); |
3245 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3246 | } | |
95a29032 | 3247 | |
89272910 | 3248 | zebra_router_sweep_route(); |
d4644d41 DS |
3249 | |
3250 | return 0; | |
718e3744 | 3251 | } |
2ea1ab1c VT |
3252 | |
3253 | /* Remove specific by protocol routes from 'table'. */ | |
d7c0a89a | 3254 | unsigned long rib_score_proto_table(uint8_t proto, unsigned short instance, |
47a08aa9 | 3255 | struct route_table *table) |
d62a17ae | 3256 | { |
3257 | struct route_node *rn; | |
3258 | struct route_entry *re; | |
3259 | struct route_entry *next; | |
3260 | unsigned long n = 0; | |
3261 | ||
3262 | if (table) | |
3263 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
a2addae8 | 3264 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 3265 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
3266 | continue; | |
3267 | if (re->type == proto | |
3268 | && re->instance == instance) { | |
3269 | rib_delnode(rn, re); | |
3270 | n++; | |
3271 | } | |
3272 | } | |
3273 | return n; | |
2ea1ab1c VT |
3274 | } |
3275 | ||
3276 | /* Remove specific by protocol routes. */ | |
d7c0a89a | 3277 | unsigned long rib_score_proto(uint8_t proto, unsigned short instance) |
2ea1ab1c | 3278 | { |
d62a17ae | 3279 | struct vrf *vrf; |
3280 | struct zebra_vrf *zvrf; | |
d8612e65 | 3281 | struct other_route_table *ort; |
d62a17ae | 3282 | unsigned long cnt = 0; |
78104b9b | 3283 | |
d8612e65 DS |
3284 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
3285 | zvrf = vrf->info; | |
3286 | if (!zvrf) | |
3287 | continue; | |
78104b9b | 3288 | |
d8612e65 DS |
3289 | cnt += rib_score_proto_table(proto, instance, |
3290 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
3291 | + rib_score_proto_table( | |
3292 | proto, instance, | |
3293 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3294 | ||
81fddbe7 | 3295 | frr_each(otable, &zvrf->other_tables, ort) cnt += |
d8612e65 DS |
3296 | rib_score_proto_table(proto, instance, ort->table); |
3297 | } | |
47a08aa9 | 3298 | |
d62a17ae | 3299 | return cnt; |
2ea1ab1c VT |
3300 | } |
3301 | ||
718e3744 | 3302 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 3303 | void rib_close_table(struct route_table *table) |
718e3744 | 3304 | { |
d62a17ae | 3305 | struct route_node *rn; |
1e9f448f | 3306 | rib_table_info_t *info; |
5f7a4718 | 3307 | rib_dest_t *dest; |
718e3744 | 3308 | |
1e9f448f DS |
3309 | if (!table) |
3310 | return; | |
9fd92e3c | 3311 | |
6ca30e9e | 3312 | info = route_table_get_info(table); |
5adc2528 | 3313 | |
5f7a4718 DS |
3314 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
3315 | dest = rib_dest_from_rnode(rn); | |
1e9f448f | 3316 | |
5f7a4718 | 3317 | if (dest && dest->selected_fib) { |
1e9f448f DS |
3318 | if (info->safi == SAFI_UNICAST) |
3319 | hook_call(rib_update, rn, NULL); | |
3320 | ||
3cdba47a DS |
3321 | rib_uninstall_kernel(rn, dest->selected_fib); |
3322 | dest->selected_fib = NULL; | |
1e9f448f | 3323 | } |
5f7a4718 | 3324 | } |
718e3744 | 3325 | } |
3326 | ||
9bd9717b MS |
3327 | /* |
3328 | * Handler for async dataplane results after a pseudowire installation | |
3329 | */ | |
3330 | static int handle_pw_result(struct zebra_dplane_ctx *ctx) | |
3331 | { | |
9bd9717b MS |
3332 | struct zebra_pw *pw; |
3333 | struct zebra_vrf *vrf; | |
3334 | ||
3335 | /* The pseudowire code assumes success - we act on an error | |
3336 | * result for installation attempts here. | |
3337 | */ | |
3338 | if (dplane_ctx_get_op(ctx) != DPLANE_OP_PW_INSTALL) | |
3339 | goto done; | |
3340 | ||
3341 | if (dplane_ctx_get_status(ctx) != ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
3342 | vrf = zebra_vrf_lookup_by_id(dplane_ctx_get_vrf(ctx)); | |
7c7ef4a8 | 3343 | pw = zebra_pw_find(vrf, dplane_ctx_get_ifname(ctx)); |
9bd9717b MS |
3344 | if (pw) |
3345 | zebra_pw_install_failure(pw); | |
3346 | } | |
3347 | ||
3348 | done: | |
3349 | ||
9f2d0354 | 3350 | return 0; |
9bd9717b MS |
3351 | } |
3352 | ||
3353 | ||
7cdb1a84 | 3354 | /* |
d37f4d6c MS |
3355 | * Handle results from the dataplane system. Dequeue update context |
3356 | * structs, dispatch to appropriate internal handlers. | |
7cdb1a84 MS |
3357 | */ |
3358 | static int rib_process_dplane_results(struct thread *thread) | |
3359 | { | |
25779064 | 3360 | struct zebra_dplane_ctx *ctx; |
4c206c8f | 3361 | struct dplane_ctx_q ctxlist; |
2fc69f03 | 3362 | bool shut_p = false; |
4c206c8f MS |
3363 | |
3364 | /* Dequeue a list of completed updates with one lock/unlock cycle */ | |
7cdb1a84 MS |
3365 | |
3366 | do { | |
4c206c8f MS |
3367 | TAILQ_INIT(&ctxlist); |
3368 | ||
7cdb1a84 | 3369 | /* Take lock controlling queue of results */ |
00dffa8c | 3370 | frr_with_mutex(&dplane_mutex) { |
d37f4d6c | 3371 | /* Dequeue list of context structs */ |
4c206c8f | 3372 | dplane_ctx_list_append(&ctxlist, &rib_dplane_q); |
7cdb1a84 | 3373 | } |
7cdb1a84 | 3374 | |
4c206c8f MS |
3375 | /* Dequeue context block */ |
3376 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3377 | ||
3378 | /* If we've emptied the results queue, we're done */ | |
3379 | if (ctx == NULL) | |
7cdb1a84 | 3380 | break; |
7cdb1a84 | 3381 | |
2fc69f03 MS |
3382 | /* If zebra is shutting down, avoid processing results, |
3383 | * just drain the results queue. | |
3384 | */ | |
3385 | shut_p = atomic_load_explicit(&zrouter.in_shutdown, | |
3386 | memory_order_relaxed); | |
3387 | if (shut_p) { | |
3388 | while (ctx) { | |
3389 | dplane_ctx_fini(&ctx); | |
3390 | ||
3391 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3392 | } | |
3393 | ||
3394 | continue; | |
3395 | } | |
3396 | ||
4c206c8f | 3397 | while (ctx) { |
d37f4d6c MS |
3398 | switch (dplane_ctx_get_op(ctx)) { |
3399 | case DPLANE_OP_ROUTE_INSTALL: | |
3400 | case DPLANE_OP_ROUTE_UPDATE: | |
3401 | case DPLANE_OP_ROUTE_DELETE: | |
188a00e0 MS |
3402 | { |
3403 | /* Bit of special case for route updates | |
3404 | * that were generated by async notifications: | |
3405 | * we don't want to continue processing these | |
3406 | * in the rib. | |
3407 | */ | |
3408 | if (dplane_ctx_get_notif_provider(ctx) == 0) | |
3409 | rib_process_result(ctx); | |
3410 | else | |
3411 | dplane_ctx_fini(&ctx); | |
3412 | } | |
3413 | break; | |
d37f4d6c | 3414 | |
54818e3b MS |
3415 | case DPLANE_OP_ROUTE_NOTIFY: |
3416 | rib_process_dplane_notify(ctx); | |
d37f4d6c MS |
3417 | break; |
3418 | ||
3419 | case DPLANE_OP_LSP_INSTALL: | |
3420 | case DPLANE_OP_LSP_UPDATE: | |
3421 | case DPLANE_OP_LSP_DELETE: | |
188a00e0 MS |
3422 | { |
3423 | /* Bit of special case for LSP updates | |
3424 | * that were generated by async notifications: | |
3425 | * we don't want to continue processing these. | |
3426 | */ | |
3427 | if (dplane_ctx_get_notif_provider(ctx) == 0) | |
3428 | zebra_mpls_lsp_dplane_result(ctx); | |
3429 | else | |
3430 | dplane_ctx_fini(&ctx); | |
3431 | } | |
3432 | break; | |
d37f4d6c | 3433 | |
104e3ad9 MS |
3434 | case DPLANE_OP_LSP_NOTIFY: |
3435 | zebra_mpls_process_dplane_notify(ctx); | |
d37f4d6c MS |
3436 | break; |
3437 | ||
9bd9717b MS |
3438 | case DPLANE_OP_PW_INSTALL: |
3439 | case DPLANE_OP_PW_UNINSTALL: | |
3440 | handle_pw_result(ctx); | |
3441 | break; | |
3442 | ||
cf363e1b MS |
3443 | case DPLANE_OP_SYS_ROUTE_ADD: |
3444 | case DPLANE_OP_SYS_ROUTE_DELETE: | |
3445 | /* No further processing in zebra for these. */ | |
3446 | dplane_ctx_fini(&ctx); | |
3447 | break; | |
3448 | ||
036d93c0 MS |
3449 | case DPLANE_OP_MAC_INSTALL: |
3450 | case DPLANE_OP_MAC_DELETE: | |
3451 | zebra_vxlan_handle_result(ctx); | |
3452 | break; | |
3453 | ||
931fa60c MS |
3454 | /* Some op codes not handled here */ |
3455 | case DPLANE_OP_ADDR_INSTALL: | |
3456 | case DPLANE_OP_ADDR_UNINSTALL: | |
3457 | case DPLANE_OP_NEIGH_INSTALL: | |
3458 | case DPLANE_OP_NEIGH_UPDATE: | |
3459 | case DPLANE_OP_NEIGH_DELETE: | |
0bbd4ff4 MS |
3460 | case DPLANE_OP_VTEP_ADD: |
3461 | case DPLANE_OP_VTEP_DELETE: | |
931fa60c | 3462 | case DPLANE_OP_NONE: |
d37f4d6c MS |
3463 | /* Don't expect this: just return the struct? */ |
3464 | dplane_ctx_fini(&ctx); | |
3465 | break; | |
931fa60c | 3466 | |
d37f4d6c | 3467 | } /* Dispatch by op code */ |
4c206c8f MS |
3468 | |
3469 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3470 | } | |
3471 | ||
5709131c | 3472 | } while (1); |
7cdb1a84 | 3473 | |
5709131c | 3474 | return 0; |
7cdb1a84 MS |
3475 | } |
3476 | ||
3477 | /* | |
3478 | * Results are returned from the dataplane subsystem, in the context of | |
1bcea841 | 3479 | * the dataplane pthread. We enqueue the results here for processing by |
7cdb1a84 MS |
3480 | * the main thread later. |
3481 | */ | |
4c206c8f | 3482 | static int rib_dplane_results(struct dplane_ctx_q *ctxlist) |
7cdb1a84 MS |
3483 | { |
3484 | /* Take lock controlling queue of results */ | |
00dffa8c | 3485 | frr_with_mutex(&dplane_mutex) { |
4c206c8f MS |
3486 | /* Enqueue context blocks */ |
3487 | dplane_ctx_list_append(&rib_dplane_q, ctxlist); | |
7cdb1a84 | 3488 | } |
7cdb1a84 | 3489 | |
4c206c8f | 3490 | /* Ensure event is signalled to zebra main pthread */ |
3801e764 | 3491 | thread_add_event(zrouter.master, rib_process_dplane_results, NULL, 0, |
7cdb1a84 MS |
3492 | &t_dplane); |
3493 | ||
5709131c | 3494 | return 0; |
7cdb1a84 MS |
3495 | } |
3496 | ||
eaa2716d SW |
3497 | /* |
3498 | * Ensure there are no empty slots in the route_info array. | |
3499 | * Every route type in zebra should be present there. | |
3500 | */ | |
3501 | static void check_route_info(void) | |
3502 | { | |
3503 | int len = array_size(route_info); | |
3504 | ||
3505 | /* | |
3506 | * ZEBRA_ROUTE_SYSTEM is special cased since | |
3507 | * its key is 0 anyway. | |
3508 | * | |
3509 | * ZEBRA_ROUTE_ALL is also ignored. | |
3510 | */ | |
3511 | for (int i = 0; i < len; i++) { | |
3512 | if (i == ZEBRA_ROUTE_SYSTEM || i == ZEBRA_ROUTE_ALL) | |
3513 | continue; | |
3514 | assert(route_info[i].key); | |
8dc7a759 | 3515 | assert(route_info[i].meta_q_map < MQ_SIZE); |
eaa2716d SW |
3516 | } |
3517 | } | |
3518 | ||
718e3744 | 3519 | /* Routing information base initialize. */ |
d62a17ae | 3520 | void rib_init(void) |
718e3744 | 3521 | { |
eaa2716d SW |
3522 | check_route_info(); |
3523 | ||
2561d12e | 3524 | rib_queue_init(); |
7cdb1a84 MS |
3525 | |
3526 | /* Init dataplane, and register for results */ | |
3527 | pthread_mutex_init(&dplane_mutex, NULL); | |
3528 | TAILQ_INIT(&rib_dplane_q); | |
4c206c8f | 3529 | zebra_dplane_init(rib_dplane_results); |
718e3744 | 3530 | } |
0915bb0c AS |
3531 | |
3532 | /* | |
3533 | * vrf_id_get_next | |
3534 | * | |
3535 | * Get the first vrf id that is greater than the given vrf id if any. | |
3536 | * | |
2951a7a4 | 3537 | * Returns true if a vrf id was found, false otherwise. |
0915bb0c | 3538 | */ |
d62a17ae | 3539 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 3540 | { |
d62a17ae | 3541 | struct vrf *vrf; |
b72ede27 | 3542 | |
d62a17ae | 3543 | vrf = vrf_lookup_by_id(vrf_id); |
3544 | if (vrf) { | |
3545 | vrf = RB_NEXT(vrf_id_head, vrf); | |
3546 | if (vrf) { | |
3547 | *next_id_p = vrf->vrf_id; | |
3548 | return 1; | |
3549 | } | |
3550 | } | |
0915bb0c | 3551 | |
d62a17ae | 3552 | return 0; |
0915bb0c AS |
3553 | } |
3554 | ||
3555 | /* | |
3556 | * rib_tables_iter_next | |
3557 | * | |
3558 | * Returns the next table in the iteration. | |
3559 | */ | |
d62a17ae | 3560 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
3561 | { | |
3562 | struct route_table *table; | |
3563 | ||
3564 | /* | |
3565 | * Array that helps us go over all AFI/SAFI combinations via one | |
3566 | * index. | |
3567 | */ | |
3568 | static struct { | |
3569 | afi_t afi; | |
3570 | safi_t safi; | |
3571 | } afi_safis[] = { | |
3572 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
3573 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
3574 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
3575 | }; | |
3576 | ||
3577 | table = NULL; | |
3578 | ||
3579 | switch (iter->state) { | |
3580 | ||
3581 | case RIB_TABLES_ITER_S_INIT: | |
3582 | iter->vrf_id = VRF_DEFAULT; | |
3583 | iter->afi_safi_ix = -1; | |
3584 | ||
3585 | /* Fall through */ | |
3586 | ||
3587 | case RIB_TABLES_ITER_S_ITERATING: | |
3588 | iter->afi_safi_ix++; | |
3589 | while (1) { | |
3590 | ||
3591 | while (iter->afi_safi_ix | |
7e3a1ec7 | 3592 | < (int)array_size(afi_safis)) { |
d62a17ae | 3593 | table = zebra_vrf_table( |
3594 | afi_safis[iter->afi_safi_ix].afi, | |
3595 | afi_safis[iter->afi_safi_ix].safi, | |
3596 | iter->vrf_id); | |
3597 | if (table) | |
3598 | break; | |
3599 | ||
3600 | iter->afi_safi_ix++; | |
3601 | } | |
3602 | ||
3603 | /* | |
3604 | * Found another table in this vrf. | |
3605 | */ | |
3606 | if (table) | |
3607 | break; | |
3608 | ||
3609 | /* | |
3610 | * Done with all tables in the current vrf, go to the | |
3611 | * next | |
3612 | * one. | |
3613 | */ | |
3614 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
3615 | break; | |
3616 | ||
3617 | iter->afi_safi_ix = 0; | |
3618 | } | |
0915bb0c | 3619 | |
0915bb0c AS |
3620 | break; |
3621 | ||
d62a17ae | 3622 | case RIB_TABLES_ITER_S_DONE: |
3623 | return NULL; | |
0915bb0c AS |
3624 | } |
3625 | ||
d62a17ae | 3626 | if (table) |
3627 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
3628 | else | |
3629 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 3630 | |
d62a17ae | 3631 | return table; |
0915bb0c | 3632 | } |