]>
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); |
d62a17ae | 200 | re->nexthop_num++; |
718e3744 | 201 | } |
202 | ||
6e26278c | 203 | |
6e26278c DS |
204 | /** |
205 | * copy_nexthop - copy a nexthop to the rib structure. | |
206 | */ | |
d62a17ae | 207 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) |
6e26278c | 208 | { |
6b468511 DS |
209 | assert(!re->ng->nexthop); |
210 | copy_nexthops(&re->ng->nexthop, nh, NULL); | |
d62a17ae | 211 | for (struct nexthop *nexthop = nh; nexthop; nexthop = nexthop->next) |
212 | re->nexthop_num++; | |
6e26278c DS |
213 | } |
214 | ||
718e3744 | 215 | /* Delete specified nexthop from the list. */ |
d62a17ae | 216 | void route_entry_nexthop_delete(struct route_entry *re, struct nexthop *nexthop) |
718e3744 | 217 | { |
d62a17ae | 218 | if (nexthop->next) |
219 | nexthop->next->prev = nexthop->prev; | |
220 | if (nexthop->prev) | |
221 | nexthop->prev->next = nexthop->next; | |
222 | else | |
6b468511 | 223 | re->ng->nexthop = nexthop->next; |
d62a17ae | 224 | re->nexthop_num--; |
718e3744 | 225 | } |
226 | ||
fa713d9e | 227 | |
d62a17ae | 228 | struct nexthop *route_entry_nexthop_ifindex_add(struct route_entry *re, |
4a7371e9 DS |
229 | ifindex_t ifindex, |
230 | vrf_id_t nh_vrf_id) | |
718e3744 | 231 | { |
d62a17ae | 232 | struct nexthop *nexthop; |
718e3744 | 233 | |
d62a17ae | 234 | nexthop = nexthop_new(); |
235 | nexthop->type = NEXTHOP_TYPE_IFINDEX; | |
236 | nexthop->ifindex = ifindex; | |
4a7371e9 | 237 | nexthop->vrf_id = nh_vrf_id; |
718e3744 | 238 | |
d62a17ae | 239 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 240 | |
d62a17ae | 241 | return nexthop; |
718e3744 | 242 | } |
243 | ||
d62a17ae | 244 | struct nexthop *route_entry_nexthop_ipv4_add(struct route_entry *re, |
245 | struct in_addr *ipv4, | |
4a7371e9 DS |
246 | struct in_addr *src, |
247 | vrf_id_t nh_vrf_id) | |
718e3744 | 248 | { |
d62a17ae | 249 | struct nexthop *nexthop; |
718e3744 | 250 | |
d62a17ae | 251 | nexthop = nexthop_new(); |
252 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
4a7371e9 | 253 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 254 | nexthop->gate.ipv4 = *ipv4; |
255 | if (src) | |
256 | nexthop->src.ipv4 = *src; | |
718e3744 | 257 | |
d62a17ae | 258 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 259 | |
d62a17ae | 260 | return nexthop; |
718e3744 | 261 | } |
262 | ||
d62a17ae | 263 | struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re, |
264 | struct in_addr *ipv4, | |
265 | struct in_addr *src, | |
4a7371e9 DS |
266 | ifindex_t ifindex, |
267 | vrf_id_t nh_vrf_id) | |
718e3744 | 268 | { |
d62a17ae | 269 | struct nexthop *nexthop; |
270 | struct interface *ifp; | |
718e3744 | 271 | |
d62a17ae | 272 | nexthop = nexthop_new(); |
4a7371e9 | 273 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 274 | nexthop->type = NEXTHOP_TYPE_IPV4_IFINDEX; |
275 | nexthop->gate.ipv4 = *ipv4; | |
276 | if (src) | |
277 | nexthop->src.ipv4 = *src; | |
278 | nexthop->ifindex = ifindex; | |
4a7371e9 | 279 | ifp = if_lookup_by_index(nexthop->ifindex, nh_vrf_id); |
d62a17ae | 280 | /*Pending: need to think if null ifp here is ok during bootup? |
281 | There was a crash because ifp here was coming to be NULL */ | |
282 | if (ifp) | |
2b83602b | 283 | if (connected_is_unnumbered(ifp)) |
d62a17ae | 284 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); |
718e3744 | 285 | |
d62a17ae | 286 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 287 | |
d62a17ae | 288 | return nexthop; |
718e3744 | 289 | } |
290 | ||
d62a17ae | 291 | struct nexthop *route_entry_nexthop_ipv6_add(struct route_entry *re, |
4a7371e9 DS |
292 | struct in6_addr *ipv6, |
293 | vrf_id_t nh_vrf_id) | |
718e3744 | 294 | { |
d62a17ae | 295 | struct nexthop *nexthop; |
718e3744 | 296 | |
d62a17ae | 297 | nexthop = nexthop_new(); |
4a7371e9 | 298 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 299 | nexthop->type = NEXTHOP_TYPE_IPV6; |
300 | nexthop->gate.ipv6 = *ipv6; | |
718e3744 | 301 | |
d62a17ae | 302 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 303 | |
d62a17ae | 304 | return nexthop; |
718e3744 | 305 | } |
306 | ||
d62a17ae | 307 | struct nexthop *route_entry_nexthop_ipv6_ifindex_add(struct route_entry *re, |
308 | struct in6_addr *ipv6, | |
4a7371e9 DS |
309 | ifindex_t ifindex, |
310 | vrf_id_t nh_vrf_id) | |
718e3744 | 311 | { |
d62a17ae | 312 | struct nexthop *nexthop; |
718e3744 | 313 | |
d62a17ae | 314 | nexthop = nexthop_new(); |
4a7371e9 | 315 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 316 | nexthop->type = NEXTHOP_TYPE_IPV6_IFINDEX; |
317 | nexthop->gate.ipv6 = *ipv6; | |
318 | nexthop->ifindex = ifindex; | |
718e3744 | 319 | |
d62a17ae | 320 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 321 | |
d62a17ae | 322 | return nexthop; |
718e3744 | 323 | } |
718e3744 | 324 | |
a8309422 | 325 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re, |
60466a63 | 326 | enum blackhole_type bh_type) |
595db7f1 | 327 | { |
d62a17ae | 328 | struct nexthop *nexthop; |
595db7f1 | 329 | |
d62a17ae | 330 | nexthop = nexthop_new(); |
4a7371e9 | 331 | nexthop->vrf_id = VRF_DEFAULT; |
d62a17ae | 332 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; |
a8309422 | 333 | nexthop->bh_type = bh_type; |
595db7f1 | 334 | |
d62a17ae | 335 | route_entry_nexthop_add(re, nexthop); |
595db7f1 | 336 | |
d62a17ae | 337 | return nexthop; |
595db7f1 | 338 | } |
339 | ||
d62a17ae | 340 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
341 | union g_addr *addr, struct route_node **rn_out) | |
342 | { | |
343 | struct prefix p; | |
344 | struct route_table *table; | |
345 | struct route_node *rn; | |
5f7a4718 | 346 | struct route_entry *match = NULL; |
d62a17ae | 347 | |
348 | /* Lookup table. */ | |
349 | table = zebra_vrf_table(afi, safi, vrf_id); | |
350 | if (!table) | |
351 | return 0; | |
352 | ||
353 | memset(&p, 0, sizeof(struct prefix)); | |
354 | p.family = afi; | |
355 | if (afi == AFI_IP) { | |
356 | p.u.prefix4 = addr->ipv4; | |
357 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
358 | } else { | |
359 | p.u.prefix6 = addr->ipv6; | |
360 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
361 | } | |
718e3744 | 362 | |
d62a17ae | 363 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 364 | |
d62a17ae | 365 | while (rn) { |
5f7a4718 DS |
366 | rib_dest_t *dest; |
367 | ||
d62a17ae | 368 | route_unlock_node(rn); |
718e3744 | 369 | |
5f7a4718 | 370 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
371 | if (dest && dest->selected_fib |
372 | && !CHECK_FLAG(dest->selected_fib->status, | |
373 | ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 374 | match = dest->selected_fib; |
718e3744 | 375 | |
d62a17ae | 376 | /* If there is no selected route or matched route is EGP, go up |
377 | tree. */ | |
378 | if (!match) { | |
379 | do { | |
380 | rn = rn->parent; | |
381 | } while (rn && rn->info == NULL); | |
382 | if (rn) | |
383 | route_lock_node(rn); | |
384 | } else { | |
385 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
677c1dd5 DS |
386 | if (!CHECK_FLAG(match->status, |
387 | ROUTE_ENTRY_INSTALLED)) | |
d62a17ae | 388 | return NULL; |
389 | } | |
390 | ||
391 | if (rn_out) | |
392 | *rn_out = rn; | |
393 | return match; | |
394 | } | |
16814f96 | 395 | } |
d62a17ae | 396 | return NULL; |
397 | } | |
398 | ||
399 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
400 | struct in_addr addr, | |
401 | struct route_node **rn_out) | |
402 | { | |
403 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
404 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
405 | union g_addr gaddr = {.ipv4 = addr}; | |
406 | ||
526052fb | 407 | switch (zrouter.ipv4_multicast_mode) { |
d62a17ae | 408 | case MCAST_MRIB_ONLY: |
409 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
410 | rn_out); | |
411 | case MCAST_URIB_ONLY: | |
412 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
413 | case MCAST_NO_CONFIG: | |
414 | case MCAST_MIX_MRIB_FIRST: | |
415 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
416 | &m_rn); | |
417 | if (!mre) | |
418 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
419 | &gaddr, &u_rn); | |
420 | break; | |
421 | case MCAST_MIX_DISTANCE: | |
422 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
423 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
424 | if (mre && ure) | |
425 | re = ure->distance < mre->distance ? ure : mre; | |
426 | else if (mre) | |
427 | re = mre; | |
428 | else if (ure) | |
429 | re = ure; | |
430 | break; | |
431 | case MCAST_MIX_PFXLEN: | |
432 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
433 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
434 | if (mre && ure) | |
435 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
436 | else if (mre) | |
437 | re = mre; | |
438 | else if (ure) | |
439 | re = ure; | |
440 | break; | |
718e3744 | 441 | } |
33550aa8 | 442 | |
d62a17ae | 443 | if (rn_out) |
444 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 445 | |
d62a17ae | 446 | if (IS_ZEBRA_DEBUG_RIB) { |
447 | char buf[BUFSIZ]; | |
448 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 449 | |
32391aff DS |
450 | zlog_debug("%s: %s: vrf: %u found %s, using %s", |
451 | __func__, buf, vrf_id, | |
d62a17ae | 452 | mre ? (ure ? "MRIB+URIB" : "MRIB") |
453 | : ure ? "URIB" : "nothing", | |
454 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
455 | } | |
456 | return re; | |
4623d897 DL |
457 | } |
458 | ||
d62a17ae | 459 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 460 | { |
d62a17ae | 461 | struct route_table *table; |
462 | struct route_node *rn; | |
5f7a4718 | 463 | struct route_entry *match = NULL; |
5f7a4718 | 464 | rib_dest_t *dest; |
718e3744 | 465 | |
d62a17ae | 466 | /* Lookup table. */ |
467 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
468 | if (!table) | |
469 | return 0; | |
718e3744 | 470 | |
d62a17ae | 471 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 472 | |
d62a17ae | 473 | /* No route for this prefix. */ |
474 | if (!rn) | |
475 | return NULL; | |
718e3744 | 476 | |
d62a17ae | 477 | /* Unlock node. */ |
478 | route_unlock_node(rn); | |
5f7a4718 | 479 | dest = rib_dest_from_rnode(rn); |
718e3744 | 480 | |
996c9314 LB |
481 | if (dest && dest->selected_fib |
482 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 483 | match = dest->selected_fib; |
718e3744 | 484 | |
d62a17ae | 485 | if (!match) |
486 | return NULL; | |
718e3744 | 487 | |
d62a17ae | 488 | if (match->type == ZEBRA_ROUTE_CONNECT) |
489 | return match; | |
f9e1b38e | 490 | |
677c1dd5 DS |
491 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_INSTALLED)) |
492 | return match; | |
718e3744 | 493 | |
d62a17ae | 494 | return NULL; |
718e3744 | 495 | } |
496 | ||
a64448ba DS |
497 | /* |
498 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
499 | * (nexthops) must have a label. | |
500 | */ | |
d62a17ae | 501 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 502 | { |
d62a17ae | 503 | struct nexthop *nexthop = NULL; |
a64448ba | 504 | |
d62a17ae | 505 | if (re->type != ZEBRA_ROUTE_BGP) |
506 | return 0; | |
a64448ba | 507 | |
6b468511 | 508 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
d62a17ae | 509 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) |
510 | return 0; | |
6b0655a2 | 511 | |
d62a17ae | 512 | return 1; |
a64448ba | 513 | } |
718e3744 | 514 | |
6ae24471 DS |
515 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
516 | * is only used for IPv4. | |
517 | */ | |
0c555cc6 DS |
518 | void rib_install_kernel(struct route_node *rn, struct route_entry *re, |
519 | struct route_entry *old) | |
718e3744 | 520 | { |
d62a17ae | 521 | struct nexthop *nexthop; |
522 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 523 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
97f5b441 MS |
524 | const struct prefix *p, *src_p; |
525 | enum zebra_dplane_result ret; | |
526 | ||
527 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
718e3744 | 528 | |
d62a17ae | 529 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 530 | |
d62a17ae | 531 | if (info->safi != SAFI_UNICAST) { |
6b468511 | 532 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
d62a17ae | 533 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 534 | return; |
25b9cb0c DL |
535 | } else { |
536 | struct nexthop *prev; | |
537 | ||
6b468511 | 538 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) { |
25b9cb0c | 539 | UNSET_FLAG (nexthop->flags, NEXTHOP_FLAG_DUPLICATE); |
6b468511 | 540 | for (ALL_NEXTHOPS_PTR(re->ng, prev)) { |
25b9cb0c DL |
541 | if (prev == nexthop) |
542 | break; | |
996c9314 LB |
543 | if (nexthop_same_firsthop(nexthop, prev)) { |
544 | SET_FLAG(nexthop->flags, | |
545 | NEXTHOP_FLAG_DUPLICATE); | |
25b9cb0c DL |
546 | break; |
547 | } | |
548 | } | |
549 | } | |
d62a17ae | 550 | } |
718e3744 | 551 | |
2063a814 DS |
552 | /* |
553 | * If this is a replace to a new RE let the originator of the RE | |
554 | * know that they've lost | |
555 | */ | |
9a9f8926 | 556 | if (old && (old != re) && (old->type != re->type)) |
28610f7e | 557 | zsend_route_notify_owner(old, p, ZAPI_ROUTE_BETTER_ADMIN_WON); |
25b9cb0c | 558 | |
97f5b441 MS |
559 | /* Update fib selection */ |
560 | dest->selected_fib = re; | |
561 | ||
d62a17ae | 562 | /* |
563 | * Make sure we update the FPM any time we send new information to | |
564 | * the kernel. | |
565 | */ | |
566 | hook_call(rib_update, rn, "installing in kernel"); | |
97f5b441 MS |
567 | |
568 | /* Send add or update */ | |
8263d1d0 | 569 | if (old) |
97f5b441 | 570 | ret = dplane_route_update(rn, re, old); |
5709131c | 571 | else |
97f5b441 | 572 | ret = dplane_route_add(rn, re); |
97f5b441 MS |
573 | |
574 | switch (ret) { | |
ea1c14f6 | 575 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
60f98b23 | 576 | SET_FLAG(re->status, ROUTE_ENTRY_QUEUED); |
efe6c026 MS |
577 | |
578 | if (old) { | |
60f98b23 | 579 | SET_FLAG(old->status, ROUTE_ENTRY_QUEUED); |
efe6c026 MS |
580 | |
581 | /* Free old FIB nexthop group */ | |
582 | if (old->fib_ng.nexthop) { | |
583 | nexthops_free(old->fib_ng.nexthop); | |
584 | old->fib_ng.nexthop = NULL; | |
585 | } | |
586 | ||
587 | if (!RIB_SYSTEM_ROUTE(old)) { | |
588 | /* Clear old route's FIB flags */ | |
6b468511 | 589 | for (ALL_NEXTHOPS_PTR(old->ng, nexthop)) { |
efe6c026 MS |
590 | UNSET_FLAG(nexthop->flags, |
591 | NEXTHOP_FLAG_FIB); | |
592 | } | |
593 | } | |
594 | } | |
595 | ||
97f5b441 MS |
596 | if (zvrf) |
597 | zvrf->installs_queued++; | |
1e885672 | 598 | break; |
ea1c14f6 | 599 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
600 | { |
601 | char str[SRCDEST2STR_BUFFER]; | |
602 | ||
603 | srcdest_rnode2str(rn, str, sizeof(str)); | |
604 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
605 | "%u:%s: Failed to enqueue dataplane install", | |
606 | re->vrf_id, str); | |
1e885672 | 607 | break; |
97f5b441 | 608 | } |
ea1c14f6 | 609 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
97f5b441 MS |
610 | if (zvrf) |
611 | zvrf->installs++; | |
1e885672 DS |
612 | break; |
613 | } | |
d62a17ae | 614 | |
0c555cc6 | 615 | return; |
718e3744 | 616 | } |
617 | ||
618 | /* Uninstall the route from kernel. */ | |
0c555cc6 | 619 | void rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 620 | { |
d62a17ae | 621 | struct nexthop *nexthop; |
622 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 623 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
05737783 | 624 | |
d62a17ae | 625 | if (info->safi != SAFI_UNICAST) { |
677c1dd5 | 626 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
6b468511 | 627 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
d6792f9d | 628 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 629 | return; |
d62a17ae | 630 | } |
416ec78d | 631 | |
d62a17ae | 632 | /* |
633 | * Make sure we update the FPM any time we send new information to | |
97f5b441 | 634 | * the dataplane. |
d62a17ae | 635 | */ |
636 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
97f5b441 MS |
637 | |
638 | switch (dplane_route_delete(rn, re)) { | |
ea1c14f6 | 639 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
97f5b441 MS |
640 | if (zvrf) |
641 | zvrf->removals_queued++; | |
1e885672 | 642 | break; |
ea1c14f6 | 643 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
644 | { |
645 | char str[SRCDEST2STR_BUFFER]; | |
646 | ||
647 | srcdest_rnode2str(rn, str, sizeof(str)); | |
648 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
649 | "%u:%s: Failed to enqueue dataplane uninstall", | |
650 | re->vrf_id, str); | |
1e885672 | 651 | break; |
97f5b441 | 652 | } |
ea1c14f6 | 653 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
1e885672 DS |
654 | if (zvrf) |
655 | zvrf->removals++; | |
656 | break; | |
657 | } | |
718e3744 | 658 | |
0c555cc6 | 659 | return; |
718e3744 | 660 | } |
661 | ||
662 | /* Uninstall the route from kernel. */ | |
d62a17ae | 663 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 664 | { |
d62a17ae | 665 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
5f7a4718 | 666 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
97f5b441 | 667 | struct nexthop *nexthop; |
416ec78d | 668 | |
5f7a4718 | 669 | if (dest && dest->selected_fib == re) { |
d62a17ae | 670 | if (info->safi == SAFI_UNICAST) |
671 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 672 | |
d62a17ae | 673 | /* If labeled-unicast route, uninstall transit LSP. */ |
674 | if (zebra_rib_labeled_unicast(re)) | |
675 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
7cdb1a84 | 676 | |
3cdba47a | 677 | rib_uninstall_kernel(rn, re); |
97f5b441 MS |
678 | |
679 | dest->selected_fib = NULL; | |
680 | ||
efe6c026 MS |
681 | /* Free FIB nexthop group, if present */ |
682 | if (re->fib_ng.nexthop) { | |
683 | nexthops_free(re->fib_ng.nexthop); | |
684 | re->fib_ng.nexthop = NULL; | |
685 | } | |
686 | ||
6b468511 | 687 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
97f5b441 | 688 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
d62a17ae | 689 | } |
446bb95e | 690 | |
d62a17ae | 691 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
86391e56 MS |
692 | const struct prefix *p, *src_p; |
693 | ||
d62a17ae | 694 | srcdest_rnode_prefixes(rn, &p, &src_p); |
05737783 | 695 | |
40f321c0 | 696 | redistribute_delete(p, src_p, re, NULL); |
d62a17ae | 697 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); |
698 | } | |
718e3744 | 699 | } |
700 | ||
9fd92e3c AS |
701 | /* |
702 | * rib_can_delete_dest | |
703 | * | |
2951a7a4 | 704 | * Returns true if the given dest can be deleted from the table. |
9fd92e3c | 705 | */ |
d62a17ae | 706 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 707 | { |
045207e2 | 708 | if (re_list_first(&dest->routes)) { |
d62a17ae | 709 | return 0; |
710 | } | |
9fd92e3c | 711 | |
c86ba6c2 DS |
712 | /* |
713 | * Unresolved rnh's are stored on the default route's list | |
714 | * | |
715 | * dest->rnode can also be the source prefix node in an | |
716 | * ipv6 sourcedest table. Fortunately the prefix of a | |
717 | * source prefix node can never be the default prefix. | |
718 | */ | |
719 | if (is_default_prefix(&dest->rnode->p)) | |
720 | return 0; | |
721 | ||
d62a17ae | 722 | /* |
723 | * Don't delete the dest if we have to update the FPM about this | |
724 | * prefix. | |
725 | */ | |
726 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
727 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
728 | return 0; | |
5adc2528 | 729 | |
d62a17ae | 730 | return 1; |
9fd92e3c AS |
731 | } |
732 | ||
699dae23 DS |
733 | void zebra_rib_evaluate_rn_nexthops(struct route_node *rn, uint32_t seq) |
734 | { | |
735 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
699dae23 DS |
736 | struct rnh *rnh; |
737 | ||
738 | /* | |
50872b08 | 739 | * We are storing the rnh's associated withb |
699dae23 DS |
740 | * the tracked nexthop as a list of the rn's. |
741 | * Unresolved rnh's are placed at the top | |
742 | * of the tree list.( 0.0.0.0/0 for v4 and 0::0/0 for v6 ) | |
743 | * As such for each rn we need to walk up the tree | |
744 | * and see if any rnh's need to see if they | |
745 | * would match a more specific route | |
746 | */ | |
747 | while (rn) { | |
50872b08 DS |
748 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
749 | char buf[PREFIX_STRLEN]; | |
750 | ||
0a7be328 | 751 | zlog_debug("%s: %s Being examined for Nexthop Tracking Count: %zd", |
50872b08 | 752 | __PRETTY_FUNCTION__, |
0a7be328 DS |
753 | srcdest_rnode2str(rn, buf, sizeof(buf)), |
754 | dest ? rnh_list_count(&dest->nht) : 0); | |
50872b08 | 755 | } |
699dae23 DS |
756 | if (!dest) { |
757 | rn = rn->parent; | |
758 | if (rn) | |
759 | dest = rib_dest_from_rnode(rn); | |
760 | continue; | |
761 | } | |
762 | /* | |
763 | * If we have any rnh's stored in the nht list | |
764 | * then we know that this route node was used for | |
765 | * nht resolution and as such we need to call the | |
766 | * nexthop tracking evaluation code | |
767 | */ | |
739c9c90 | 768 | frr_each_safe(rnh_list, &dest->nht, rnh) { |
699dae23 DS |
769 | struct zebra_vrf *zvrf = |
770 | zebra_vrf_lookup_by_id(rnh->vrf_id); | |
771 | struct prefix *p = &rnh->node->p; | |
772 | ||
50872b08 | 773 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
699dae23 DS |
774 | char buf1[PREFIX_STRLEN]; |
775 | char buf2[PREFIX_STRLEN]; | |
776 | ||
0a7be328 | 777 | zlog_debug("%u:%s has Nexthop(%s) Type: %s depending on it, evaluating %u:%u", |
699dae23 | 778 | zvrf->vrf->vrf_id, |
50872b08 | 779 | srcdest_rnode2str(rn, buf1, |
699dae23 DS |
780 | sizeof(buf1)), |
781 | prefix2str(p, buf2, sizeof(buf2)), | |
0a7be328 | 782 | rnh_type2str(rnh->type), |
699dae23 DS |
783 | seq, rnh->seqno); |
784 | } | |
785 | ||
786 | /* | |
787 | * If we have evaluated this node on this pass | |
788 | * already, due to following the tree up | |
789 | * then we know that we can move onto the next | |
790 | * rnh to process. | |
791 | * | |
792 | * Additionally we call zebra_evaluate_rnh | |
793 | * when we gc the dest. In this case we know | |
794 | * that there must be no other re's where | |
795 | * we were originally as such we know that | |
796 | * that sequence number is ok to respect. | |
797 | */ | |
50872b08 DS |
798 | if (rnh->seqno == seq) { |
799 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) | |
800 | zlog_debug( | |
801 | "\tNode processed and moved already"); | |
699dae23 | 802 | continue; |
50872b08 | 803 | } |
699dae23 DS |
804 | |
805 | rnh->seqno = seq; | |
806 | zebra_evaluate_rnh(zvrf, family2afi(p->family), 0, | |
807 | rnh->type, p); | |
808 | } | |
809 | ||
810 | rn = rn->parent; | |
811 | if (rn) | |
812 | dest = rib_dest_from_rnode(rn); | |
813 | } | |
814 | } | |
815 | ||
9fd92e3c AS |
816 | /* |
817 | * rib_gc_dest | |
818 | * | |
819 | * Garbage collect the rib dest corresponding to the given route node | |
820 | * if appropriate. | |
821 | * | |
2951a7a4 | 822 | * Returns true if the dest was deleted, false otherwise. |
9fd92e3c | 823 | */ |
d62a17ae | 824 | int rib_gc_dest(struct route_node *rn) |
9fd92e3c | 825 | { |
d62a17ae | 826 | rib_dest_t *dest; |
9fd92e3c | 827 | |
d62a17ae | 828 | dest = rib_dest_from_rnode(rn); |
829 | if (!dest) | |
830 | return 0; | |
9fd92e3c | 831 | |
d62a17ae | 832 | if (!rib_can_delete_dest(dest)) |
833 | return 0; | |
9fd92e3c | 834 | |
c9abf558 DS |
835 | if (IS_ZEBRA_DEBUG_RIB) { |
836 | struct zebra_vrf *zvrf; | |
837 | ||
838 | zvrf = rib_dest_vrf(dest); | |
d62a17ae | 839 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); |
c9abf558 | 840 | } |
9fd92e3c | 841 | |
699dae23 DS |
842 | zebra_rib_evaluate_rn_nexthops(rn, zebra_router_get_next_sequence()); |
843 | ||
d62a17ae | 844 | dest->rnode = NULL; |
aa57abfb | 845 | rnh_list_fini(&dest->nht); |
d62a17ae | 846 | XFREE(MTYPE_RIB_DEST, dest); |
847 | rn->info = NULL; | |
9fd92e3c | 848 | |
d62a17ae | 849 | /* |
850 | * Release the one reference that we keep on the route node. | |
851 | */ | |
852 | route_unlock_node(rn); | |
853 | return 1; | |
9fd92e3c AS |
854 | } |
855 | ||
d62a17ae | 856 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
857 | struct route_entry *new) | |
3e5c6e00 | 858 | { |
d62a17ae | 859 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 860 | |
d62a17ae | 861 | /* Update real nexthop. This may actually determine if nexthop is active |
862 | * or not. */ | |
6b468511 | 863 | if (!nexthop_group_active_nexthop_num(new->ng)) { |
d62a17ae | 864 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
865 | return; | |
866 | } | |
3e5c6e00 | 867 | |
d62a17ae | 868 | if (IS_ZEBRA_DEBUG_RIB) { |
869 | char buf[SRCDEST2STR_BUFFER]; | |
870 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
871 | zlog_debug("%u:%s: Adding route rn %p, re %p (%s)", |
872 | zvrf_id(zvrf), buf, rn, new, | |
873 | zebra_route_string(new->type)); | |
d62a17ae | 874 | } |
3e5c6e00 | 875 | |
d62a17ae | 876 | /* If labeled-unicast route, install transit LSP. */ |
877 | if (zebra_rib_labeled_unicast(new)) | |
878 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 879 | |
3cdba47a | 880 | rib_install_kernel(rn, new, NULL); |
3e5c6e00 | 881 | |
d62a17ae | 882 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 883 | } |
884 | ||
d62a17ae | 885 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
886 | struct route_entry *old) | |
3e5c6e00 | 887 | { |
d62a17ae | 888 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 889 | |
d62a17ae | 890 | /* Uninstall from kernel. */ |
891 | if (IS_ZEBRA_DEBUG_RIB) { | |
892 | char buf[SRCDEST2STR_BUFFER]; | |
893 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
894 | zlog_debug("%u:%s: Deleting route rn %p, re %p (%s)", |
895 | zvrf_id(zvrf), buf, rn, old, | |
896 | zebra_route_string(old->type)); | |
d62a17ae | 897 | } |
3e5c6e00 | 898 | |
d62a17ae | 899 | /* If labeled-unicast route, uninstall transit LSP. */ |
900 | if (zebra_rib_labeled_unicast(old)) | |
901 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
902 | ||
3cdba47a | 903 | rib_uninstall_kernel(rn, old); |
d62a17ae | 904 | |
905 | /* Update nexthop for route, reset changed flag. */ | |
ef57f35f DL |
906 | /* Note: this code also handles the Linux case when an interface goes |
907 | * down, causing the kernel to delete routes without sending DELROUTE | |
908 | * notifications | |
909 | */ | |
6883bf8d | 910 | if (RIB_KERNEL_ROUTE(old)) |
212df1de PG |
911 | SET_FLAG(old->status, ROUTE_ENTRY_REMOVED); |
912 | else | |
913 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 914 | } |
915 | ||
916 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
917 | struct route_node *rn, | |
918 | struct route_entry *old, | |
919 | struct route_entry *new) | |
920 | { | |
d62a17ae | 921 | int nh_active = 0; |
d62a17ae | 922 | |
923 | /* | |
924 | * We have to install or update if a new route has been selected or | |
925 | * something has changed. | |
926 | */ | |
927 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
928 | hook_call(rib_update, rn, "updating existing route"); | |
929 | ||
930 | /* Update the nexthop; we could determine here that nexthop is | |
931 | * inactive. */ | |
6b468511 | 932 | if (nexthop_group_active_nexthop_num(new->ng)) |
d62a17ae | 933 | nh_active = 1; |
934 | ||
935 | /* If nexthop is active, install the selected route, if | |
936 | * appropriate. If | |
937 | * the install succeeds, cleanup flags for prior route, if | |
938 | * different from | |
939 | * newly selected. | |
940 | */ | |
941 | if (nh_active) { | |
942 | if (IS_ZEBRA_DEBUG_RIB) { | |
943 | char buf[SRCDEST2STR_BUFFER]; | |
944 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
945 | if (new != old) | |
946 | zlog_debug( | |
2da33d6b | 947 | "%u:%s: Updating route rn %p, re %p (%s) old %p (%s)", |
d62a17ae | 948 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
949 | zebra_route_string(new->type), |
950 | old, | |
951 | zebra_route_string(old->type)); | |
d62a17ae | 952 | else |
953 | zlog_debug( | |
2da33d6b | 954 | "%u:%s: Updating route rn %p, re %p (%s)", |
d62a17ae | 955 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 956 | zebra_route_string(new->type)); |
d62a17ae | 957 | } |
958 | ||
959 | /* If labeled-unicast route, uninstall transit LSP. */ | |
960 | if (zebra_rib_labeled_unicast(old)) | |
961 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
962 | ||
3cdba47a DS |
963 | /* |
964 | * Non-system route should be installed. | |
965 | * If labeled-unicast route, install transit | |
966 | * LSP. | |
967 | */ | |
968 | if (zebra_rib_labeled_unicast(new)) | |
969 | zebra_mpls_lsp_install(zvrf, rn, new); | |
d62a17ae | 970 | |
3cdba47a | 971 | rib_install_kernel(rn, new, old); |
d62a17ae | 972 | } |
a64448ba | 973 | |
d62a17ae | 974 | /* |
975 | * If nexthop for selected route is not active or install | |
976 | * failed, we | |
977 | * may need to uninstall and delete for redistribution. | |
978 | */ | |
ed216282 | 979 | if (!nh_active) { |
d62a17ae | 980 | if (IS_ZEBRA_DEBUG_RIB) { |
981 | char buf[SRCDEST2STR_BUFFER]; | |
982 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
983 | if (new != old) | |
984 | zlog_debug( | |
2da33d6b | 985 | "%u:%s: Deleting route rn %p, re %p (%s) old %p (%s) - nexthop inactive", |
d62a17ae | 986 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
987 | zebra_route_string(new->type), |
988 | old, | |
989 | zebra_route_string(old->type)); | |
d62a17ae | 990 | else |
991 | zlog_debug( | |
2da33d6b | 992 | "%u:%s: Deleting route rn %p, re %p (%s) - nexthop inactive", |
d62a17ae | 993 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 994 | zebra_route_string(new->type)); |
d62a17ae | 995 | } |
996 | ||
997 | /* If labeled-unicast route, uninstall transit LSP. */ | |
998 | if (zebra_rib_labeled_unicast(old)) | |
999 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1000 | ||
3cdba47a | 1001 | rib_uninstall_kernel(rn, old); |
d62a17ae | 1002 | } |
1003 | } else { | |
1004 | /* | |
1005 | * Same route selected; check if in the FIB and if not, | |
3cdba47a DS |
1006 | * re-install. This is housekeeping code to deal with |
1007 | * race conditions in kernel with linux netlink reporting | |
1008 | * interface up before IPv4 or IPv6 protocol is ready | |
d62a17ae | 1009 | * to add routes. |
1010 | */ | |
3cdba47a DS |
1011 | if (!CHECK_FLAG(new->status, ROUTE_ENTRY_INSTALLED) || |
1012 | RIB_SYSTEM_ROUTE(new)) | |
677c1dd5 | 1013 | rib_install_kernel(rn, new, NULL); |
d62a17ae | 1014 | } |
1015 | ||
1016 | /* Update prior route. */ | |
6883bf8d | 1017 | if (new != old) |
d62a17ae | 1018 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1019 | |
d62a17ae | 1020 | /* Clear changed flag. */ |
1021 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 1022 | } |
1023 | ||
d62a17ae | 1024 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
1025 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1026 | struct route_entry *alternate) | |
1027 | { | |
1028 | if (current == NULL) | |
1029 | return alternate; | |
1030 | ||
1031 | /* filter route selection in following order: | |
1032 | * - connected beats other types | |
fec4ca19 | 1033 | * - if both connected, loopback or vrf wins |
d62a17ae | 1034 | * - lower distance beats higher |
1035 | * - lower metric beats higher for equal distance | |
1036 | * - last, hence oldest, route wins tie break. | |
1037 | */ | |
1038 | ||
fec4ca19 DS |
1039 | /* Connected routes. Check to see if either are a vrf |
1040 | * or loopback interface. If not, pick the last connected | |
d62a17ae | 1041 | * route of the set of lowest metric connected routes. |
1042 | */ | |
1043 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
fec4ca19 DS |
1044 | if (current->type != ZEBRA_ROUTE_CONNECT) |
1045 | return alternate; | |
1046 | ||
1047 | /* both are connected. are either loop or vrf? */ | |
1048 | struct nexthop *nexthop = NULL; | |
1049 | ||
6b468511 | 1050 | for (ALL_NEXTHOPS_PTR(alternate->ng, nexthop)) { |
873fde8c | 1051 | struct interface *ifp = if_lookup_by_index( |
1052 | nexthop->ifindex, alternate->vrf_id); | |
1053 | ||
1054 | if (ifp && if_is_loopback_or_vrf(ifp)) | |
fec4ca19 DS |
1055 | return alternate; |
1056 | } | |
1057 | ||
6b468511 | 1058 | for (ALL_NEXTHOPS_PTR(current->ng, nexthop)) { |
873fde8c | 1059 | struct interface *ifp = if_lookup_by_index( |
1060 | nexthop->ifindex, current->vrf_id); | |
1061 | ||
1062 | if (ifp && if_is_loopback_or_vrf(ifp)) | |
fec4ca19 DS |
1063 | return current; |
1064 | } | |
1065 | ||
1066 | /* Neither are loop or vrf so pick best metric */ | |
1067 | if (alternate->metric <= current->metric) | |
d62a17ae | 1068 | return alternate; |
1069 | ||
1070 | return current; | |
1071 | } | |
3e5c6e00 | 1072 | |
d62a17ae | 1073 | if (current->type == ZEBRA_ROUTE_CONNECT) |
1074 | return current; | |
3e5c6e00 | 1075 | |
d62a17ae | 1076 | /* higher distance loses */ |
1077 | if (alternate->distance < current->distance) | |
1078 | return alternate; | |
1079 | if (current->distance < alternate->distance) | |
1080 | return current; | |
3e5c6e00 | 1081 | |
d62a17ae | 1082 | /* metric tie-breaks equal distance */ |
1083 | if (alternate->metric <= current->metric) | |
1084 | return alternate; | |
3e5c6e00 | 1085 | |
d62a17ae | 1086 | return current; |
3e5c6e00 | 1087 | } |
1088 | ||
d62a17ae | 1089 | /* Core function for processing routing information base. */ |
1090 | static void rib_process(struct route_node *rn) | |
1091 | { | |
1092 | struct route_entry *re; | |
1093 | struct route_entry *next; | |
1094 | struct route_entry *old_selected = NULL; | |
1095 | struct route_entry *new_selected = NULL; | |
1096 | struct route_entry *old_fib = NULL; | |
1097 | struct route_entry *new_fib = NULL; | |
1098 | struct route_entry *best = NULL; | |
1099 | char buf[SRCDEST2STR_BUFFER]; | |
1100 | rib_dest_t *dest; | |
1101 | struct zebra_vrf *zvrf = NULL; | |
86391e56 MS |
1102 | const struct prefix *p, *src_p; |
1103 | ||
d62a17ae | 1104 | srcdest_rnode_prefixes(rn, &p, &src_p); |
1105 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1106 | ||
1107 | assert(rn); | |
1108 | ||
1109 | dest = rib_dest_from_rnode(rn); | |
1110 | if (dest) { | |
1111 | zvrf = rib_dest_vrf(dest); | |
1112 | vrf_id = zvrf_id(zvrf); | |
1113 | } | |
bab85d4f | 1114 | |
d62a17ae | 1115 | if (IS_ZEBRA_DEBUG_RIB) |
1116 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1117 | |
d62a17ae | 1118 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1119 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1120 | |
607425e5 DS |
1121 | /* |
1122 | * we can have rn's that have a NULL info pointer | |
1123 | * (dest). As such let's not let the deref happen | |
1124 | * additionally we know RNODE_FOREACH_RE_SAFE | |
1125 | * will not iterate so we are ok. | |
1126 | */ | |
1127 | if (dest) | |
1128 | old_fib = dest->selected_fib; | |
5f7a4718 | 1129 | |
a2addae8 | 1130 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1131 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1132 | zlog_debug( | |
2da33d6b DS |
1133 | "%u:%s: Examine re %p (%s) status %x flags %x dist %d metric %d", |
1134 | vrf_id, buf, re, zebra_route_string(re->type), | |
1135 | re->status, re->flags, re->distance, | |
1136 | re->metric); | |
d62a17ae | 1137 | |
d62a17ae | 1138 | /* Currently selected re. */ |
1139 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1140 | assert(old_selected == NULL); | |
1141 | old_selected = re; | |
1142 | } | |
bab85d4f | 1143 | |
d62a17ae | 1144 | /* Skip deleted entries from selection */ |
1145 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1146 | continue; | |
1147 | ||
1148 | /* Skip unreachable nexthop. */ | |
1149 | /* This first call to nexthop_active_update is merely to | |
6883bf8d DS |
1150 | * determine if there's any change to nexthops associated |
1151 | * with this RIB entry. Now, rib_process() can be invoked due | |
1152 | * to an external event such as link down or due to | |
1153 | * next-hop-tracking evaluation. In the latter case, | |
d62a17ae | 1154 | * a decision has already been made that the NHs have changed. |
6883bf8d DS |
1155 | * So, no need to invoke a potentially expensive call again. |
1156 | * Further, since the change might be in a recursive NH which | |
1157 | * is not caught in the nexthop_active_update() code. Thus, we | |
1158 | * might miss changes to recursive NHs. | |
d62a17ae | 1159 | */ |
6883bf8d | 1160 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) |
99eabcec | 1161 | && !nexthop_active_update(rn, re)) { |
d62a17ae | 1162 | if (re->type == ZEBRA_ROUTE_TABLE) { |
1163 | /* XXX: HERE BE DRAGONS!!!!! | |
1164 | * In all honesty, I have not yet figured out | |
6883bf8d DS |
1165 | * what this part does or why the |
1166 | * ROUTE_ENTRY_CHANGED test above is correct | |
d62a17ae | 1167 | * or why we need to delete a route here, and |
6883bf8d DS |
1168 | * also not whether this concerns both selected |
1169 | * and fib route, or only selected | |
1170 | * or only fib | |
1171 | * | |
1172 | * This entry was denied by the 'ip protocol | |
1173 | * table' route-map, we need to delete it */ | |
d62a17ae | 1174 | if (re != old_selected) { |
1175 | if (IS_ZEBRA_DEBUG_RIB) | |
1176 | zlog_debug( | |
32391aff | 1177 | "%s: %u:%s: imported via import-table but denied " |
d62a17ae | 1178 | "by the ip protocol table route-map", |
32391aff | 1179 | __func__, vrf_id, buf); |
d62a17ae | 1180 | rib_unlink(rn, re); |
1181 | } else | |
1182 | SET_FLAG(re->status, | |
1183 | ROUTE_ENTRY_REMOVED); | |
1184 | } | |
1185 | ||
1186 | continue; | |
1187 | } | |
bab85d4f | 1188 | |
d62a17ae | 1189 | /* Infinite distance. */ |
1190 | if (re->distance == DISTANCE_INFINITY) { | |
1191 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1192 | continue; | |
1193 | } | |
bab85d4f | 1194 | |
d62a17ae | 1195 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1196 | best = rib_choose_best(new_fib, re); | |
1197 | if (new_fib && best != new_fib) | |
1198 | UNSET_FLAG(new_fib->status, | |
1199 | ROUTE_ENTRY_CHANGED); | |
1200 | new_fib = best; | |
1201 | } else { | |
1202 | best = rib_choose_best(new_selected, re); | |
1203 | if (new_selected && best != new_selected) | |
1204 | UNSET_FLAG(new_selected->status, | |
1205 | ROUTE_ENTRY_CHANGED); | |
1206 | new_selected = best; | |
1207 | } | |
1208 | if (best != re) | |
1209 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1210 | } /* RNODE_FOREACH_RE */ | |
1211 | ||
1212 | /* If no FIB override route, use the selected route also for FIB */ | |
1213 | if (new_fib == NULL) | |
1214 | new_fib = new_selected; | |
1215 | ||
1216 | /* After the cycle is finished, the following pointers will be set: | |
1217 | * old_selected --- RE entry currently having SELECTED | |
1218 | * new_selected --- RE entry that is newly SELECTED | |
1219 | * old_fib --- RE entry currently in kernel FIB | |
1220 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1221 | * | |
1222 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1223 | * the zclients. new_fib (which can be new_selected) will be installed | |
1224 | * in kernel. | |
1225 | */ | |
1226 | ||
1227 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1228 | zlog_debug( | |
1229 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1230 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1231 | (void *)old_fib, (void *)new_fib); | |
1232 | } | |
446bb95e | 1233 | |
d62a17ae | 1234 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1235 | * fib == selected */ | |
9d303b37 DL |
1236 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, |
1237 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1238 | |
1239 | /* Update fib according to selection results */ | |
1240 | if (new_fib && old_fib) | |
1241 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1242 | else if (new_fib) | |
1243 | rib_process_add_fib(zvrf, rn, new_fib); | |
1244 | else if (old_fib) | |
1245 | rib_process_del_fib(zvrf, rn, old_fib); | |
1246 | ||
8cb41cd6 | 1247 | /* Update SELECTED entry */ |
d62a17ae | 1248 | if (old_selected != new_selected || selected_changed) { |
93bdadae | 1249 | |
6883bf8d | 1250 | if (new_selected && new_selected != new_fib) |
d62a17ae | 1251 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); |
41ec9222 | 1252 | |
3cdba47a | 1253 | if (new_selected) |
5af4b346 MS |
1254 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); |
1255 | ||
d62a17ae | 1256 | if (old_selected) { |
40f321c0 MS |
1257 | /* |
1258 | * If we're removing the old entry, we should tell | |
1259 | * redist subscribers about that *if* they aren't | |
1260 | * going to see a redist for the new entry. | |
1261 | */ | |
1262 | if (!new_selected || CHECK_FLAG(old_selected->status, | |
1263 | ROUTE_ENTRY_REMOVED)) | |
1264 | redistribute_delete(p, src_p, | |
1265 | old_selected, | |
1266 | new_selected); | |
1267 | ||
d62a17ae | 1268 | if (old_selected != new_selected) |
1269 | UNSET_FLAG(old_selected->flags, | |
1270 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1271 | } |
d62a17ae | 1272 | } |
3e5c6e00 | 1273 | |
d62a17ae | 1274 | /* Remove all RE entries queued for removal */ |
a2addae8 | 1275 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1276 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { |
1277 | if (IS_ZEBRA_DEBUG_RIB) { | |
1278 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1279 | (void *)rn, (void *)re); | |
1280 | } | |
1281 | rib_unlink(rn, re); | |
1282 | } | |
1283 | } | |
4d38fdb4 | 1284 | |
d62a17ae | 1285 | /* |
1286 | * Check if the dest can be deleted now. | |
1287 | */ | |
1288 | rib_gc_dest(rn); | |
e96f9203 DO |
1289 | } |
1290 | ||
a1494c25 DS |
1291 | static void zebra_rib_evaluate_mpls(struct route_node *rn) |
1292 | { | |
1293 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
1294 | struct zebra_vrf *zvrf = vrf_info_lookup(VRF_DEFAULT); | |
1295 | ||
1296 | if (!dest) | |
1297 | return; | |
1298 | ||
1299 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_LSPS)) { | |
1300 | if (IS_ZEBRA_DEBUG_MPLS) | |
1301 | zlog_debug( | |
1302 | "%u: Scheduling all LSPs upon RIB completion", | |
1303 | zvrf_id(zvrf)); | |
1304 | zebra_mpls_lsp_schedule(zvrf); | |
1305 | mpls_unmark_lsps_for_processing(rn); | |
1306 | } | |
1307 | } | |
1308 | ||
e5ac2adf MS |
1309 | /* |
1310 | * Utility to match route with dplane context data | |
1311 | */ | |
1312 | static bool rib_route_match_ctx(const struct route_entry *re, | |
25779064 MS |
1313 | const struct zebra_dplane_ctx *ctx, |
1314 | bool is_update) | |
e5ac2adf MS |
1315 | { |
1316 | bool result = false; | |
1317 | ||
1318 | if (is_update) { | |
1319 | /* | |
1320 | * In 'update' case, we test info about the 'previous' or | |
1321 | * 'old' route | |
1322 | */ | |
1323 | if ((re->type == dplane_ctx_get_old_type(ctx)) && | |
1324 | (re->instance == dplane_ctx_get_old_instance(ctx))) { | |
1325 | result = true; | |
1326 | ||
1327 | /* TODO -- we're using this extra test, but it's not | |
1328 | * exactly clear why. | |
1329 | */ | |
1330 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1331 | (re->distance != dplane_ctx_get_old_distance(ctx) || | |
1332 | re->tag != dplane_ctx_get_old_tag(ctx))) { | |
1333 | result = false; | |
1334 | } | |
1335 | } | |
1336 | ||
1337 | } else { | |
1338 | /* | |
1339 | * Ordinary, single-route case using primary context info | |
1340 | */ | |
1341 | if ((dplane_ctx_get_op(ctx) != DPLANE_OP_ROUTE_DELETE) && | |
1342 | CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1343 | /* Skip route that's been deleted */ | |
1344 | goto done; | |
1345 | } | |
1346 | ||
1347 | if ((re->type == dplane_ctx_get_type(ctx)) && | |
1348 | (re->instance == dplane_ctx_get_instance(ctx))) { | |
1349 | result = true; | |
1350 | ||
1351 | /* TODO -- we're using this extra test, but it's not | |
1352 | * exactly clear why. | |
1353 | */ | |
1354 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1355 | (re->distance != dplane_ctx_get_distance(ctx) || | |
1356 | re->tag != dplane_ctx_get_tag(ctx))) { | |
1357 | result = false; | |
1358 | } | |
1359 | } | |
1360 | } | |
1361 | ||
1362 | done: | |
1363 | ||
1364 | return (result); | |
1365 | } | |
1366 | ||
3cdba47a DS |
1367 | static void zebra_rib_fixup_system(struct route_node *rn) |
1368 | { | |
1369 | struct route_entry *re; | |
1370 | ||
1371 | RNODE_FOREACH_RE(rn, re) { | |
1372 | struct nexthop *nhop; | |
1373 | ||
1374 | if (!RIB_SYSTEM_ROUTE(re)) | |
1375 | continue; | |
1376 | ||
1377 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1378 | continue; | |
1379 | ||
1380 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
6d0ee6a0 | 1381 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); |
3cdba47a | 1382 | |
6b468511 | 1383 | for (ALL_NEXTHOPS_PTR(re->ng, nhop)) { |
3cdba47a DS |
1384 | if (CHECK_FLAG(nhop->flags, NEXTHOP_FLAG_RECURSIVE)) |
1385 | continue; | |
1386 | ||
1387 | SET_FLAG(nhop->flags, NEXTHOP_FLAG_FIB); | |
1388 | } | |
1389 | } | |
1390 | } | |
1391 | ||
e5ac2adf | 1392 | /* |
78bf56b0 MS |
1393 | * Update a route from a dplane context. This consolidates common code |
1394 | * that can be used in processing of results from FIB updates, and in | |
1395 | * async notification processing. | |
efe6c026 | 1396 | * The return is 'true' if the installed nexthops changed; 'false' otherwise. |
e5ac2adf | 1397 | */ |
efe6c026 MS |
1398 | static bool rib_update_re_from_ctx(struct route_entry *re, |
1399 | struct route_node *rn, | |
1400 | struct zebra_dplane_ctx *ctx) | |
78bf56b0 | 1401 | { |
941e261c | 1402 | char dest_str[PREFIX_STRLEN] = ""; |
efe6c026 | 1403 | char nh_str[NEXTHOP_STRLEN]; |
78bf56b0 | 1404 | struct nexthop *nexthop, *ctx_nexthop; |
941e261c MS |
1405 | bool matched; |
1406 | const struct nexthop_group *ctxnhg; | |
efe6c026 MS |
1407 | bool is_selected = false; /* Is 're' currently the selected re? */ |
1408 | bool changed_p = false; /* Change to nexthops? */ | |
1409 | rib_dest_t *dest; | |
78bf56b0 | 1410 | |
941e261c MS |
1411 | /* Note well: only capturing the prefix string if debug is enabled here; |
1412 | * unconditional log messages will have to generate the string. | |
1413 | */ | |
1414 | if (IS_ZEBRA_DEBUG_RIB) | |
1415 | prefix2str(&(rn->p), dest_str, sizeof(dest_str)); | |
78bf56b0 | 1416 | |
efe6c026 MS |
1417 | dest = rib_dest_from_rnode(rn); |
1418 | if (dest) | |
1419 | is_selected = (re == dest->selected_fib); | |
1420 | ||
1421 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1422 | zlog_debug("update_from_ctx: %u:%s: %sSELECTED", | |
1423 | re->vrf_id, dest_str, (is_selected ? "" : "NOT ")); | |
1424 | ||
1425 | /* Update zebra's nexthop FIB flag for each nexthop that was installed. | |
941e261c | 1426 | * If the installed set differs from the set requested by the rib/owner, |
efe6c026 MS |
1427 | * we use the fib-specific nexthop-group to record the actual FIB |
1428 | * status. | |
78bf56b0 | 1429 | */ |
941e261c MS |
1430 | |
1431 | /* | |
1432 | * First check the fib nexthop-group, if it's present. The comparison | |
efe6c026 | 1433 | * here is quite strict: we require that the fib sets match exactly. |
941e261c MS |
1434 | */ |
1435 | matched = false; | |
1436 | do { | |
1437 | if (re->fib_ng.nexthop == NULL) | |
1438 | break; | |
1439 | ||
1440 | matched = true; | |
1441 | ||
1442 | /* First check the route's fib nexthops */ | |
1443 | for (ALL_NEXTHOPS(re->fib_ng, nexthop)) { | |
1444 | ||
1445 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1446 | continue; | |
1447 | ||
efe6c026 | 1448 | ctx_nexthop = NULL; |
941e261c MS |
1449 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), |
1450 | ctx_nexthop)) { | |
1451 | if (nexthop_same(ctx_nexthop, nexthop)) | |
1452 | break; | |
1453 | } | |
1454 | ||
1455 | if (ctx_nexthop == NULL) { | |
1456 | /* Nexthop not in the new installed set */ | |
efe6c026 MS |
1457 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1458 | nexthop2str(nexthop, nh_str, | |
1459 | sizeof(nh_str)); | |
1460 | zlog_debug("update_from_ctx: no match for fib nh %s", | |
1461 | nh_str); | |
1462 | } | |
1463 | ||
941e261c MS |
1464 | matched = false; |
1465 | break; | |
1466 | } | |
1467 | } | |
1468 | ||
1469 | if (!matched) | |
1470 | break; | |
1471 | ||
1472 | /* Check the new installed set */ | |
efe6c026 | 1473 | ctx_nexthop = NULL; |
941e261c MS |
1474 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), ctx_nexthop)) { |
1475 | ||
1476 | if (CHECK_FLAG(ctx_nexthop->flags, | |
1477 | NEXTHOP_FLAG_RECURSIVE)) | |
1478 | continue; | |
1479 | ||
1480 | /* Compare with the current group's nexthops */ | |
efe6c026 MS |
1481 | nexthop = NULL; |
1482 | for (ALL_NEXTHOPS(re->fib_ng, nexthop)) { | |
941e261c MS |
1483 | if (nexthop_same(nexthop, ctx_nexthop)) |
1484 | break; | |
1485 | } | |
1486 | ||
1487 | if (nexthop == NULL) { | |
1488 | /* Nexthop not in the old installed set */ | |
efe6c026 MS |
1489 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1490 | nexthop2str(ctx_nexthop, nh_str, | |
1491 | sizeof(nh_str)); | |
1492 | zlog_debug("update_from_ctx: no fib match for notif nh %s", | |
1493 | nh_str); | |
1494 | } | |
941e261c MS |
1495 | matched = false; |
1496 | break; | |
1497 | } | |
1498 | } | |
1499 | ||
1500 | } while (0); | |
1501 | ||
1502 | /* If the new FIB set matches the existing FIB set, we're done. */ | |
1503 | if (matched) { | |
1504 | if (IS_ZEBRA_DEBUG_RIB) | |
1505 | zlog_debug("%u:%s update_from_ctx(): existing fib nhg, no change", | |
1506 | re->vrf_id, dest_str); | |
1507 | goto done; | |
1508 | ||
1509 | } else if (re->fib_ng.nexthop) { | |
1510 | /* | |
1511 | * Free stale fib list and move on to check the rib nhg. | |
1512 | */ | |
1513 | if (IS_ZEBRA_DEBUG_RIB) | |
1514 | zlog_debug("%u:%s update_from_ctx(): replacing fib nhg", | |
1515 | re->vrf_id, dest_str); | |
1516 | nexthops_free(re->fib_ng.nexthop); | |
1517 | re->fib_ng.nexthop = NULL; | |
efe6c026 MS |
1518 | |
1519 | /* Note that the installed nexthops have changed */ | |
1520 | changed_p = true; | |
941e261c MS |
1521 | } else { |
1522 | if (IS_ZEBRA_DEBUG_RIB) | |
1523 | zlog_debug("%u:%s update_from_ctx(): no fib nhg", | |
1524 | re->vrf_id, dest_str); | |
1525 | } | |
1526 | ||
1527 | /* | |
1528 | * Compare with the rib nexthop group. The comparison here is different: | |
1529 | * the RIB group may be a superset of the list installed in the FIB. We | |
1530 | * walk the RIB group, looking for the 'installable' candidate | |
1531 | * nexthops, and then check those against the set | |
1532 | * that is actually installed. | |
1533 | */ | |
1534 | matched = true; | |
6b468511 | 1535 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) { |
78bf56b0 | 1536 | |
5695d9ac | 1537 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) |
78bf56b0 MS |
1538 | continue; |
1539 | ||
941e261c MS |
1540 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) |
1541 | continue; | |
5695d9ac | 1542 | |
941e261c | 1543 | /* Check for a FIB nexthop corresponding to the RIB nexthop */ |
efe6c026 | 1544 | ctx_nexthop = NULL; |
941e261c | 1545 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), ctx_nexthop)) { |
78bf56b0 MS |
1546 | if (nexthop_same(ctx_nexthop, nexthop)) |
1547 | break; | |
1548 | } | |
1549 | ||
5695d9ac MS |
1550 | /* If the FIB doesn't know about the nexthop, |
1551 | * it's not installed | |
1552 | */ | |
1553 | if (ctx_nexthop == NULL) { | |
efe6c026 MS |
1554 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1555 | nexthop2str(nexthop, nh_str, sizeof(nh_str)); | |
1556 | zlog_debug("update_from_ctx: no notif match for rib nh %s", | |
1557 | nh_str); | |
1558 | } | |
941e261c | 1559 | matched = false; |
efe6c026 MS |
1560 | |
1561 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1562 | changed_p = true; | |
1563 | ||
1564 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
11260e70 MS |
1565 | |
1566 | /* Keep checking nexthops */ | |
1567 | continue; | |
5695d9ac | 1568 | } |
78bf56b0 | 1569 | |
efe6c026 MS |
1570 | if (CHECK_FLAG(ctx_nexthop->flags, NEXTHOP_FLAG_FIB)) { |
1571 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1572 | changed_p = true; | |
1573 | ||
5695d9ac | 1574 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
efe6c026 MS |
1575 | } else { |
1576 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1577 | changed_p = true; | |
1578 | ||
5695d9ac | 1579 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
efe6c026 | 1580 | } |
78bf56b0 MS |
1581 | } |
1582 | ||
941e261c MS |
1583 | /* If all nexthops were processed, we're done */ |
1584 | if (matched) { | |
1585 | if (IS_ZEBRA_DEBUG_RIB) | |
efe6c026 MS |
1586 | zlog_debug("%u:%s update_from_ctx(): rib nhg matched, changed '%s'", |
1587 | re->vrf_id, dest_str, | |
1588 | (changed_p ? "true" : "false")); | |
941e261c MS |
1589 | goto done; |
1590 | } | |
78bf56b0 | 1591 | |
941e261c MS |
1592 | /* FIB nexthop set differs from the RIB set: |
1593 | * create a fib-specific nexthop-group | |
1594 | */ | |
1595 | if (IS_ZEBRA_DEBUG_RIB) | |
efe6c026 MS |
1596 | zlog_debug("%u:%s update_from_ctx(): changed %s, adding new fib nhg", |
1597 | re->vrf_id, dest_str, | |
1598 | (changed_p ? "true" : "false")); | |
941e261c MS |
1599 | |
1600 | ctxnhg = dplane_ctx_get_ng(ctx); | |
188a00e0 MS |
1601 | |
1602 | if (ctxnhg->nexthop) | |
1603 | copy_nexthops(&(re->fib_ng.nexthop), ctxnhg->nexthop, NULL); | |
1604 | else { | |
1605 | /* Bit of a special case when the fib has _no_ installed | |
1606 | * nexthops. | |
1607 | */ | |
1608 | nexthop = nexthop_new(); | |
1609 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
50d89650 | 1610 | _nexthop_add(&(re->fib_ng.nexthop), nexthop); |
188a00e0 | 1611 | } |
941e261c MS |
1612 | |
1613 | done: | |
efe6c026 | 1614 | return changed_p; |
78bf56b0 MS |
1615 | } |
1616 | ||
e5ac2adf | 1617 | /* |
fad4d69c MS |
1618 | * Helper to locate a zebra route-node from a dplane context. This is used |
1619 | * when processing dplane results, e.g. Note well: the route-node is returned | |
1620 | * with a ref held - route_unlock_node() must be called eventually. | |
e5ac2adf | 1621 | */ |
fad4d69c MS |
1622 | static struct route_node * |
1623 | rib_find_rn_from_ctx(const struct zebra_dplane_ctx *ctx) | |
e5ac2adf MS |
1624 | { |
1625 | struct route_table *table = NULL; | |
1626 | struct route_node *rn = NULL; | |
e5ac2adf MS |
1627 | const struct prefix *dest_pfx, *src_pfx; |
1628 | ||
1629 | /* Locate rn and re(s) from ctx */ | |
1630 | ||
1631 | table = zebra_vrf_table_with_table_id(dplane_ctx_get_afi(ctx), | |
1632 | dplane_ctx_get_safi(ctx), | |
1633 | dplane_ctx_get_vrf(ctx), | |
1634 | dplane_ctx_get_table(ctx)); | |
1635 | if (table == NULL) { | |
1636 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fad4d69c | 1637 | zlog_debug("Failed to find route for ctx: no table for afi %d, safi %d, vrf %u", |
e5ac2adf MS |
1638 | dplane_ctx_get_afi(ctx), |
1639 | dplane_ctx_get_safi(ctx), | |
1640 | dplane_ctx_get_vrf(ctx)); | |
1641 | } | |
1642 | goto done; | |
1643 | } | |
1644 | ||
fad4d69c MS |
1645 | dest_pfx = dplane_ctx_get_dest(ctx); |
1646 | src_pfx = dplane_ctx_get_src(ctx); | |
1647 | ||
1648 | rn = srcdest_rnode_get(table, dest_pfx, | |
1649 | src_pfx ? (struct prefix_ipv6 *)src_pfx : NULL); | |
1650 | ||
1651 | done: | |
1652 | return rn; | |
1653 | } | |
1654 | ||
1655 | ||
97f5b441 | 1656 | |
fad4d69c MS |
1657 | /* |
1658 | * Route-update results processing after async dataplane update. | |
1659 | */ | |
1660 | static void rib_process_result(struct zebra_dplane_ctx *ctx) | |
1661 | { | |
1662 | struct zebra_vrf *zvrf = NULL; | |
1663 | struct route_node *rn = NULL; | |
1664 | struct route_entry *re = NULL, *old_re = NULL, *rib; | |
1665 | bool is_update = false; | |
1666 | char dest_str[PREFIX_STRLEN] = ""; | |
1667 | enum dplane_op_e op; | |
1668 | enum zebra_dplane_result status; | |
1669 | const struct prefix *dest_pfx, *src_pfx; | |
1670 | uint32_t seq; | |
efe6c026 | 1671 | bool fib_changed = false; |
97f5b441 | 1672 | |
fad4d69c | 1673 | zvrf = vrf_info_lookup(dplane_ctx_get_vrf(ctx)); |
e5ac2adf MS |
1674 | dest_pfx = dplane_ctx_get_dest(ctx); |
1675 | ||
1676 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1677 | * unconditional log messages will have to generate the string. | |
1678 | */ | |
5709131c | 1679 | if (IS_ZEBRA_DEBUG_DPLANE) |
e5ac2adf | 1680 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
e5ac2adf | 1681 | |
fad4d69c MS |
1682 | /* Locate rn and re(s) from ctx */ |
1683 | rn = rib_find_rn_from_ctx(ctx); | |
e5ac2adf MS |
1684 | if (rn == NULL) { |
1685 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fe2c53d4 | 1686 | zlog_debug("Failed to process dplane results: no route for %u:%s", |
e5ac2adf MS |
1687 | dplane_ctx_get_vrf(ctx), dest_str); |
1688 | } | |
1689 | goto done; | |
1690 | } | |
1691 | ||
1692 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); | |
1693 | ||
1694 | op = dplane_ctx_get_op(ctx); | |
1695 | status = dplane_ctx_get_status(ctx); | |
1696 | ||
c831033f | 1697 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
f183e380 | 1698 | zlog_debug("%u:%s Processing dplane ctx %p, op %s result %s", |
e5ac2adf | 1699 | dplane_ctx_get_vrf(ctx), dest_str, ctx, |
f183e380 | 1700 | dplane_op2str(op), dplane_res2str(status)); |
e5ac2adf | 1701 | |
e5ac2adf MS |
1702 | /* |
1703 | * Update is a bit of a special case, where we may have both old and new | |
1704 | * routes to post-process. | |
1705 | */ | |
1706 | is_update = dplane_ctx_is_update(ctx); | |
1707 | ||
1708 | /* | |
1709 | * Take a pass through the routes, look for matches with the context | |
1710 | * info. | |
1711 | */ | |
1712 | RNODE_FOREACH_RE(rn, rib) { | |
1713 | ||
1714 | if (re == NULL) { | |
5709131c | 1715 | if (rib_route_match_ctx(rib, ctx, false)) |
e5ac2adf | 1716 | re = rib; |
e5ac2adf MS |
1717 | } |
1718 | ||
1719 | /* Check for old route match */ | |
1720 | if (is_update && (old_re == NULL)) { | |
5709131c | 1721 | if (rib_route_match_ctx(rib, ctx, true /*is_update*/)) |
e5ac2adf | 1722 | old_re = rib; |
e5ac2adf MS |
1723 | } |
1724 | ||
1725 | /* Have we found the routes we need to work on? */ | |
5709131c | 1726 | if (re && ((!is_update || old_re))) |
e5ac2adf | 1727 | break; |
e5ac2adf MS |
1728 | } |
1729 | ||
699dae23 DS |
1730 | seq = dplane_ctx_get_seq(ctx); |
1731 | ||
e5ac2adf MS |
1732 | /* |
1733 | * Check sequence number(s) to detect stale results before continuing | |
1734 | */ | |
60f98b23 | 1735 | if (re) { |
699dae23 | 1736 | if (re->dplane_sequence != seq) { |
60f98b23 DS |
1737 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
1738 | zlog_debug("%u:%s Stale dplane result for re %p", | |
1739 | dplane_ctx_get_vrf(ctx), | |
1740 | dest_str, re); | |
1741 | } else | |
1742 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); | |
e5ac2adf MS |
1743 | } |
1744 | ||
60f98b23 DS |
1745 | if (old_re) { |
1746 | if (old_re->dplane_sequence != dplane_ctx_get_old_seq(ctx)) { | |
1747 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
1748 | zlog_debug("%u:%s Stale dplane result for old_re %p", | |
1749 | dplane_ctx_get_vrf(ctx), | |
1750 | dest_str, old_re); | |
1751 | } else | |
9ef0c6ba | 1752 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_QUEUED); |
e5ac2adf MS |
1753 | } |
1754 | ||
12e7fe3a | 1755 | switch (op) { |
12e7fe3a DS |
1756 | case DPLANE_OP_ROUTE_INSTALL: |
1757 | case DPLANE_OP_ROUTE_UPDATE: | |
1758 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
677c1dd5 DS |
1759 | if (re) { |
1760 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
1761 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1762 | } | |
b9f0e5ee DS |
1763 | /* |
1764 | * On an update operation from the same route type | |
1765 | * context retrieval currently has no way to know | |
1766 | * which was the old and which was the new. | |
1767 | * So don't unset our flags that we just set. | |
1768 | * We know redistribution is ok because the | |
1769 | * old_re in this case is used for nothing | |
1770 | * more than knowing whom to contact if necessary. | |
1771 | */ | |
1772 | if (old_re && old_re != re) { | |
677c1dd5 DS |
1773 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
1774 | UNSET_FLAG(old_re->status, | |
1775 | ROUTE_ENTRY_INSTALLED); | |
1776 | } | |
f183e380 | 1777 | |
78bf56b0 | 1778 | /* Update zebra route based on the results in |
941e261c | 1779 | * the context struct. |
78bf56b0 | 1780 | */ |
941e261c | 1781 | if (re) { |
efe6c026 MS |
1782 | fib_changed = |
1783 | rib_update_re_from_ctx(re, rn, ctx); | |
1784 | ||
1785 | if (!fib_changed) { | |
1786 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
1787 | zlog_debug("%u:%s no fib change for re", | |
1788 | dplane_ctx_get_vrf( | |
1789 | ctx), | |
1790 | dest_str); | |
12e7fe3a DS |
1791 | } |
1792 | ||
941e261c MS |
1793 | /* Redistribute */ |
1794 | redistribute_update(dest_pfx, src_pfx, | |
228a811a | 1795 | re, old_re); |
f183e380 MS |
1796 | } |
1797 | ||
3cdba47a DS |
1798 | /* |
1799 | * System routes are weird in that they | |
1800 | * allow multiple to be installed that match | |
1801 | * to the same prefix, so after we get the | |
1802 | * result we need to clean them up so that | |
1803 | * we can actually use them. | |
1804 | */ | |
1805 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
1806 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
1807 | zebra_rib_fixup_system(rn); | |
1808 | ||
f4c6e2a8 | 1809 | if (zvrf) |
12e7fe3a | 1810 | zvrf->installs++; |
f183e380 | 1811 | |
12e7fe3a | 1812 | /* Notify route owner */ |
677c1dd5 | 1813 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_INSTALLED); |
e5ac2adf | 1814 | |
12e7fe3a | 1815 | } else { |
7a230a9d | 1816 | if (re) { |
677c1dd5 | 1817 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); |
7a230a9d DS |
1818 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
1819 | } if (old_re) | |
677c1dd5 DS |
1820 | SET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
1821 | if (re) | |
1822 | zsend_route_notify_owner(re, dest_pfx, | |
1823 | ZAPI_ROUTE_FAIL_INSTALL); | |
97f5b441 | 1824 | |
12e7fe3a DS |
1825 | zlog_warn("%u:%s: Route install failed", |
1826 | dplane_ctx_get_vrf(ctx), | |
1827 | prefix2str(dest_pfx, | |
1828 | dest_str, sizeof(dest_str))); | |
1829 | } | |
1830 | break; | |
1831 | case DPLANE_OP_ROUTE_DELETE: | |
677c1dd5 DS |
1832 | if (re) |
1833 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
1834 | /* |
1835 | * In the delete case, the zebra core datastructs were | |
1836 | * updated (or removed) at the time the delete was issued, | |
1837 | * so we're just notifying the route owner. | |
e5ac2adf | 1838 | */ |
12e7fe3a | 1839 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { |
677c1dd5 DS |
1840 | if (re) { |
1841 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1842 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
1843 | } | |
12e7fe3a | 1844 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_REMOVED); |
e5ac2adf | 1845 | |
12e7fe3a DS |
1846 | if (zvrf) |
1847 | zvrf->removals++; | |
1848 | } else { | |
677c1dd5 DS |
1849 | if (re) |
1850 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
1851 | zsend_route_notify_owner_ctx(ctx, |
1852 | ZAPI_ROUTE_REMOVE_FAIL); | |
e5ac2adf | 1853 | |
12e7fe3a DS |
1854 | zlog_warn("%u:%s: Route Deletion failure", |
1855 | dplane_ctx_get_vrf(ctx), | |
1856 | prefix2str(dest_pfx, | |
1857 | dest_str, sizeof(dest_str))); | |
1858 | } | |
3cdba47a DS |
1859 | |
1860 | /* | |
1861 | * System routes are weird in that they | |
1862 | * allow multiple to be installed that match | |
1863 | * to the same prefix, so after we get the | |
1864 | * result we need to clean them up so that | |
1865 | * we can actually use them. | |
1866 | */ | |
1867 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
1868 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
1869 | zebra_rib_fixup_system(rn); | |
12e7fe3a | 1870 | break; |
d37f4d6c MS |
1871 | default: |
1872 | break; | |
e5ac2adf | 1873 | } |
699dae23 DS |
1874 | |
1875 | zebra_rib_evaluate_rn_nexthops(rn, seq); | |
a1494c25 | 1876 | zebra_rib_evaluate_mpls(rn); |
e5ac2adf MS |
1877 | done: |
1878 | ||
5f27bcba DS |
1879 | if (rn) |
1880 | route_unlock_node(rn); | |
1881 | ||
e5ac2adf MS |
1882 | /* Return context to dataplane module */ |
1883 | dplane_ctx_fini(&ctx); | |
1884 | } | |
1885 | ||
54818e3b MS |
1886 | /* |
1887 | * Handle notification from async dataplane: the dataplane has detected | |
1888 | * some change to a route, and notifies zebra so that the control plane | |
1889 | * can reflect that change. | |
1890 | */ | |
1891 | static void rib_process_dplane_notify(struct zebra_dplane_ctx *ctx) | |
1892 | { | |
1893 | struct route_node *rn = NULL; | |
efe6c026 MS |
1894 | struct route_entry *re = NULL; |
1895 | struct nexthop *nexthop; | |
54818e3b MS |
1896 | char dest_str[PREFIX_STRLEN] = ""; |
1897 | const struct prefix *dest_pfx, *src_pfx; | |
941e261c | 1898 | rib_dest_t *dest; |
efe6c026 | 1899 | bool fib_changed = false; |
941e261c | 1900 | bool debug_p = IS_ZEBRA_DEBUG_DPLANE | IS_ZEBRA_DEBUG_RIB; |
efe6c026 | 1901 | int start_count, end_count; |
54818e3b MS |
1902 | dest_pfx = dplane_ctx_get_dest(ctx); |
1903 | ||
1904 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1905 | * unconditional log messages will have to generate the string. | |
1906 | */ | |
941e261c | 1907 | if (debug_p) |
54818e3b MS |
1908 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
1909 | ||
1910 | /* Locate rn and re(s) from ctx */ | |
1911 | rn = rib_find_rn_from_ctx(ctx); | |
1912 | if (rn == NULL) { | |
941e261c MS |
1913 | if (debug_p) { |
1914 | zlog_debug("Failed to process dplane notification: no routes for %u:%s", | |
54818e3b MS |
1915 | dplane_ctx_get_vrf(ctx), dest_str); |
1916 | } | |
1917 | goto done; | |
1918 | } | |
1919 | ||
941e261c | 1920 | dest = rib_dest_from_rnode(rn); |
54818e3b MS |
1921 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); |
1922 | ||
efe6c026 | 1923 | if (debug_p) |
54818e3b MS |
1924 | zlog_debug("%u:%s Processing dplane notif ctx %p", |
1925 | dplane_ctx_get_vrf(ctx), dest_str, ctx); | |
1926 | ||
1927 | /* | |
1928 | * Take a pass through the routes, look for matches with the context | |
1929 | * info. | |
1930 | */ | |
efe6c026 MS |
1931 | RNODE_FOREACH_RE(rn, re) { |
1932 | if (rib_route_match_ctx(re, ctx, false /*!update*/)) | |
54818e3b MS |
1933 | break; |
1934 | } | |
1935 | ||
1936 | /* No match? Nothing we can do */ | |
1937 | if (re == NULL) { | |
941e261c MS |
1938 | if (debug_p) |
1939 | zlog_debug("%u:%s Unable to process dplane notification: no entry for type %s", | |
54818e3b | 1940 | dplane_ctx_get_vrf(ctx), dest_str, |
941e261c MS |
1941 | zebra_route_string( |
1942 | dplane_ctx_get_type(ctx))); | |
54818e3b MS |
1943 | |
1944 | goto done; | |
1945 | } | |
1946 | ||
272e8903 MS |
1947 | /* Ensure we clear the QUEUED flag */ |
1948 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); | |
1949 | ||
941e261c MS |
1950 | /* Is this a notification that ... matters? We only really care about |
1951 | * the route that is currently selected for installation. | |
1952 | */ | |
1953 | if (re != dest->selected_fib) { | |
1954 | /* TODO -- don't skip processing entirely? We might like to | |
1955 | * at least report on the event. | |
1956 | */ | |
1957 | if (debug_p) | |
efe6c026 | 1958 | zlog_debug("%u:%s dplane notif, but type %s not selected_fib", |
941e261c MS |
1959 | dplane_ctx_get_vrf(ctx), dest_str, |
1960 | zebra_route_string( | |
1961 | dplane_ctx_get_type(ctx))); | |
1962 | goto done; | |
1963 | } | |
1964 | ||
efe6c026 MS |
1965 | /* We'll want to determine whether the installation status of the |
1966 | * route has changed: we'll check the status before processing, | |
1967 | * and then again if there's been a change. | |
1968 | */ | |
1969 | start_count = 0; | |
1970 | for (ALL_NEXTHOPS_PTR(rib_active_nhg(re), nexthop)) { | |
1971 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1972 | start_count++; | |
1973 | } | |
1974 | ||
54818e3b MS |
1975 | /* Update zebra's nexthop FIB flags based on the context struct's |
1976 | * nexthops. | |
1977 | */ | |
efe6c026 | 1978 | fib_changed = rib_update_re_from_ctx(re, rn, ctx); |
54818e3b | 1979 | |
efe6c026 MS |
1980 | if (!fib_changed) { |
1981 | if (debug_p) | |
54818e3b MS |
1982 | zlog_debug("%u:%s No change from dplane notification", |
1983 | dplane_ctx_get_vrf(ctx), dest_str); | |
efe6c026 MS |
1984 | |
1985 | goto done; | |
54818e3b MS |
1986 | } |
1987 | ||
188a00e0 MS |
1988 | /* |
1989 | * Perform follow-up work if the actual status of the prefix | |
efe6c026 | 1990 | * changed. |
941e261c MS |
1991 | */ |
1992 | ||
efe6c026 MS |
1993 | end_count = 0; |
1994 | for (ALL_NEXTHOPS_PTR(rib_active_nhg(re), nexthop)) { | |
1995 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1996 | end_count++; | |
1997 | } | |
941e261c | 1998 | |
188a00e0 MS |
1999 | /* Various fib transitions: changed nexthops; from installed to |
2000 | * not-installed; or not-installed to installed. | |
efe6c026 | 2001 | */ |
188a00e0 | 2002 | if (start_count > 0 && end_count > 0) { |
11260e70 MS |
2003 | if (debug_p) |
2004 | zlog_debug("%u:%s applied nexthop changes from dplane notification", | |
2005 | dplane_ctx_get_vrf(ctx), dest_str); | |
188a00e0 MS |
2006 | |
2007 | /* Changed nexthops - update kernel/others */ | |
2008 | dplane_route_notif_update(rn, re, | |
2009 | DPLANE_OP_ROUTE_UPDATE, ctx); | |
2010 | ||
2011 | } else if (start_count == 0 && end_count > 0) { | |
efe6c026 MS |
2012 | if (debug_p) |
2013 | zlog_debug("%u:%s installed transition from dplane notification", | |
2014 | dplane_ctx_get_vrf(ctx), dest_str); | |
2015 | ||
2016 | /* We expect this to be the selected route, so we want | |
272e8903 | 2017 | * to tell others about this transition. |
efe6c026 MS |
2018 | */ |
2019 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
2020 | ||
188a00e0 MS |
2021 | /* Changed nexthops - update kernel/others */ |
2022 | dplane_route_notif_update(rn, re, DPLANE_OP_ROUTE_INSTALL, ctx); | |
2023 | ||
efe6c026 MS |
2024 | /* Redistribute, lsp, and nht update */ |
2025 | redistribute_update(dest_pfx, src_pfx, re, NULL); | |
2026 | ||
2027 | zebra_rib_evaluate_rn_nexthops( | |
2028 | rn, zebra_router_get_next_sequence()); | |
2029 | ||
2030 | zebra_rib_evaluate_mpls(rn); | |
2031 | ||
2032 | } else if (start_count > 0 && end_count == 0) { | |
2033 | if (debug_p) | |
2034 | zlog_debug("%u:%s un-installed transition from dplane notification", | |
2035 | dplane_ctx_get_vrf(ctx), dest_str); | |
2036 | ||
2037 | /* Transition from _something_ installed to _nothing_ | |
2038 | * installed. | |
2039 | */ | |
2040 | /* We expect this to be the selected route, so we want | |
2041 | * to tell others about this transistion. | |
2042 | */ | |
2043 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
2044 | ||
188a00e0 MS |
2045 | /* Changed nexthops - update kernel/others */ |
2046 | dplane_route_notif_update(rn, re, DPLANE_OP_ROUTE_DELETE, ctx); | |
2047 | ||
efe6c026 | 2048 | /* Redistribute, lsp, and nht update */ |
40f321c0 | 2049 | redistribute_delete(dest_pfx, src_pfx, re, NULL); |
efe6c026 MS |
2050 | |
2051 | zebra_rib_evaluate_rn_nexthops( | |
2052 | rn, zebra_router_get_next_sequence()); | |
2053 | ||
2054 | zebra_rib_evaluate_mpls(rn); | |
2055 | } | |
941e261c | 2056 | |
54818e3b | 2057 | done: |
941e261c MS |
2058 | if (rn) |
2059 | route_unlock_node(rn); | |
2060 | ||
54818e3b MS |
2061 | /* Return context to dataplane module */ |
2062 | dplane_ctx_fini(&ctx); | |
2063 | } | |
2064 | ||
5110a0c6 | 2065 | /* Take a list of route_node structs and return 1, if there was a record |
d62a17ae | 2066 | * picked from it and processed by rib_process(). Don't process more, |
5110a0c6 | 2067 | * than one RN record; operate only in the specified sub-queue. |
e96f9203 | 2068 | */ |
d7c0a89a | 2069 | static unsigned int process_subq(struct list *subq, uint8_t qindex) |
e96f9203 | 2070 | { |
d62a17ae | 2071 | struct listnode *lnode = listhead(subq); |
2072 | struct route_node *rnode; | |
2073 | rib_dest_t *dest; | |
2074 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 2075 | |
d62a17ae | 2076 | if (!lnode) |
2077 | return 0; | |
5110a0c6 | 2078 | |
d62a17ae | 2079 | rnode = listgetdata(lnode); |
2080 | dest = rib_dest_from_rnode(rnode); | |
2081 | if (dest) | |
2082 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 2083 | |
d62a17ae | 2084 | rib_process(rnode); |
5110a0c6 | 2085 | |
d62a17ae | 2086 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
2087 | char buf[SRCDEST2STR_BUFFER]; | |
efe6c026 | 2088 | |
d62a17ae | 2089 | srcdest_rnode2str(rnode, buf, sizeof(buf)); |
2090 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
2091 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
2092 | } | |
41ec9222 | 2093 | |
d62a17ae | 2094 | if (rnode->info) |
2095 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
2096 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 2097 | |
67b9467f | 2098 | #if 0 |
5110a0c6 SH |
2099 | else |
2100 | { | |
2101 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
2102 | __func__, rnode, rnode->lock); | |
2103 | zlog_backtrace(LOG_DEBUG); | |
2104 | } | |
67b9467f | 2105 | #endif |
d62a17ae | 2106 | route_unlock_node(rnode); |
2107 | list_delete_node(subq, lnode); | |
2108 | return 1; | |
e96f9203 DO |
2109 | } |
2110 | ||
2111 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from | |
d62a17ae | 2112 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
2113 | * data | |
e96f9203 DO |
2114 | * is pointed to the meta queue structure. |
2115 | */ | |
d62a17ae | 2116 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 2117 | { |
d62a17ae | 2118 | struct meta_queue *mq = data; |
2119 | unsigned i; | |
91f16812 MS |
2120 | uint32_t queue_len, queue_limit; |
2121 | ||
2122 | /* Ensure there's room for more dataplane updates */ | |
2123 | queue_limit = dplane_get_in_queue_limit(); | |
2124 | queue_len = dplane_get_in_queue_len(); | |
2125 | if (queue_len > queue_limit) { | |
2126 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
8b962e77 MS |
2127 | zlog_debug("rib queue: dplane queue len %u, limit %u, retrying", |
2128 | queue_len, queue_limit); | |
91f16812 MS |
2129 | |
2130 | /* Ensure that the meta-queue is actually enqueued */ | |
489a9614 | 2131 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2132 | work_queue_add(zrouter.ribq, zrouter.mq); |
91f16812 MS |
2133 | |
2134 | return WQ_QUEUE_BLOCKED; | |
2135 | } | |
5110a0c6 | 2136 | |
d62a17ae | 2137 | for (i = 0; i < MQ_SIZE; i++) |
2138 | if (process_subq(mq->subq[i], i)) { | |
2139 | mq->size--; | |
2140 | break; | |
2141 | } | |
2142 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
2143 | } |
2144 | ||
f52ed677 DS |
2145 | |
2146 | /* | |
2147 | * Look into the RN and queue it into the highest priority queue | |
2148 | * at this point in time for processing. | |
2149 | * | |
2150 | * We will enqueue a route node only once per invocation. | |
2151 | * | |
2152 | * There are two possibilities here that should be kept in mind. | |
2153 | * If the original invocation has not been pulled off for processing | |
2154 | * yet, A subsuquent invocation can have a route entry with a better | |
2155 | * meta queue index value and we can have a situation where | |
2156 | * we might have the same node enqueued 2 times. Not necessarily | |
2157 | * an optimal situation but it should be ok. | |
2158 | * | |
2159 | * The other possibility is that the original invocation has not | |
2160 | * been pulled off for processing yet, A subsusquent invocation | |
2161 | * doesn't have a route_entry with a better meta-queue and the | |
2162 | * original metaqueue index value will win and we'll end up with | |
2163 | * the route node enqueued once. | |
e96f9203 | 2164 | */ |
d62a17ae | 2165 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) |
e96f9203 | 2166 | { |
f52ed677 DS |
2167 | struct route_entry *re = NULL, *curr_re = NULL; |
2168 | uint8_t qindex = MQ_SIZE, curr_qindex = MQ_SIZE; | |
5110a0c6 | 2169 | |
f52ed677 DS |
2170 | RNODE_FOREACH_RE (rn, curr_re) { |
2171 | curr_qindex = route_info[curr_re->type].meta_q_map; | |
d62a17ae | 2172 | |
f52ed677 DS |
2173 | if (curr_qindex <= qindex) { |
2174 | re = curr_re; | |
2175 | qindex = curr_qindex; | |
d62a17ae | 2176 | } |
f52ed677 | 2177 | } |
5110a0c6 | 2178 | |
f52ed677 DS |
2179 | if (!re) |
2180 | return; | |
5110a0c6 | 2181 | |
f52ed677 DS |
2182 | /* Invariant: at this point we always have rn->info set. */ |
2183 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
2184 | RIB_ROUTE_QUEUED(qindex))) { | |
d62a17ae | 2185 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2186 | rnode_debug(rn, re->vrf_id, | |
f52ed677 | 2187 | "rn %p is already queued in sub-queue %u", |
d62a17ae | 2188 | (void *)rn, qindex); |
f52ed677 | 2189 | return; |
d62a17ae | 2190 | } |
f52ed677 DS |
2191 | |
2192 | SET_FLAG(rib_dest_from_rnode(rn)->flags, RIB_ROUTE_QUEUED(qindex)); | |
2193 | listnode_add(mq->subq[qindex], rn); | |
2194 | route_lock_node(rn); | |
2195 | mq->size++; | |
2196 | ||
2197 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2198 | rnode_debug(rn, re->vrf_id, "queued rn %p into sub-queue %u", | |
2199 | (void *)rn, qindex); | |
4d38fdb4 | 2200 | } |
2201 | ||
6d691129 | 2202 | /* Add route_node to work queue and schedule processing */ |
d62a17ae | 2203 | void rib_queue_add(struct route_node *rn) |
4d38fdb4 | 2204 | { |
d62a17ae | 2205 | assert(rn); |
fc328ac9 | 2206 | |
d62a17ae | 2207 | /* Pointless to queue a route_node with no RIB entries to add or remove |
2208 | */ | |
2209 | if (!rnode_to_ribs(rn)) { | |
2210 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
2211 | __func__, (void *)rn, rn->lock); | |
2212 | zlog_backtrace(LOG_DEBUG); | |
2213 | return; | |
2214 | } | |
4d38fdb4 | 2215 | |
489a9614 | 2216 | if (zrouter.ribq == NULL) { |
e914ccbe | 2217 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2218 | "%s: work_queue does not exist!", __func__); |
d62a17ae | 2219 | return; |
2220 | } | |
2221 | ||
2222 | /* | |
2223 | * The RIB queue should normally be either empty or holding the only | |
2224 | * work_queue_item element. In the latter case this element would | |
2225 | * hold a pointer to the meta queue structure, which must be used to | |
2226 | * actually queue the route nodes to process. So create the MQ | |
2227 | * holder, if necessary, then push the work into it in any case. | |
2228 | * This semantics was introduced after 0.99.9 release. | |
2229 | */ | |
489a9614 | 2230 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2231 | work_queue_add(zrouter.ribq, zrouter.mq); |
e96f9203 | 2232 | |
ea45a4e7 | 2233 | rib_meta_queue_add(zrouter.mq, rn); |
fc328ac9 | 2234 | |
d62a17ae | 2235 | return; |
4d38fdb4 | 2236 | } |
2237 | ||
5110a0c6 SH |
2238 | /* Create new meta queue. |
2239 | A destructor function doesn't seem to be necessary here. | |
2240 | */ | |
d62a17ae | 2241 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 2242 | { |
d62a17ae | 2243 | struct meta_queue *new; |
2244 | unsigned i; | |
5110a0c6 | 2245 | |
d62a17ae | 2246 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
e96f9203 | 2247 | |
d62a17ae | 2248 | for (i = 0; i < MQ_SIZE; i++) { |
2249 | new->subq[i] = list_new(); | |
2250 | assert(new->subq[i]); | |
2251 | } | |
5110a0c6 | 2252 | |
d62a17ae | 2253 | return new; |
e96f9203 DO |
2254 | } |
2255 | ||
d62a17ae | 2256 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 2257 | { |
d62a17ae | 2258 | unsigned i; |
5a8dfcd8 | 2259 | |
d62a17ae | 2260 | for (i = 0; i < MQ_SIZE; i++) |
6a154c88 | 2261 | list_delete(&mq->subq[i]); |
5a8dfcd8 | 2262 | |
d62a17ae | 2263 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
2264 | } |
2265 | ||
4d38fdb4 | 2266 | /* initialise zebra rib work queue */ |
2561d12e | 2267 | static void rib_queue_init(void) |
4d38fdb4 | 2268 | { |
489a9614 DS |
2269 | if (!(zrouter.ribq = work_queue_new(zrouter.master, |
2270 | "route_node processing"))) { | |
e914ccbe | 2271 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2272 | "%s: could not initialise work queue!", __func__); |
d62a17ae | 2273 | return; |
2274 | } | |
4d38fdb4 | 2275 | |
d62a17ae | 2276 | /* fill in the work queue spec */ |
489a9614 DS |
2277 | zrouter.ribq->spec.workfunc = &meta_queue_process; |
2278 | zrouter.ribq->spec.errorfunc = NULL; | |
46a4e345 | 2279 | zrouter.ribq->spec.completion_func = NULL; |
d62a17ae | 2280 | /* XXX: TODO: These should be runtime configurable via vty */ |
489a9614 DS |
2281 | zrouter.ribq->spec.max_retries = 3; |
2282 | zrouter.ribq->spec.hold = ZEBRA_RIB_PROCESS_HOLD_TIME; | |
2283 | zrouter.ribq->spec.retry = ZEBRA_RIB_PROCESS_RETRY_TIME; | |
d62a17ae | 2284 | |
ea45a4e7 | 2285 | if (!(zrouter.mq = meta_queue_new())) { |
e914ccbe | 2286 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2287 | "%s: could not initialise meta queue!", __func__); |
d62a17ae | 2288 | return; |
2289 | } | |
2290 | return; | |
718e3744 | 2291 | } |
2292 | ||
434434f7 DS |
2293 | rib_dest_t *zebra_rib_create_dest(struct route_node *rn) |
2294 | { | |
2295 | rib_dest_t *dest; | |
2296 | ||
2297 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); | |
aa57abfb | 2298 | rnh_list_init(&dest->nht); |
434434f7 DS |
2299 | route_lock_node(rn); /* rn route table reference */ |
2300 | rn->info = dest; | |
2301 | dest->rnode = rn; | |
2302 | ||
2303 | return dest; | |
2304 | } | |
2305 | ||
6d691129 PJ |
2306 | /* RIB updates are processed via a queue of pointers to route_nodes. |
2307 | * | |
2308 | * The queue length is bounded by the maximal size of the routing table, | |
2309 | * as a route_node will not be requeued, if already queued. | |
2310 | * | |
f0f77c9a DS |
2311 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
2312 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 2313 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 2314 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 2315 | * |
f0f77c9a | 2316 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
2317 | * |
2318 | * rib_addnode | |
f0f77c9a DS |
2319 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
2320 | * |-------->| | best RE, if required | |
3c0755dc PJ |
2321 | * | | |
2322 | * static_install->|->rib_addqueue...... -> rib_process | |
2323 | * | | | |
2324 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
2325 | * |-> set ROUTE_ENTRY_REMOVE | |
2326 | * rib_delnode (RE freed) | |
6d691129 | 2327 | * |
9fd92e3c AS |
2328 | * The 'info' pointer of a route_node points to a rib_dest_t |
2329 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
2330 | * dest is created on-demand by rib_link() and is kept around at least | |
2331 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 2332 | * |
6d691129 PJ |
2333 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
2334 | * | |
2335 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
2336 | * - dest attached to route_node: |
2337 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
2338 | * - route_node processing queue |
2339 | * - managed by: rib_addqueue, rib_process. | |
2340 | * | |
2341 | */ | |
d62a17ae | 2342 | |
f0f77c9a | 2343 | /* Add RE to head of the route node. */ |
d62a17ae | 2344 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
2345 | { | |
d62a17ae | 2346 | rib_dest_t *dest; |
2347 | afi_t afi; | |
2348 | const char *rmap_name; | |
9fd92e3c | 2349 | |
d62a17ae | 2350 | assert(re && rn); |
9fd92e3c | 2351 | |
d62a17ae | 2352 | dest = rib_dest_from_rnode(rn); |
2353 | if (!dest) { | |
2354 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2355 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 2356 | |
434434f7 | 2357 | dest = zebra_rib_create_dest(rn); |
d62a17ae | 2358 | } |
2263a412 | 2359 | |
045207e2 | 2360 | re_list_add_head(&dest->routes, re); |
d62a17ae | 2361 | |
2362 | afi = (rn->p.family == AF_INET) | |
2363 | ? AFI_IP | |
2364 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
fe257ae7 DS |
2365 | if (is_zebra_import_table_enabled(afi, re->vrf_id, re->table)) { |
2366 | struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(re->vrf_id); | |
2367 | ||
d62a17ae | 2368 | rmap_name = zebra_get_import_table_route_map(afi, re->table); |
fe257ae7 | 2369 | zebra_add_import_table_entry(zvrf, rn, re, rmap_name); |
d62a17ae | 2370 | } else if (process) |
2371 | rib_queue_add(rn); | |
2372 | } | |
2373 | ||
7e24fdf3 DS |
2374 | static void rib_addnode(struct route_node *rn, |
2375 | struct route_entry *re, int process) | |
d62a17ae | 2376 | { |
2377 | /* RE node has been un-removed before route-node is processed. | |
2378 | * route_node must hence already be on the queue for processing.. | |
2379 | */ | |
2380 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
2381 | if (IS_ZEBRA_DEBUG_RIB) | |
2382 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
2383 | (void *)rn, (void *)re); | |
2384 | ||
2385 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2386 | return; | |
2387 | } | |
2388 | rib_link(rn, re, process); | |
6d691129 PJ |
2389 | } |
2390 | ||
9fd92e3c AS |
2391 | /* |
2392 | * rib_unlink | |
2393 | * | |
2394 | * Detach a rib structure from a route_node. | |
2395 | * | |
2396 | * Note that a call to rib_unlink() should be followed by a call to | |
2397 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
2398 | * longer required to be deleted. | |
2399 | */ | |
d62a17ae | 2400 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 2401 | { |
d62a17ae | 2402 | rib_dest_t *dest; |
22bcedb2 | 2403 | rib_table_info_t *info; |
9fd92e3c | 2404 | |
d62a17ae | 2405 | assert(rn && re); |
6d691129 | 2406 | |
d62a17ae | 2407 | if (IS_ZEBRA_DEBUG_RIB) |
2408 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
2409 | (void *)re); | |
6d691129 | 2410 | |
d62a17ae | 2411 | dest = rib_dest_from_rnode(rn); |
6d691129 | 2412 | |
045207e2 | 2413 | re_list_del(&dest->routes, re); |
7a4bb9c5 | 2414 | |
2eb07de3 DS |
2415 | if (dest->selected_fib == re) |
2416 | dest->selected_fib = NULL; | |
2417 | ||
22bcedb2 DS |
2418 | info = srcdest_rnode_table_info(rn); |
2419 | zebra_nhg_release(info->afi, re); | |
2420 | ||
6b468511 DS |
2421 | nexthops_free(re->ng->nexthop); |
2422 | nexthop_group_delete(&re->ng); | |
ee5e8a48 MS |
2423 | nexthops_free(re->fib_ng.nexthop); |
2424 | ||
d62a17ae | 2425 | XFREE(MTYPE_RE, re); |
2426 | } | |
2427 | ||
2428 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2429 | { | |
2430 | afi_t afi; | |
2431 | ||
2432 | if (IS_ZEBRA_DEBUG_RIB) | |
2433 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2434 | (void *)rn, (void *)re); | |
2435 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2436 | ||
2437 | afi = (rn->p.family == AF_INET) | |
2438 | ? AFI_IP | |
2439 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
fe257ae7 DS |
2440 | if (is_zebra_import_table_enabled(afi, re->vrf_id, re->table)) { |
2441 | struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(re->vrf_id); | |
2442 | ||
2443 | zebra_del_import_table_entry(zvrf, rn, re); | |
d62a17ae | 2444 | /* Just clean up if non main table */ |
2445 | if (IS_ZEBRA_DEBUG_RIB) { | |
2446 | char buf[SRCDEST2STR_BUFFER]; | |
2447 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
2448 | zlog_debug("%u:%s: Freeing route rn %p, re %p (%s)", |
2449 | re->vrf_id, buf, rn, re, | |
2450 | zebra_route_string(re->type)); | |
d62a17ae | 2451 | } |
7a4bb9c5 | 2452 | |
d62a17ae | 2453 | rib_unlink(rn, re); |
2454 | } else { | |
2455 | rib_queue_add(rn); | |
2456 | } | |
718e3744 | 2457 | } |
2458 | ||
f0f77c9a | 2459 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2460 | * standard debug log. Calling function name and IP prefix in |
2461 | * question are passed as 1st and 2nd arguments. | |
2462 | */ | |
2463 | ||
d62a17ae | 2464 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2465 | union prefixconstptr src_pp, | |
2466 | const struct route_entry *re) | |
2467 | { | |
d62a17ae | 2468 | const struct prefix *src_p = src_pp.p; |
2469 | bool is_srcdst = src_p && src_p->prefixlen; | |
2470 | char straddr[PREFIX_STRLEN]; | |
2471 | char srcaddr[PREFIX_STRLEN]; | |
53c16fbe | 2472 | char nhname[PREFIX_STRLEN]; |
d62a17ae | 2473 | struct nexthop *nexthop; |
2474 | ||
2475 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2476 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2477 | is_srcdst ? " from " : "", | |
2478 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2479 | : "", | |
2480 | re->vrf_id); | |
cc54cfee | 2481 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", |
53c16fbe | 2482 | straddr, (unsigned long)re->uptime, re->type, re->instance, |
cc54cfee | 2483 | re->table); |
d62a17ae | 2484 | zlog_debug( |
2485 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
53c16fbe DS |
2486 | straddr, re->metric, re->mtu, re->distance, re->flags, re->status); |
2487 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", straddr, | |
d62a17ae | 2488 | re->nexthop_num, re->nexthop_active_num); |
2489 | ||
6b468511 | 2490 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) { |
2d68a0f2 DS |
2491 | struct interface *ifp; |
2492 | struct vrf *vrf = vrf_lookup_by_id(nexthop->vrf_id); | |
2493 | ||
2494 | switch (nexthop->type) { | |
2495 | case NEXTHOP_TYPE_BLACKHOLE: | |
53c16fbe | 2496 | sprintf(nhname, "Blackhole"); |
2d68a0f2 DS |
2497 | break; |
2498 | case NEXTHOP_TYPE_IFINDEX: | |
2499 | ifp = if_lookup_by_index(nexthop->ifindex, | |
2500 | nexthop->vrf_id); | |
53c16fbe | 2501 | sprintf(nhname, "%s", ifp ? ifp->name : "Unknown"); |
2d68a0f2 DS |
2502 | break; |
2503 | case NEXTHOP_TYPE_IPV4: | |
2504 | /* fallthrough */ | |
2505 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
53c16fbe | 2506 | inet_ntop(AF_INET, &nexthop->gate, nhname, |
2d68a0f2 DS |
2507 | INET6_ADDRSTRLEN); |
2508 | break; | |
2509 | case NEXTHOP_TYPE_IPV6: | |
2510 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
53c16fbe | 2511 | inet_ntop(AF_INET6, &nexthop->gate, nhname, |
2d68a0f2 DS |
2512 | INET6_ADDRSTRLEN); |
2513 | break; | |
2514 | } | |
df38b099 | 2515 | zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s%s%s%s", |
53c16fbe | 2516 | straddr, (nexthop->rparent ? " NH" : "NH"), nhname, |
2d68a0f2 DS |
2517 | nexthop->ifindex, vrf ? vrf->name : "Unknown", |
2518 | nexthop->vrf_id, | |
d62a17ae | 2519 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) |
2520 | ? "ACTIVE " | |
2521 | : ""), | |
677c1dd5 | 2522 | (CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED) |
d62a17ae | 2523 | ? "FIB " |
2524 | : ""), | |
2525 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
df38b099 DS |
2526 | ? "RECURSIVE " |
2527 | : ""), | |
2528 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK) | |
2529 | ? "ONLINK " | |
2530 | : ""), | |
2531 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_MATCHED) | |
2532 | ? "MATCHED " | |
2533 | : ""), | |
2534 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_DUPLICATE) | |
2535 | ? "DUPLICATE " | |
d62a17ae | 2536 | : "")); |
2537 | } | |
53c16fbe | 2538 | zlog_debug("%s: dump complete", straddr); |
dc95824a DO |
2539 | } |
2540 | ||
2541 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2542 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2543 | */ |
2544 | ||
d62a17ae | 2545 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2546 | { | |
2547 | struct route_table *table; | |
2548 | struct route_node *rn; | |
2549 | struct route_entry *re; | |
2550 | char prefix_buf[INET_ADDRSTRLEN]; | |
2551 | ||
2552 | /* Lookup table. */ | |
2553 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2554 | if (!table) { | |
e914ccbe | 2555 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2556 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2557 | vrf_id); | |
d62a17ae | 2558 | return; |
2559 | } | |
2560 | ||
2561 | /* Scan the RIB table for exactly matching RE entry. */ | |
2562 | rn = route_node_lookup(table, (struct prefix *)p); | |
2563 | ||
2564 | /* No route for this prefix. */ | |
2565 | if (!rn) { | |
32391aff | 2566 | zlog_debug("%s:%u lookup failed for %s", __func__, vrf_id, |
d62a17ae | 2567 | prefix2str((struct prefix *)p, prefix_buf, |
2568 | sizeof(prefix_buf))); | |
2569 | return; | |
2570 | } | |
2571 | ||
2572 | /* Unlock node. */ | |
2573 | route_unlock_node(rn); | |
2574 | ||
2575 | /* let's go */ | |
a2addae8 | 2576 | RNODE_FOREACH_RE (rn, re) { |
32391aff DS |
2577 | zlog_debug("%s:%u rn %p, re %p: %s, %s", |
2578 | __func__, vrf_id, | |
2579 | (void *)rn, (void *)re, | |
d62a17ae | 2580 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) |
2581 | ? "removed" | |
2582 | : "NOT removed"), | |
2583 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2584 | ? "selected" | |
2585 | : "NOT selected")); | |
2586 | route_entry_dump(p, NULL, re); | |
2587 | } | |
dc95824a DO |
2588 | } |
2589 | ||
20e5ff0a DO |
2590 | /* Check if requested address assignment will fail due to another |
2591 | * route being installed by zebra in FIB already. Take necessary | |
2592 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2593 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2594 | */ |
d62a17ae | 2595 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2596 | { | |
2597 | struct route_table *table; | |
2598 | struct route_node *rn; | |
5f7a4718 | 2599 | rib_dest_t *dest; |
d62a17ae | 2600 | |
2601 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
e914ccbe | 2602 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2603 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2604 | vrf_id); | |
d62a17ae | 2605 | return; |
2606 | } | |
2607 | ||
2608 | /* No matches would be the simplest case. */ | |
2609 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2610 | return; | |
2611 | ||
2612 | /* Unlock node. */ | |
2613 | route_unlock_node(rn); | |
2614 | ||
5f7a4718 | 2615 | dest = rib_dest_from_rnode(rn); |
d62a17ae | 2616 | /* Check all RE entries. In case any changes have to be done, requeue |
2617 | * the RN into RIBQ head. If the routing message about the new connected | |
2618 | * route (generated by the IP address we are going to assign very soon) | |
2619 | * comes before the RIBQ is processed, the new RE entry will join | |
2620 | * RIBQ record already on head. This is necessary for proper | |
2621 | * revalidation | |
2622 | * of the rest of the RE. | |
2623 | */ | |
3cdba47a | 2624 | if (dest->selected_fib) { |
5f7a4718 DS |
2625 | if (IS_ZEBRA_DEBUG_RIB) { |
2626 | char buf[PREFIX_STRLEN]; | |
2627 | ||
2628 | zlog_debug("%u:%s: freeing way for connected prefix", | |
2629 | dest->selected_fib->vrf_id, | |
2630 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2631 | route_entry_dump(&rn->p, NULL, dest->selected_fib); | |
d62a17ae | 2632 | } |
5f7a4718 | 2633 | rib_uninstall(rn, dest->selected_fib); |
d62a17ae | 2634 | rib_queue_add(rn); |
ade4a886 | 2635 | } |
20e5ff0a DO |
2636 | } |
2637 | ||
d62a17ae | 2638 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
2639 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
718e3744 | 2640 | { |
d62a17ae | 2641 | struct route_table *table; |
2642 | struct route_node *rn; | |
40ecd8e4 | 2643 | struct route_entry *same = NULL; |
d62a17ae | 2644 | int ret = 0; |
b4c034b0 | 2645 | |
d62a17ae | 2646 | if (!re) |
2647 | return 0; | |
b4c034b0 | 2648 | |
1f610a1f | 2649 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
05737783 | 2650 | |
d62a17ae | 2651 | /* Lookup table. */ |
7865c65d RW |
2652 | table = zebra_vrf_table_with_table_id(afi, safi, re->vrf_id, re->table); |
2653 | if (!table) { | |
6b468511 DS |
2654 | nexthops_free(re->ng->nexthop); |
2655 | nexthop_group_delete(&re->ng); | |
7865c65d | 2656 | XFREE(MTYPE_RE, re); |
d62a17ae | 2657 | return 0; |
7865c65d | 2658 | } |
cddf391b | 2659 | |
d62a17ae | 2660 | /* Make it sure prefixlen is applied to the prefix. */ |
2661 | apply_mask(p); | |
2662 | if (src_p) | |
2663 | apply_mask_ipv6(src_p); | |
718e3744 | 2664 | |
6b468511 | 2665 | zebra_nhg_find(afi, re->ng, re); |
d62a17ae | 2666 | /* Set default distance by route type. */ |
eea2ef07 | 2667 | if (re->distance == 0) |
0492eea0 | 2668 | re->distance = route_distance(re->type); |
718e3744 | 2669 | |
d62a17ae | 2670 | /* Lookup route node.*/ |
2671 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2672 | |
40ecd8e4 DS |
2673 | /* |
2674 | * If same type of route are installed, treat it as a implicit | |
2675 | * withdraw. | |
2676 | * If the user has specified the No route replace semantics | |
2677 | * for the install don't do a route replace. | |
2678 | */ | |
a2addae8 | 2679 | RNODE_FOREACH_RE (rn, same) { |
d62a17ae | 2680 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) |
2681 | continue; | |
41ec9222 | 2682 | |
eb327fa5 RW |
2683 | if (same->type != re->type) |
2684 | continue; | |
2685 | if (same->instance != re->instance) | |
2686 | continue; | |
996c9314 LB |
2687 | if (same->type == ZEBRA_ROUTE_KERNEL |
2688 | && same->metric != re->metric) | |
eb327fa5 | 2689 | continue; |
40ecd8e4 DS |
2690 | |
2691 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && | |
2692 | same->distance != re->distance) | |
2693 | continue; | |
2694 | ||
844b3a87 | 2695 | /* |
40ecd8e4 DS |
2696 | * We should allow duplicate connected routes |
2697 | * because of IPv6 link-local routes and unnumbered | |
2698 | * interfaces on Linux. | |
844b3a87 RW |
2699 | */ |
2700 | if (same->type != ZEBRA_ROUTE_CONNECT) | |
d62a17ae | 2701 | break; |
2702 | } | |
718e3744 | 2703 | |
cf363e1b MS |
2704 | /* If this route is kernel/connected route, notify the dataplane. */ |
2705 | if (RIB_SYSTEM_ROUTE(re)) { | |
2706 | /* Notify dataplane */ | |
2707 | dplane_sys_route_add(rn, re); | |
2708 | } | |
2709 | ||
d62a17ae | 2710 | /* Link new re to node.*/ |
2711 | if (IS_ZEBRA_DEBUG_RIB) { | |
2da33d6b DS |
2712 | rnode_debug(rn, re->vrf_id, |
2713 | "Inserting route rn %p, re %p (%s) existing %p", | |
2714 | rn, re, zebra_route_string(re->type), same); | |
718e3744 | 2715 | |
d62a17ae | 2716 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2717 | route_entry_dump(p, src_p, re); | |
718e3744 | 2718 | } |
6883bf8d DS |
2719 | |
2720 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 2721 | rib_addnode(rn, re, 1); |
2722 | ret = 1; | |
6b0655a2 | 2723 | |
d62a17ae | 2724 | /* Free implicit route.*/ |
2725 | if (same) { | |
2726 | rib_delnode(rn, same); | |
2727 | ret = -1; | |
2728 | } | |
718e3744 | 2729 | |
d62a17ae | 2730 | route_unlock_node(rn); |
2731 | return ret; | |
2732 | } | |
2733 | ||
2734 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
d7c0a89a | 2735 | unsigned short instance, int flags, struct prefix *p, |
fd36be7e | 2736 | struct prefix_ipv6 *src_p, const struct nexthop *nh, |
40ecd8e4 DS |
2737 | uint32_t table_id, uint32_t metric, uint8_t distance, |
2738 | bool fromkernel) | |
d62a17ae | 2739 | { |
2740 | struct route_table *table; | |
2741 | struct route_node *rn; | |
2742 | struct route_entry *re; | |
2743 | struct route_entry *fib = NULL; | |
2744 | struct route_entry *same = NULL; | |
fd36be7e | 2745 | struct nexthop *rtnh; |
d62a17ae | 2746 | char buf2[INET6_ADDRSTRLEN]; |
5f7a4718 | 2747 | rib_dest_t *dest; |
d62a17ae | 2748 | |
1f610a1f | 2749 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
d62a17ae | 2750 | |
2751 | /* Lookup table. */ | |
2752 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2753 | if (!table) | |
2754 | return; | |
2755 | ||
2756 | /* Apply mask. */ | |
2757 | apply_mask(p); | |
2758 | if (src_p) | |
2759 | apply_mask_ipv6(src_p); | |
2760 | ||
2761 | /* Lookup route node. */ | |
2762 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2763 | if (!rn) { | |
2764 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2765 | ||
2766 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2767 | if (src_p && src_p->prefixlen) | |
2768 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2769 | else | |
2770 | src_buf[0] = '\0'; | |
2771 | ||
82d7d0e2 DS |
2772 | if (IS_ZEBRA_DEBUG_RIB) { |
2773 | struct vrf *vrf = vrf_lookup_by_id(vrf_id); | |
2774 | ||
2775 | zlog_debug("%s[%d]:%s%s%s doesn't exist in rib", | |
2776 | vrf->name, table_id, dst_buf, | |
d62a17ae | 2777 | (src_buf[0] != '\0') ? " from " : "", |
2778 | src_buf); | |
82d7d0e2 | 2779 | } |
d62a17ae | 2780 | return; |
2781 | } | |
718e3744 | 2782 | |
5f7a4718 DS |
2783 | dest = rib_dest_from_rnode(rn); |
2784 | fib = dest->selected_fib; | |
2785 | ||
d62a17ae | 2786 | /* Lookup same type route. */ |
a2addae8 | 2787 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2788 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2789 | continue; | |
2790 | ||
d62a17ae | 2791 | if (re->type != type) |
2792 | continue; | |
2793 | if (re->instance != instance) | |
2794 | continue; | |
40ecd8e4 DS |
2795 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && |
2796 | distance != re->distance) | |
2797 | continue; | |
2798 | ||
996c9314 | 2799 | if (re->type == ZEBRA_ROUTE_KERNEL && re->metric != metric) |
f19435a8 | 2800 | continue; |
6b468511 | 2801 | if (re->type == ZEBRA_ROUTE_CONNECT && (rtnh = re->ng->nexthop) |
fd36be7e DL |
2802 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { |
2803 | if (rtnh->ifindex != nh->ifindex) | |
d62a17ae | 2804 | continue; |
d62a17ae | 2805 | same = re; |
2806 | break; | |
2807 | } | |
2808 | /* Make sure that the route found has the same gateway. */ | |
2809 | else { | |
fd36be7e | 2810 | if (nh == NULL) { |
d62a17ae | 2811 | same = re; |
2812 | break; | |
2813 | } | |
6b468511 | 2814 | for (ALL_NEXTHOPS_PTR(re->ng, rtnh)) |
a5a2d802 SW |
2815 | /* |
2816 | * No guarantee all kernel send nh with labels | |
2817 | * on delete. | |
2818 | */ | |
2819 | if (nexthop_same_no_labels(rtnh, nh)) { | |
d62a17ae | 2820 | same = re; |
2821 | break; | |
2822 | } | |
2823 | if (same) | |
2824 | break; | |
2825 | } | |
2826 | } | |
2827 | /* If same type of route can't be found and this message is from | |
2828 | kernel. */ | |
2829 | if (!same) { | |
5dfeba19 DS |
2830 | /* |
2831 | * In the past(HA!) we could get here because | |
2832 | * we were receiving a route delete from the | |
2833 | * kernel and we're not marking the proto | |
2834 | * as coming from it's appropriate originator. | |
2835 | * Now that we are properly noticing the fact | |
2836 | * that the kernel has deleted our route we | |
2837 | * are not going to get called in this path | |
2838 | * I am going to leave this here because | |
2839 | * this might still work this way on non-linux | |
2840 | * platforms as well as some weird state I have | |
2841 | * not properly thought of yet. | |
2842 | * If we can show that this code path is | |
2843 | * dead then we can remove it. | |
2844 | */ | |
b8faa875 | 2845 | if (fib && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { |
d62a17ae | 2846 | if (IS_ZEBRA_DEBUG_RIB) { |
2da33d6b DS |
2847 | rnode_debug(rn, vrf_id, |
2848 | "rn %p, re %p (%s) was deleted from kernel, adding", | |
2849 | rn, fib, | |
2850 | zebra_route_string(fib->type)); | |
d62a17ae | 2851 | } |
2852 | if (allow_delete) { | |
677c1dd5 | 2853 | UNSET_FLAG(fib->status, ROUTE_ENTRY_INSTALLED); |
d62a17ae | 2854 | /* Unset flags. */ |
6b468511 | 2855 | for (rtnh = fib->ng->nexthop; rtnh; |
fd36be7e DL |
2856 | rtnh = rtnh->next) |
2857 | UNSET_FLAG(rtnh->flags, | |
d62a17ae | 2858 | NEXTHOP_FLAG_FIB); |
2859 | ||
ed216282 DS |
2860 | /* |
2861 | * This is a non FRR route | |
2862 | * as such we should mark | |
2863 | * it as deleted | |
2864 | */ | |
5f7a4718 | 2865 | dest->selected_fib = NULL; |
d62a17ae | 2866 | } else { |
2867 | /* This means someone else, other than Zebra, | |
2868 | * has deleted | |
2869 | * a Zebra router from the kernel. We will add | |
2870 | * it back */ | |
2871 | rib_install_kernel(rn, fib, NULL); | |
2872 | } | |
2873 | } else { | |
2874 | if (IS_ZEBRA_DEBUG_RIB) { | |
fd36be7e | 2875 | if (nh) |
d62a17ae | 2876 | rnode_debug( |
2877 | rn, vrf_id, | |
2878 | "via %s ifindex %d type %d " | |
2879 | "doesn't exist in rib", | |
36228974 | 2880 | inet_ntop(afi2family(afi), |
2881 | &nh->gate, buf2, | |
2882 | sizeof(buf2)), | |
2883 | nh->ifindex, type); | |
d62a17ae | 2884 | else |
2885 | rnode_debug( | |
2886 | rn, vrf_id, | |
fd36be7e DL |
2887 | "type %d doesn't exist in rib", |
2888 | type); | |
d62a17ae | 2889 | } |
2890 | route_unlock_node(rn); | |
2891 | return; | |
2892 | } | |
2893 | } | |
718e3744 | 2894 | |
5dfeba19 | 2895 | if (same) { |
996c9314 LB |
2896 | if (fromkernel && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE) |
2897 | && !allow_delete) { | |
5dfeba19 DS |
2898 | rib_install_kernel(rn, same, NULL); |
2899 | route_unlock_node(rn); | |
2900 | ||
2901 | return; | |
2902 | } | |
6134fd82 | 2903 | |
2b83602b | 2904 | /* Special handling for IPv4 or IPv6 routes sourced from |
2905 | * EVPN - the nexthop (and associated MAC) need to be | |
2906 | * uninstalled if no more refs. | |
2907 | */ | |
90264d64 | 2908 | if (CHECK_FLAG(flags, ZEBRA_FLAG_EVPN_ROUTE)) { |
6134fd82 | 2909 | struct nexthop *tmp_nh; |
2910 | ||
6b468511 | 2911 | for (ALL_NEXTHOPS_PTR(re->ng, tmp_nh)) { |
6134fd82 | 2912 | struct ipaddr vtep_ip; |
2913 | ||
2914 | memset(&vtep_ip, 0, sizeof(struct ipaddr)); | |
1ec31309 | 2915 | if (afi == AFI_IP) { |
2916 | vtep_ip.ipa_type = IPADDR_V4; | |
2917 | memcpy(&(vtep_ip.ipaddr_v4), | |
2918 | &(tmp_nh->gate.ipv4), | |
2919 | sizeof(struct in_addr)); | |
2920 | } else { | |
2921 | vtep_ip.ipa_type = IPADDR_V6; | |
2922 | memcpy(&(vtep_ip.ipaddr_v6), | |
2923 | &(tmp_nh->gate.ipv6), | |
2924 | sizeof(struct in6_addr)); | |
2925 | } | |
a317a9b9 | 2926 | zebra_vxlan_evpn_vrf_route_del(re->vrf_id, |
6134fd82 | 2927 | &vtep_ip, p); |
2928 | } | |
2929 | } | |
cf363e1b MS |
2930 | |
2931 | /* Notify dplane if system route changes */ | |
2932 | if (RIB_SYSTEM_ROUTE(re)) | |
2933 | dplane_sys_route_del(rn, same); | |
2934 | ||
d62a17ae | 2935 | rib_delnode(rn, same); |
5dfeba19 | 2936 | } |
05737783 | 2937 | |
d62a17ae | 2938 | route_unlock_node(rn); |
2939 | return; | |
2940 | } | |
718e3744 | 2941 | |
718e3744 | 2942 | |
d7c0a89a QY |
2943 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, |
2944 | unsigned short instance, int flags, struct prefix *p, | |
2945 | struct prefix_ipv6 *src_p, const struct nexthop *nh, | |
2946 | uint32_t table_id, uint32_t metric, uint32_t mtu, uint8_t distance, | |
2947 | route_tag_t tag) | |
d62a17ae | 2948 | { |
2949 | struct route_entry *re; | |
66af6845 | 2950 | struct nexthop *nexthop; |
718e3744 | 2951 | |
66af6845 | 2952 | /* Allocate new route_entry structure. */ |
d62a17ae | 2953 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); |
d62a17ae | 2954 | re->type = type; |
2955 | re->instance = instance; | |
2956 | re->distance = distance; | |
2957 | re->flags = flags; | |
2958 | re->metric = metric; | |
2959 | re->mtu = mtu; | |
2960 | re->table = table_id; | |
2961 | re->vrf_id = vrf_id; | |
2962 | re->nexthop_num = 0; | |
98572489 | 2963 | re->uptime = monotime(NULL); |
4e40b6d6 | 2964 | re->tag = tag; |
6b468511 | 2965 | re->ng = nexthop_group_new(); |
d62a17ae | 2966 | |
66af6845 RW |
2967 | /* Add nexthop. */ |
2968 | nexthop = nexthop_new(); | |
2969 | *nexthop = *nh; | |
2970 | route_entry_nexthop_add(re, nexthop); | |
718e3744 | 2971 | |
66af6845 | 2972 | return rib_add_multipath(afi, safi, p, src_p, re); |
718e3744 | 2973 | } |
2974 | ||
2a99ab95 SW |
2975 | static const char *rib_update_event2str(rib_update_event_t event) |
2976 | { | |
2977 | const char *ret = "UNKNOWN"; | |
2978 | ||
2979 | switch (event) { | |
2980 | case RIB_UPDATE_KERNEL: | |
2981 | ret = "RIB_UPDATE_KERNEL"; | |
2982 | break; | |
2983 | case RIB_UPDATE_RMAP_CHANGE: | |
2984 | ret = "RIB_UPDATE_RMAP_CHANGE"; | |
2985 | break; | |
2986 | case RIB_UPDATE_OTHER: | |
2987 | ret = "RIB_UPDATE_OTHER"; | |
2988 | break; | |
2989 | case RIB_UPDATE_MAX: | |
2990 | break; | |
2991 | } | |
2992 | ||
2993 | return ret; | |
2994 | } | |
2995 | ||
2996 | ||
2997 | /* Schedule route nodes to be processed if they match the type */ | |
2998 | static void rib_update_route_node(struct route_node *rn, int type) | |
2999 | { | |
3000 | struct route_entry *re, *next; | |
3001 | bool re_changed = false; | |
3002 | ||
3003 | RNODE_FOREACH_RE_SAFE (rn, re, next) { | |
3004 | if (type == ZEBRA_ROUTE_ALL || type == re->type) { | |
3005 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
3006 | re_changed = true; | |
3007 | } | |
3008 | } | |
3009 | ||
3010 | if (re_changed) | |
3011 | rib_queue_add(rn); | |
3012 | } | |
3013 | ||
1c848137 | 3014 | /* Schedule routes of a particular table (address-family) based on event. */ |
d5b8c216 | 3015 | void rib_update_table(struct route_table *table, rib_update_event_t event) |
d62a17ae | 3016 | { |
3017 | struct route_node *rn; | |
2a99ab95 SW |
3018 | |
3019 | if (IS_ZEBRA_DEBUG_EVENT) { | |
3020 | struct zebra_vrf *zvrf; | |
3021 | struct vrf *vrf; | |
3022 | ||
3023 | zvrf = table->info ? ((rib_table_info_t *)table->info)->zvrf | |
3024 | : NULL; | |
3025 | vrf = zvrf ? zvrf->vrf : NULL; | |
3026 | ||
3027 | zlog_debug("%s: %s VRF %s Table %u event %s", __func__, | |
3028 | table->info ? afi2str( | |
3029 | ((rib_table_info_t *)table->info)->afi) | |
3030 | : "Unknown", | |
3031 | vrf ? vrf->name : "Unknown", | |
3032 | zvrf ? zvrf->table_id : 0, | |
3033 | rib_update_event2str(event)); | |
3034 | } | |
d62a17ae | 3035 | |
3036 | /* Walk all routes and queue for processing, if appropriate for | |
3037 | * the trigger event. | |
3038 | */ | |
3039 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
1ca60f2c DS |
3040 | /* |
3041 | * If we are looking at a route node and the node | |
3042 | * has already been queued we don't | |
3043 | * need to queue it up again | |
3044 | */ | |
2a99ab95 SW |
3045 | if (rn->info |
3046 | && CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
3047 | RIB_ROUTE_ANY_QUEUED)) | |
1ca60f2c | 3048 | continue; |
2a99ab95 | 3049 | |
d62a17ae | 3050 | switch (event) { |
2a99ab95 SW |
3051 | case RIB_UPDATE_KERNEL: |
3052 | rib_update_route_node(rn, ZEBRA_ROUTE_KERNEL); | |
3053 | break; | |
d62a17ae | 3054 | case RIB_UPDATE_RMAP_CHANGE: |
3055 | case RIB_UPDATE_OTHER: | |
2a99ab95 | 3056 | rib_update_route_node(rn, ZEBRA_ROUTE_ALL); |
d62a17ae | 3057 | break; |
d62a17ae | 3058 | default: |
3059 | break; | |
3060 | } | |
3061 | } | |
b84c7253 | 3062 | } |
3063 | ||
2a99ab95 | 3064 | static void rib_update_handle_vrf(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 3065 | { |
d62a17ae | 3066 | struct route_table *table; |
1c848137 | 3067 | |
2a99ab95 SW |
3068 | if (IS_ZEBRA_DEBUG_EVENT) |
3069 | zlog_debug("%s: Handling VRF %s event %s", __func__, | |
3070 | vrf_id_to_name(vrf_id), rib_update_event2str(event)); | |
3071 | ||
d62a17ae | 3072 | /* Process routes of interested address-families. */ |
3073 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2a99ab95 | 3074 | if (table) |
d62a17ae | 3075 | rib_update_table(table, event); |
718e3744 | 3076 | |
d62a17ae | 3077 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
2a99ab95 | 3078 | if (table) |
d62a17ae | 3079 | rib_update_table(table, event); |
2a99ab95 SW |
3080 | } |
3081 | ||
3082 | static void rib_update_handle_vrf_all(rib_update_event_t event) | |
3083 | { | |
3084 | struct zebra_router_table *zrt; | |
3085 | ||
3086 | if (IS_ZEBRA_DEBUG_EVENT) | |
3087 | zlog_debug("%s: Handling VRF (ALL) event %s", __func__, | |
3088 | rib_update_event2str(event)); | |
3089 | ||
3090 | /* Just iterate over all the route tables, rather than vrf lookups */ | |
3091 | RB_FOREACH (zrt, zebra_router_table_head, &zrouter.tables) | |
3092 | rib_update_table(zrt->table, event); | |
3093 | } | |
3094 | ||
3095 | struct rib_update_ctx { | |
3096 | rib_update_event_t event; | |
3097 | bool vrf_all; | |
3098 | vrf_id_t vrf_id; | |
3099 | }; | |
3100 | ||
3101 | static struct rib_update_ctx *rib_update_ctx_init(vrf_id_t vrf_id, | |
3102 | rib_update_event_t event) | |
3103 | { | |
3104 | struct rib_update_ctx *ctx; | |
3105 | ||
3106 | ctx = XCALLOC(MTYPE_RIB_UPDATE_CTX, sizeof(struct rib_update_ctx)); | |
3107 | ||
3108 | ctx->event = event; | |
3109 | ctx->vrf_id = vrf_id; | |
3110 | ||
3111 | return ctx; | |
3112 | } | |
3113 | ||
3114 | static void rib_update_ctx_fini(struct rib_update_ctx **ctx) | |
3115 | { | |
3116 | XFREE(MTYPE_RIB_UPDATE_CTX, *ctx); | |
3117 | ||
3118 | *ctx = NULL; | |
3119 | } | |
3120 | ||
3121 | static int rib_update_handler(struct thread *thread) | |
3122 | { | |
3123 | struct rib_update_ctx *ctx; | |
3124 | ||
3125 | ctx = THREAD_ARG(thread); | |
3126 | ||
3127 | if (ctx->vrf_all) | |
3128 | rib_update_handle_vrf_all(ctx->event); | |
3129 | else | |
3130 | rib_update_handle_vrf(ctx->vrf_id, ctx->event); | |
3131 | ||
3132 | rib_update_ctx_fini(&ctx); | |
3133 | ||
3134 | return 0; | |
3135 | } | |
3136 | ||
3137 | /* | |
3138 | * Thread list to ensure we don't schedule a ton of events | |
3139 | * if interfaces are flapping for instance. | |
3140 | */ | |
3141 | static struct thread *t_rib_update_threads[RIB_UPDATE_MAX]; | |
3142 | ||
3143 | /* Schedule a RIB update event for specific vrf */ | |
3144 | void rib_update_vrf(vrf_id_t vrf_id, rib_update_event_t event) | |
3145 | { | |
3146 | struct rib_update_ctx *ctx; | |
3147 | ||
3148 | ctx = rib_update_ctx_init(vrf_id, event); | |
3149 | ||
3150 | /* Don't worry about making sure multiple rib updates for specific vrf | |
3151 | * are scheduled at once for now. If it becomes a problem, we can use a | |
3152 | * lookup of some sort to keep track of running threads via t_vrf_id | |
3153 | * like how we are doing it in t_rib_update_threads[]. | |
3154 | */ | |
3155 | thread_add_event(zrouter.master, rib_update_handler, ctx, 0, NULL); | |
3156 | ||
3157 | if (IS_ZEBRA_DEBUG_EVENT) | |
3158 | zlog_debug("%s: Scheduled VRF %s, event %s", __func__, | |
3159 | vrf_id_to_name(ctx->vrf_id), | |
3160 | rib_update_event2str(event)); | |
3161 | } | |
3162 | ||
3163 | /* Schedule a RIB update event for all vrfs */ | |
3164 | void rib_update(rib_update_event_t event) | |
3165 | { | |
3166 | struct rib_update_ctx *ctx; | |
3167 | ||
3168 | ctx = rib_update_ctx_init(0, event); | |
3169 | ||
3170 | ctx->vrf_all = true; | |
3171 | ||
3172 | if (!thread_add_event(zrouter.master, rib_update_handler, ctx, 0, | |
3173 | &t_rib_update_threads[event])) | |
3174 | rib_update_ctx_fini(&ctx); /* Already scheduled */ | |
3175 | else if (IS_ZEBRA_DEBUG_EVENT) | |
3176 | zlog_debug("%s: Schedued VRF (ALL), event %s", __func__, | |
3177 | rib_update_event2str(event)); | |
718e3744 | 3178 | } |
3179 | ||
718e3744 | 3180 | /* Delete self installed routes after zebra is relaunched. */ |
95a29032 | 3181 | void rib_sweep_table(struct route_table *table) |
d62a17ae | 3182 | { |
3183 | struct route_node *rn; | |
3184 | struct route_entry *re; | |
3185 | struct route_entry *next; | |
915902cb | 3186 | struct nexthop *nexthop; |
d62a17ae | 3187 | |
915902cb DS |
3188 | if (!table) |
3189 | return; | |
d62a17ae | 3190 | |
915902cb | 3191 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
a2addae8 | 3192 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d4644d41 | 3193 | |
915902cb DS |
3194 | if (IS_ZEBRA_DEBUG_RIB) |
3195 | route_entry_dump(&rn->p, NULL, re); | |
3196 | ||
3197 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
3198 | continue; | |
3199 | ||
3200 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
3201 | continue; | |
3202 | ||
d4644d41 DS |
3203 | /* |
3204 | * If routes are older than startup_time then | |
3205 | * we know we read them in from the kernel. | |
3206 | * As such we can safely remove them. | |
3207 | */ | |
3208 | if (zrouter.startup_time < re->uptime) | |
3209 | continue; | |
3210 | ||
915902cb DS |
3211 | /* |
3212 | * So we are starting up and have received | |
3213 | * routes from the kernel that we have installed | |
3214 | * from a previous run of zebra but not cleaned | |
3215 | * up ( say a kill -9 ) | |
3216 | * But since we haven't actually installed | |
3217 | * them yet( we received them from the kernel ) | |
3218 | * we don't think they are active. | |
3219 | * So let's pretend they are active to actually | |
3220 | * remove them. | |
3221 | * In all honesty I'm not sure if we should | |
3222 | * mark them as active when we receive them | |
3223 | * This is startup only so probably ok. | |
3224 | * | |
3225 | * If we ever decide to move rib_sweep_table | |
3226 | * to a different spot (ie startup ) | |
3227 | * this decision needs to be revisited | |
3228 | */ | |
677c1dd5 | 3229 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
6b468511 | 3230 | for (ALL_NEXTHOPS_PTR(re->ng, nexthop)) |
915902cb DS |
3231 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
3232 | ||
0c555cc6 DS |
3233 | rib_uninstall_kernel(rn, re); |
3234 | rib_delnode(rn, re); | |
915902cb DS |
3235 | } |
3236 | } | |
718e3744 | 3237 | } |
3238 | ||
3239 | /* Sweep all RIB tables. */ | |
d4644d41 | 3240 | int rib_sweep_route(struct thread *t) |
718e3744 | 3241 | { |
d62a17ae | 3242 | struct vrf *vrf; |
3243 | struct zebra_vrf *zvrf; | |
78104b9b | 3244 | |
a2addae8 | 3245 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
915902cb DS |
3246 | if ((zvrf = vrf->info) == NULL) |
3247 | continue; | |
3248 | ||
d62a17ae | 3249 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); |
3250 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3251 | } | |
95a29032 | 3252 | |
89272910 | 3253 | zebra_router_sweep_route(); |
d4644d41 DS |
3254 | |
3255 | return 0; | |
718e3744 | 3256 | } |
2ea1ab1c VT |
3257 | |
3258 | /* Remove specific by protocol routes from 'table'. */ | |
d7c0a89a | 3259 | unsigned long rib_score_proto_table(uint8_t proto, unsigned short instance, |
47a08aa9 | 3260 | struct route_table *table) |
d62a17ae | 3261 | { |
3262 | struct route_node *rn; | |
3263 | struct route_entry *re; | |
3264 | struct route_entry *next; | |
3265 | unsigned long n = 0; | |
3266 | ||
3267 | if (table) | |
3268 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
a2addae8 | 3269 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 3270 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
3271 | continue; | |
3272 | if (re->type == proto | |
3273 | && re->instance == instance) { | |
3274 | rib_delnode(rn, re); | |
3275 | n++; | |
3276 | } | |
3277 | } | |
3278 | return n; | |
2ea1ab1c VT |
3279 | } |
3280 | ||
3281 | /* Remove specific by protocol routes. */ | |
d7c0a89a | 3282 | unsigned long rib_score_proto(uint8_t proto, unsigned short instance) |
2ea1ab1c | 3283 | { |
d62a17ae | 3284 | struct vrf *vrf; |
3285 | struct zebra_vrf *zvrf; | |
d8612e65 | 3286 | struct other_route_table *ort; |
d62a17ae | 3287 | unsigned long cnt = 0; |
78104b9b | 3288 | |
d8612e65 DS |
3289 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
3290 | zvrf = vrf->info; | |
3291 | if (!zvrf) | |
3292 | continue; | |
78104b9b | 3293 | |
d8612e65 DS |
3294 | cnt += rib_score_proto_table(proto, instance, |
3295 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
3296 | + rib_score_proto_table( | |
3297 | proto, instance, | |
3298 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3299 | ||
81fddbe7 | 3300 | frr_each(otable, &zvrf->other_tables, ort) cnt += |
d8612e65 DS |
3301 | rib_score_proto_table(proto, instance, ort->table); |
3302 | } | |
47a08aa9 | 3303 | |
d62a17ae | 3304 | return cnt; |
2ea1ab1c VT |
3305 | } |
3306 | ||
718e3744 | 3307 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 3308 | void rib_close_table(struct route_table *table) |
718e3744 | 3309 | { |
d62a17ae | 3310 | struct route_node *rn; |
1e9f448f | 3311 | rib_table_info_t *info; |
5f7a4718 | 3312 | rib_dest_t *dest; |
718e3744 | 3313 | |
1e9f448f DS |
3314 | if (!table) |
3315 | return; | |
9fd92e3c | 3316 | |
6ca30e9e | 3317 | info = route_table_get_info(table); |
5adc2528 | 3318 | |
5f7a4718 DS |
3319 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
3320 | dest = rib_dest_from_rnode(rn); | |
1e9f448f | 3321 | |
5f7a4718 | 3322 | if (dest && dest->selected_fib) { |
1e9f448f DS |
3323 | if (info->safi == SAFI_UNICAST) |
3324 | hook_call(rib_update, rn, NULL); | |
3325 | ||
3cdba47a DS |
3326 | rib_uninstall_kernel(rn, dest->selected_fib); |
3327 | dest->selected_fib = NULL; | |
1e9f448f | 3328 | } |
5f7a4718 | 3329 | } |
718e3744 | 3330 | } |
3331 | ||
9bd9717b MS |
3332 | /* |
3333 | * Handler for async dataplane results after a pseudowire installation | |
3334 | */ | |
3335 | static int handle_pw_result(struct zebra_dplane_ctx *ctx) | |
3336 | { | |
9bd9717b MS |
3337 | struct zebra_pw *pw; |
3338 | struct zebra_vrf *vrf; | |
3339 | ||
3340 | /* The pseudowire code assumes success - we act on an error | |
3341 | * result for installation attempts here. | |
3342 | */ | |
3343 | if (dplane_ctx_get_op(ctx) != DPLANE_OP_PW_INSTALL) | |
3344 | goto done; | |
3345 | ||
3346 | if (dplane_ctx_get_status(ctx) != ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
3347 | vrf = zebra_vrf_lookup_by_id(dplane_ctx_get_vrf(ctx)); | |
7c7ef4a8 | 3348 | pw = zebra_pw_find(vrf, dplane_ctx_get_ifname(ctx)); |
9bd9717b MS |
3349 | if (pw) |
3350 | zebra_pw_install_failure(pw); | |
3351 | } | |
3352 | ||
3353 | done: | |
3354 | ||
9f2d0354 | 3355 | return 0; |
9bd9717b MS |
3356 | } |
3357 | ||
3358 | ||
7cdb1a84 | 3359 | /* |
d37f4d6c MS |
3360 | * Handle results from the dataplane system. Dequeue update context |
3361 | * structs, dispatch to appropriate internal handlers. | |
7cdb1a84 MS |
3362 | */ |
3363 | static int rib_process_dplane_results(struct thread *thread) | |
3364 | { | |
25779064 | 3365 | struct zebra_dplane_ctx *ctx; |
4c206c8f | 3366 | struct dplane_ctx_q ctxlist; |
2fc69f03 | 3367 | bool shut_p = false; |
4c206c8f MS |
3368 | |
3369 | /* Dequeue a list of completed updates with one lock/unlock cycle */ | |
7cdb1a84 MS |
3370 | |
3371 | do { | |
4c206c8f MS |
3372 | TAILQ_INIT(&ctxlist); |
3373 | ||
7cdb1a84 | 3374 | /* Take lock controlling queue of results */ |
00dffa8c | 3375 | frr_with_mutex(&dplane_mutex) { |
d37f4d6c | 3376 | /* Dequeue list of context structs */ |
4c206c8f | 3377 | dplane_ctx_list_append(&ctxlist, &rib_dplane_q); |
7cdb1a84 | 3378 | } |
7cdb1a84 | 3379 | |
4c206c8f MS |
3380 | /* Dequeue context block */ |
3381 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3382 | ||
3383 | /* If we've emptied the results queue, we're done */ | |
3384 | if (ctx == NULL) | |
7cdb1a84 | 3385 | break; |
7cdb1a84 | 3386 | |
2fc69f03 MS |
3387 | /* If zebra is shutting down, avoid processing results, |
3388 | * just drain the results queue. | |
3389 | */ | |
3390 | shut_p = atomic_load_explicit(&zrouter.in_shutdown, | |
3391 | memory_order_relaxed); | |
3392 | if (shut_p) { | |
3393 | while (ctx) { | |
3394 | dplane_ctx_fini(&ctx); | |
3395 | ||
3396 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3397 | } | |
3398 | ||
3399 | continue; | |
3400 | } | |
3401 | ||
4c206c8f | 3402 | while (ctx) { |
d37f4d6c MS |
3403 | switch (dplane_ctx_get_op(ctx)) { |
3404 | case DPLANE_OP_ROUTE_INSTALL: | |
3405 | case DPLANE_OP_ROUTE_UPDATE: | |
3406 | case DPLANE_OP_ROUTE_DELETE: | |
188a00e0 MS |
3407 | { |
3408 | /* Bit of special case for route updates | |
3409 | * that were generated by async notifications: | |
3410 | * we don't want to continue processing these | |
3411 | * in the rib. | |
3412 | */ | |
3413 | if (dplane_ctx_get_notif_provider(ctx) == 0) | |
3414 | rib_process_result(ctx); | |
3415 | else | |
3416 | dplane_ctx_fini(&ctx); | |
3417 | } | |
3418 | break; | |
d37f4d6c | 3419 | |
54818e3b MS |
3420 | case DPLANE_OP_ROUTE_NOTIFY: |
3421 | rib_process_dplane_notify(ctx); | |
d37f4d6c MS |
3422 | break; |
3423 | ||
3424 | case DPLANE_OP_LSP_INSTALL: | |
3425 | case DPLANE_OP_LSP_UPDATE: | |
3426 | case DPLANE_OP_LSP_DELETE: | |
188a00e0 MS |
3427 | { |
3428 | /* Bit of special case for LSP updates | |
3429 | * that were generated by async notifications: | |
3430 | * we don't want to continue processing these. | |
3431 | */ | |
3432 | if (dplane_ctx_get_notif_provider(ctx) == 0) | |
3433 | zebra_mpls_lsp_dplane_result(ctx); | |
3434 | else | |
3435 | dplane_ctx_fini(&ctx); | |
3436 | } | |
3437 | break; | |
d37f4d6c | 3438 | |
104e3ad9 MS |
3439 | case DPLANE_OP_LSP_NOTIFY: |
3440 | zebra_mpls_process_dplane_notify(ctx); | |
d37f4d6c MS |
3441 | break; |
3442 | ||
9bd9717b MS |
3443 | case DPLANE_OP_PW_INSTALL: |
3444 | case DPLANE_OP_PW_UNINSTALL: | |
3445 | handle_pw_result(ctx); | |
3446 | break; | |
3447 | ||
cf363e1b MS |
3448 | case DPLANE_OP_SYS_ROUTE_ADD: |
3449 | case DPLANE_OP_SYS_ROUTE_DELETE: | |
3450 | /* No further processing in zebra for these. */ | |
3451 | dplane_ctx_fini(&ctx); | |
3452 | break; | |
3453 | ||
036d93c0 MS |
3454 | case DPLANE_OP_MAC_INSTALL: |
3455 | case DPLANE_OP_MAC_DELETE: | |
3456 | zebra_vxlan_handle_result(ctx); | |
3457 | break; | |
3458 | ||
931fa60c MS |
3459 | /* Some op codes not handled here */ |
3460 | case DPLANE_OP_ADDR_INSTALL: | |
3461 | case DPLANE_OP_ADDR_UNINSTALL: | |
3462 | case DPLANE_OP_NEIGH_INSTALL: | |
3463 | case DPLANE_OP_NEIGH_UPDATE: | |
3464 | case DPLANE_OP_NEIGH_DELETE: | |
0bbd4ff4 MS |
3465 | case DPLANE_OP_VTEP_ADD: |
3466 | case DPLANE_OP_VTEP_DELETE: | |
931fa60c | 3467 | case DPLANE_OP_NONE: |
d37f4d6c MS |
3468 | /* Don't expect this: just return the struct? */ |
3469 | dplane_ctx_fini(&ctx); | |
3470 | break; | |
931fa60c | 3471 | |
d37f4d6c | 3472 | } /* Dispatch by op code */ |
4c206c8f MS |
3473 | |
3474 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3475 | } | |
3476 | ||
5709131c | 3477 | } while (1); |
7cdb1a84 | 3478 | |
5709131c | 3479 | return 0; |
7cdb1a84 MS |
3480 | } |
3481 | ||
3482 | /* | |
3483 | * Results are returned from the dataplane subsystem, in the context of | |
1bcea841 | 3484 | * the dataplane pthread. We enqueue the results here for processing by |
7cdb1a84 MS |
3485 | * the main thread later. |
3486 | */ | |
4c206c8f | 3487 | static int rib_dplane_results(struct dplane_ctx_q *ctxlist) |
7cdb1a84 MS |
3488 | { |
3489 | /* Take lock controlling queue of results */ | |
00dffa8c | 3490 | frr_with_mutex(&dplane_mutex) { |
4c206c8f MS |
3491 | /* Enqueue context blocks */ |
3492 | dplane_ctx_list_append(&rib_dplane_q, ctxlist); | |
7cdb1a84 | 3493 | } |
7cdb1a84 | 3494 | |
4c206c8f | 3495 | /* Ensure event is signalled to zebra main pthread */ |
3801e764 | 3496 | thread_add_event(zrouter.master, rib_process_dplane_results, NULL, 0, |
7cdb1a84 MS |
3497 | &t_dplane); |
3498 | ||
5709131c | 3499 | return 0; |
7cdb1a84 MS |
3500 | } |
3501 | ||
eaa2716d SW |
3502 | /* |
3503 | * Ensure there are no empty slots in the route_info array. | |
3504 | * Every route type in zebra should be present there. | |
3505 | */ | |
3506 | static void check_route_info(void) | |
3507 | { | |
3508 | int len = array_size(route_info); | |
3509 | ||
3510 | /* | |
3511 | * ZEBRA_ROUTE_SYSTEM is special cased since | |
3512 | * its key is 0 anyway. | |
3513 | * | |
3514 | * ZEBRA_ROUTE_ALL is also ignored. | |
3515 | */ | |
3516 | for (int i = 0; i < len; i++) { | |
3517 | if (i == ZEBRA_ROUTE_SYSTEM || i == ZEBRA_ROUTE_ALL) | |
3518 | continue; | |
3519 | assert(route_info[i].key); | |
8dc7a759 | 3520 | assert(route_info[i].meta_q_map < MQ_SIZE); |
eaa2716d SW |
3521 | } |
3522 | } | |
3523 | ||
718e3744 | 3524 | /* Routing information base initialize. */ |
d62a17ae | 3525 | void rib_init(void) |
718e3744 | 3526 | { |
eaa2716d SW |
3527 | check_route_info(); |
3528 | ||
2561d12e | 3529 | rib_queue_init(); |
7cdb1a84 MS |
3530 | |
3531 | /* Init dataplane, and register for results */ | |
3532 | pthread_mutex_init(&dplane_mutex, NULL); | |
3533 | TAILQ_INIT(&rib_dplane_q); | |
4c206c8f | 3534 | zebra_dplane_init(rib_dplane_results); |
718e3744 | 3535 | } |
0915bb0c AS |
3536 | |
3537 | /* | |
3538 | * vrf_id_get_next | |
3539 | * | |
3540 | * Get the first vrf id that is greater than the given vrf id if any. | |
3541 | * | |
2951a7a4 | 3542 | * Returns true if a vrf id was found, false otherwise. |
0915bb0c | 3543 | */ |
d62a17ae | 3544 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 3545 | { |
d62a17ae | 3546 | struct vrf *vrf; |
b72ede27 | 3547 | |
d62a17ae | 3548 | vrf = vrf_lookup_by_id(vrf_id); |
3549 | if (vrf) { | |
3550 | vrf = RB_NEXT(vrf_id_head, vrf); | |
3551 | if (vrf) { | |
3552 | *next_id_p = vrf->vrf_id; | |
3553 | return 1; | |
3554 | } | |
3555 | } | |
0915bb0c | 3556 | |
d62a17ae | 3557 | return 0; |
0915bb0c AS |
3558 | } |
3559 | ||
3560 | /* | |
3561 | * rib_tables_iter_next | |
3562 | * | |
3563 | * Returns the next table in the iteration. | |
3564 | */ | |
d62a17ae | 3565 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
3566 | { | |
3567 | struct route_table *table; | |
3568 | ||
3569 | /* | |
3570 | * Array that helps us go over all AFI/SAFI combinations via one | |
3571 | * index. | |
3572 | */ | |
3573 | static struct { | |
3574 | afi_t afi; | |
3575 | safi_t safi; | |
3576 | } afi_safis[] = { | |
3577 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
3578 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
3579 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
3580 | }; | |
3581 | ||
3582 | table = NULL; | |
3583 | ||
3584 | switch (iter->state) { | |
3585 | ||
3586 | case RIB_TABLES_ITER_S_INIT: | |
3587 | iter->vrf_id = VRF_DEFAULT; | |
3588 | iter->afi_safi_ix = -1; | |
3589 | ||
3590 | /* Fall through */ | |
3591 | ||
3592 | case RIB_TABLES_ITER_S_ITERATING: | |
3593 | iter->afi_safi_ix++; | |
3594 | while (1) { | |
3595 | ||
3596 | while (iter->afi_safi_ix | |
7e3a1ec7 | 3597 | < (int)array_size(afi_safis)) { |
d62a17ae | 3598 | table = zebra_vrf_table( |
3599 | afi_safis[iter->afi_safi_ix].afi, | |
3600 | afi_safis[iter->afi_safi_ix].safi, | |
3601 | iter->vrf_id); | |
3602 | if (table) | |
3603 | break; | |
3604 | ||
3605 | iter->afi_safi_ix++; | |
3606 | } | |
3607 | ||
3608 | /* | |
3609 | * Found another table in this vrf. | |
3610 | */ | |
3611 | if (table) | |
3612 | break; | |
3613 | ||
3614 | /* | |
3615 | * Done with all tables in the current vrf, go to the | |
3616 | * next | |
3617 | * one. | |
3618 | */ | |
3619 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
3620 | break; | |
3621 | ||
3622 | iter->afi_safi_ix = 0; | |
3623 | } | |
0915bb0c | 3624 | |
0915bb0c AS |
3625 | break; |
3626 | ||
d62a17ae | 3627 | case RIB_TABLES_ITER_S_DONE: |
3628 | return NULL; | |
0915bb0c AS |
3629 | } |
3630 | ||
d62a17ae | 3631 | if (table) |
3632 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
3633 | else | |
3634 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 3635 | |
d62a17ae | 3636 | return table; |
0915bb0c | 3637 | } |