]>
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" |
718e3744 | 40 | |
89272910 | 41 | #include "zebra/zebra_router.h" |
43e52561 QY |
42 | #include "zebra/connected.h" |
43 | #include "zebra/debug.h" | |
44 | #include "zebra/interface.h" | |
45 | #include "zebra/redistribute.h" | |
718e3744 | 46 | #include "zebra/rib.h" |
47 | #include "zebra/rt.h" | |
43e52561 QY |
48 | #include "zebra/zapi_msg.h" |
49 | #include "zebra/zebra_errors.h" | |
50 | #include "zebra/zebra_memory.h" | |
7c551956 | 51 | #include "zebra/zebra_ns.h" |
fb018d25 | 52 | #include "zebra/zebra_rnh.h" |
43e52561 QY |
53 | #include "zebra/zebra_routemap.h" |
54 | #include "zebra/zebra_vrf.h" | |
6134fd82 | 55 | #include "zebra/zebra_vxlan.h" |
7cdb1a84 MS |
56 | #include "zebra/zapi_msg.h" |
57 | #include "zebra/zebra_dplane.h" | |
ad28e79a | 58 | #include "zebra/zebra_nhg.h" |
7cdb1a84 MS |
59 | |
60 | /* | |
61 | * Event, list, and mutex for delivery of dataplane results | |
62 | */ | |
63 | static pthread_mutex_t dplane_mutex; | |
64 | static struct thread *t_dplane; | |
25779064 | 65 | static struct dplane_ctx_q rib_dplane_q; |
718e3744 | 66 | |
d62a17ae | 67 | DEFINE_HOOK(rib_update, (struct route_node * rn, const char *reason), |
68 | (rn, reason)) | |
4f8ea50c | 69 | |
6baf7bb8 DS |
70 | /* Should we allow non Quagga processes to delete our routes */ |
71 | extern int allow_delete; | |
72 | ||
718e3744 | 73 | /* Each route type's string and default distance value. */ |
d62a17ae | 74 | static const struct { |
75 | int key; | |
8dc7a759 | 76 | uint8_t distance; |
73547a75 | 77 | uint8_t meta_q_map; |
d62a17ae | 78 | } route_info[ZEBRA_ROUTE_MAX] = { |
73547a75 DS |
79 | [ZEBRA_ROUTE_SYSTEM] = {ZEBRA_ROUTE_SYSTEM, 0, 4}, |
80 | [ZEBRA_ROUTE_KERNEL] = {ZEBRA_ROUTE_KERNEL, 0, 0}, | |
81 | [ZEBRA_ROUTE_CONNECT] = {ZEBRA_ROUTE_CONNECT, 0, 0}, | |
82 | [ZEBRA_ROUTE_STATIC] = {ZEBRA_ROUTE_STATIC, 1, 1}, | |
83 | [ZEBRA_ROUTE_RIP] = {ZEBRA_ROUTE_RIP, 120, 2}, | |
84 | [ZEBRA_ROUTE_RIPNG] = {ZEBRA_ROUTE_RIPNG, 120, 2}, | |
85 | [ZEBRA_ROUTE_OSPF] = {ZEBRA_ROUTE_OSPF, 110, 2}, | |
86 | [ZEBRA_ROUTE_OSPF6] = {ZEBRA_ROUTE_OSPF6, 110, 2}, | |
87 | [ZEBRA_ROUTE_ISIS] = {ZEBRA_ROUTE_ISIS, 115, 2}, | |
88 | [ZEBRA_ROUTE_BGP] = {ZEBRA_ROUTE_BGP, 20 /* IBGP is 200. */, 3}, | |
89 | [ZEBRA_ROUTE_PIM] = {ZEBRA_ROUTE_PIM, 255, 4}, | |
90 | [ZEBRA_ROUTE_EIGRP] = {ZEBRA_ROUTE_EIGRP, 90, 2}, | |
91 | [ZEBRA_ROUTE_NHRP] = {ZEBRA_ROUTE_NHRP, 10, 2}, | |
92 | [ZEBRA_ROUTE_HSLS] = {ZEBRA_ROUTE_HSLS, 255, 4}, | |
93 | [ZEBRA_ROUTE_OLSR] = {ZEBRA_ROUTE_OLSR, 255, 4}, | |
94 | [ZEBRA_ROUTE_TABLE] = {ZEBRA_ROUTE_TABLE, 150, 1}, | |
95 | [ZEBRA_ROUTE_LDP] = {ZEBRA_ROUTE_LDP, 150, 4}, | |
96 | [ZEBRA_ROUTE_VNC] = {ZEBRA_ROUTE_VNC, 20, 3}, | |
97 | [ZEBRA_ROUTE_VNC_DIRECT] = {ZEBRA_ROUTE_VNC_DIRECT, 20, 3}, | |
98 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = {ZEBRA_ROUTE_VNC_DIRECT_RH, 20, 3}, | |
99 | [ZEBRA_ROUTE_BGP_DIRECT] = {ZEBRA_ROUTE_BGP_DIRECT, 20, 3}, | |
100 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = {ZEBRA_ROUTE_BGP_DIRECT_EXT, 20, 3}, | |
101 | [ZEBRA_ROUTE_BABEL] = {ZEBRA_ROUTE_BABEL, 100, 2}, | |
102 | [ZEBRA_ROUTE_SHARP] = {ZEBRA_ROUTE_SHARP, 150, 4}, | |
98156652 | 103 | [ZEBRA_ROUTE_PBR] = {ZEBRA_ROUTE_PBR, 200, 4}, |
42d96b73 | 104 | [ZEBRA_ROUTE_BFD] = {ZEBRA_ROUTE_BFD, 255, 4}, |
eab7b6e3 | 105 | [ZEBRA_ROUTE_OPENFABRIC] = {ZEBRA_ROUTE_OPENFABRIC, 115, 2}, |
303b93cd | 106 | [ZEBRA_ROUTE_VRRP] = {ZEBRA_ROUTE_VRRP, 255, 4} |
d6abd8b0 | 107 | /* Any new route type added to zebra, should be mirrored here */ |
c710b277 | 108 | |
d62a17ae | 109 | /* no entry/default: 150 */ |
718e3744 | 110 | }; |
6b0655a2 | 111 | |
d62a17ae | 112 | static void __attribute__((format(printf, 5, 6))) |
113 | _rnode_zlog(const char *_func, vrf_id_t vrf_id, struct route_node *rn, | |
114 | int priority, const char *msgfmt, ...) | |
2263a412 | 115 | { |
d62a17ae | 116 | char buf[SRCDEST2STR_BUFFER + sizeof(" (MRIB)")]; |
117 | char msgbuf[512]; | |
118 | va_list ap; | |
2263a412 | 119 | |
d62a17ae | 120 | va_start(ap, msgfmt); |
121 | vsnprintf(msgbuf, sizeof(msgbuf), msgfmt, ap); | |
122 | va_end(ap); | |
2263a412 | 123 | |
d62a17ae | 124 | if (rn) { |
125 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
126 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
cb653491 | 127 | |
d62a17ae | 128 | if (info->safi == SAFI_MULTICAST) |
eab4a5c2 | 129 | strlcat(buf, " (MRIB)", sizeof(buf)); |
d62a17ae | 130 | } else { |
131 | snprintf(buf, sizeof(buf), "{(route_node *) NULL}"); | |
132 | } | |
2263a412 | 133 | |
d62a17ae | 134 | zlog(priority, "%s: %d:%s: %s", _func, vrf_id, buf, msgbuf); |
2263a412 DL |
135 | } |
136 | ||
d62a17ae | 137 | #define rnode_debug(node, vrf_id, ...) \ |
2263a412 | 138 | _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__) |
d62a17ae | 139 | #define rnode_info(node, ...) \ |
2263a412 DL |
140 | _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__) |
141 | ||
fd289fc8 | 142 | uint8_t route_distance(int type) |
40c7bdb0 | 143 | { |
fd289fc8 | 144 | uint8_t distance; |
40c7bdb0 | 145 | |
d62a17ae | 146 | if ((unsigned)type >= array_size(route_info)) |
147 | distance = 150; | |
148 | else | |
149 | distance = route_info[type].distance; | |
40c7bdb0 | 150 | |
d62a17ae | 151 | return distance; |
40c7bdb0 | 152 | } |
153 | ||
d7c0a89a | 154 | int is_zebra_valid_kernel_table(uint32_t table_id) |
7a4bb9c5 | 155 | { |
8f500a1c | 156 | #ifdef linux |
d62a17ae | 157 | if ((table_id == RT_TABLE_UNSPEC) || (table_id == RT_TABLE_LOCAL) |
158 | || (table_id == RT_TABLE_COMPAT)) | |
159 | return 0; | |
8f500a1c RW |
160 | #endif |
161 | ||
d62a17ae | 162 | return 1; |
7a4bb9c5 DS |
163 | } |
164 | ||
d7c0a89a | 165 | int is_zebra_main_routing_table(uint32_t table_id) |
7a4bb9c5 | 166 | { |
c447ad08 | 167 | if (table_id == RT_TABLE_MAIN) |
d62a17ae | 168 | return 1; |
169 | return 0; | |
7a4bb9c5 DS |
170 | } |
171 | ||
86391e56 | 172 | int zebra_check_addr(const struct prefix *p) |
0aabccc0 | 173 | { |
d62a17ae | 174 | if (p->family == AF_INET) { |
d7c0a89a | 175 | uint32_t addr; |
0aabccc0 | 176 | |
d62a17ae | 177 | addr = p->u.prefix4.s_addr; |
178 | addr = ntohl(addr); | |
0aabccc0 | 179 | |
d62a17ae | 180 | if (IPV4_NET127(addr) || IN_CLASSD(addr) |
181 | || IPV4_LINKLOCAL(addr)) | |
182 | return 0; | |
183 | } | |
184 | if (p->family == AF_INET6) { | |
185 | if (IN6_IS_ADDR_LOOPBACK(&p->u.prefix6)) | |
186 | return 0; | |
187 | if (IN6_IS_ADDR_LINKLOCAL(&p->u.prefix6)) | |
188 | return 0; | |
189 | } | |
190 | return 1; | |
0aabccc0 DD |
191 | } |
192 | ||
fa713d9e | 193 | /* Add nexthop to the end of a rib node's nexthop list */ |
d62a17ae | 194 | void route_entry_nexthop_add(struct route_entry *re, struct nexthop *nexthop) |
fa713d9e | 195 | { |
cf05bc94 | 196 | _nexthop_group_add_sorted(&re->ng, nexthop); |
d62a17ae | 197 | re->nexthop_num++; |
718e3744 | 198 | } |
199 | ||
6e26278c | 200 | |
6e26278c DS |
201 | /** |
202 | * copy_nexthop - copy a nexthop to the rib structure. | |
203 | */ | |
d62a17ae | 204 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) |
6e26278c | 205 | { |
7ee30f28 DS |
206 | assert(!re->ng.nexthop); |
207 | copy_nexthops(&re->ng.nexthop, nh, NULL); | |
d62a17ae | 208 | for (struct nexthop *nexthop = nh; nexthop; nexthop = nexthop->next) |
209 | re->nexthop_num++; | |
6e26278c DS |
210 | } |
211 | ||
718e3744 | 212 | /* Delete specified nexthop from the list. */ |
d62a17ae | 213 | void route_entry_nexthop_delete(struct route_entry *re, struct nexthop *nexthop) |
718e3744 | 214 | { |
d62a17ae | 215 | if (nexthop->next) |
216 | nexthop->next->prev = nexthop->prev; | |
217 | if (nexthop->prev) | |
218 | nexthop->prev->next = nexthop->next; | |
219 | else | |
7ee30f28 | 220 | re->ng.nexthop = nexthop->next; |
d62a17ae | 221 | re->nexthop_num--; |
718e3744 | 222 | } |
223 | ||
fa713d9e | 224 | |
d62a17ae | 225 | struct nexthop *route_entry_nexthop_ifindex_add(struct route_entry *re, |
4a7371e9 DS |
226 | ifindex_t ifindex, |
227 | vrf_id_t nh_vrf_id) | |
718e3744 | 228 | { |
d62a17ae | 229 | struct nexthop *nexthop; |
718e3744 | 230 | |
d62a17ae | 231 | nexthop = nexthop_new(); |
232 | nexthop->type = NEXTHOP_TYPE_IFINDEX; | |
233 | nexthop->ifindex = ifindex; | |
4a7371e9 | 234 | nexthop->vrf_id = nh_vrf_id; |
718e3744 | 235 | |
d62a17ae | 236 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 237 | |
d62a17ae | 238 | return nexthop; |
718e3744 | 239 | } |
240 | ||
d62a17ae | 241 | struct nexthop *route_entry_nexthop_ipv4_add(struct route_entry *re, |
242 | struct in_addr *ipv4, | |
4a7371e9 DS |
243 | struct in_addr *src, |
244 | vrf_id_t nh_vrf_id) | |
718e3744 | 245 | { |
d62a17ae | 246 | struct nexthop *nexthop; |
718e3744 | 247 | |
d62a17ae | 248 | nexthop = nexthop_new(); |
249 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
4a7371e9 | 250 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 251 | nexthop->gate.ipv4 = *ipv4; |
252 | if (src) | |
253 | nexthop->src.ipv4 = *src; | |
718e3744 | 254 | |
d62a17ae | 255 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 256 | |
d62a17ae | 257 | return nexthop; |
718e3744 | 258 | } |
259 | ||
d62a17ae | 260 | struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re, |
261 | struct in_addr *ipv4, | |
262 | struct in_addr *src, | |
4a7371e9 DS |
263 | ifindex_t ifindex, |
264 | vrf_id_t nh_vrf_id) | |
718e3744 | 265 | { |
d62a17ae | 266 | struct nexthop *nexthop; |
267 | struct interface *ifp; | |
718e3744 | 268 | |
d62a17ae | 269 | nexthop = nexthop_new(); |
4a7371e9 | 270 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 271 | nexthop->type = NEXTHOP_TYPE_IPV4_IFINDEX; |
272 | nexthop->gate.ipv4 = *ipv4; | |
273 | if (src) | |
274 | nexthop->src.ipv4 = *src; | |
275 | nexthop->ifindex = ifindex; | |
4a7371e9 | 276 | ifp = if_lookup_by_index(nexthop->ifindex, nh_vrf_id); |
d62a17ae | 277 | /*Pending: need to think if null ifp here is ok during bootup? |
278 | There was a crash because ifp here was coming to be NULL */ | |
279 | if (ifp) | |
2b83602b | 280 | if (connected_is_unnumbered(ifp)) |
d62a17ae | 281 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); |
718e3744 | 282 | |
d62a17ae | 283 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 284 | |
d62a17ae | 285 | return nexthop; |
718e3744 | 286 | } |
287 | ||
d62a17ae | 288 | struct nexthop *route_entry_nexthop_ipv6_add(struct route_entry *re, |
4a7371e9 DS |
289 | struct in6_addr *ipv6, |
290 | vrf_id_t nh_vrf_id) | |
718e3744 | 291 | { |
d62a17ae | 292 | struct nexthop *nexthop; |
718e3744 | 293 | |
d62a17ae | 294 | nexthop = nexthop_new(); |
4a7371e9 | 295 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 296 | nexthop->type = NEXTHOP_TYPE_IPV6; |
297 | nexthop->gate.ipv6 = *ipv6; | |
718e3744 | 298 | |
d62a17ae | 299 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 300 | |
d62a17ae | 301 | return nexthop; |
718e3744 | 302 | } |
303 | ||
d62a17ae | 304 | struct nexthop *route_entry_nexthop_ipv6_ifindex_add(struct route_entry *re, |
305 | struct in6_addr *ipv6, | |
4a7371e9 DS |
306 | ifindex_t ifindex, |
307 | vrf_id_t nh_vrf_id) | |
718e3744 | 308 | { |
d62a17ae | 309 | struct nexthop *nexthop; |
718e3744 | 310 | |
d62a17ae | 311 | nexthop = nexthop_new(); |
4a7371e9 | 312 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 313 | nexthop->type = NEXTHOP_TYPE_IPV6_IFINDEX; |
314 | nexthop->gate.ipv6 = *ipv6; | |
315 | nexthop->ifindex = ifindex; | |
718e3744 | 316 | |
d62a17ae | 317 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 318 | |
d62a17ae | 319 | return nexthop; |
718e3744 | 320 | } |
718e3744 | 321 | |
a8309422 | 322 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re, |
60466a63 | 323 | enum blackhole_type bh_type) |
595db7f1 | 324 | { |
d62a17ae | 325 | struct nexthop *nexthop; |
595db7f1 | 326 | |
d62a17ae | 327 | nexthop = nexthop_new(); |
4a7371e9 | 328 | nexthop->vrf_id = VRF_DEFAULT; |
d62a17ae | 329 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; |
a8309422 | 330 | nexthop->bh_type = bh_type; |
595db7f1 | 331 | |
d62a17ae | 332 | route_entry_nexthop_add(re, nexthop); |
595db7f1 | 333 | |
d62a17ae | 334 | return nexthop; |
595db7f1 | 335 | } |
336 | ||
d62a17ae | 337 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
338 | union g_addr *addr, struct route_node **rn_out) | |
339 | { | |
340 | struct prefix p; | |
341 | struct route_table *table; | |
342 | struct route_node *rn; | |
5f7a4718 | 343 | struct route_entry *match = NULL; |
d62a17ae | 344 | |
345 | /* Lookup table. */ | |
346 | table = zebra_vrf_table(afi, safi, vrf_id); | |
347 | if (!table) | |
348 | return 0; | |
349 | ||
350 | memset(&p, 0, sizeof(struct prefix)); | |
351 | p.family = afi; | |
352 | if (afi == AFI_IP) { | |
353 | p.u.prefix4 = addr->ipv4; | |
354 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
355 | } else { | |
356 | p.u.prefix6 = addr->ipv6; | |
357 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
358 | } | |
718e3744 | 359 | |
d62a17ae | 360 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 361 | |
d62a17ae | 362 | while (rn) { |
5f7a4718 DS |
363 | rib_dest_t *dest; |
364 | ||
d62a17ae | 365 | route_unlock_node(rn); |
718e3744 | 366 | |
5f7a4718 | 367 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
368 | if (dest && dest->selected_fib |
369 | && !CHECK_FLAG(dest->selected_fib->status, | |
370 | ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 371 | match = dest->selected_fib; |
718e3744 | 372 | |
d62a17ae | 373 | /* If there is no selected route or matched route is EGP, go up |
374 | tree. */ | |
375 | if (!match) { | |
376 | do { | |
377 | rn = rn->parent; | |
378 | } while (rn && rn->info == NULL); | |
379 | if (rn) | |
380 | route_lock_node(rn); | |
381 | } else { | |
382 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
677c1dd5 DS |
383 | if (!CHECK_FLAG(match->status, |
384 | ROUTE_ENTRY_INSTALLED)) | |
d62a17ae | 385 | return NULL; |
386 | } | |
387 | ||
388 | if (rn_out) | |
389 | *rn_out = rn; | |
390 | return match; | |
391 | } | |
16814f96 | 392 | } |
d62a17ae | 393 | return NULL; |
394 | } | |
395 | ||
396 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
397 | struct in_addr addr, | |
398 | struct route_node **rn_out) | |
399 | { | |
400 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
401 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
402 | union g_addr gaddr = {.ipv4 = addr}; | |
403 | ||
526052fb | 404 | switch (zrouter.ipv4_multicast_mode) { |
d62a17ae | 405 | case MCAST_MRIB_ONLY: |
406 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
407 | rn_out); | |
408 | case MCAST_URIB_ONLY: | |
409 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
410 | case MCAST_NO_CONFIG: | |
411 | case MCAST_MIX_MRIB_FIRST: | |
412 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
413 | &m_rn); | |
414 | if (!mre) | |
415 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
416 | &gaddr, &u_rn); | |
417 | break; | |
418 | case MCAST_MIX_DISTANCE: | |
419 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
420 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
421 | if (mre && ure) | |
422 | re = ure->distance < mre->distance ? ure : mre; | |
423 | else if (mre) | |
424 | re = mre; | |
425 | else if (ure) | |
426 | re = ure; | |
427 | break; | |
428 | case MCAST_MIX_PFXLEN: | |
429 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
430 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
431 | if (mre && ure) | |
432 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
433 | else if (mre) | |
434 | re = mre; | |
435 | else if (ure) | |
436 | re = ure; | |
437 | break; | |
718e3744 | 438 | } |
33550aa8 | 439 | |
d62a17ae | 440 | if (rn_out) |
441 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 442 | |
d62a17ae | 443 | if (IS_ZEBRA_DEBUG_RIB) { |
444 | char buf[BUFSIZ]; | |
445 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 446 | |
32391aff DS |
447 | zlog_debug("%s: %s: vrf: %u found %s, using %s", |
448 | __func__, buf, vrf_id, | |
d62a17ae | 449 | mre ? (ure ? "MRIB+URIB" : "MRIB") |
450 | : ure ? "URIB" : "nothing", | |
451 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
452 | } | |
453 | return re; | |
4623d897 DL |
454 | } |
455 | ||
d62a17ae | 456 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 457 | { |
d62a17ae | 458 | struct route_table *table; |
459 | struct route_node *rn; | |
5f7a4718 | 460 | struct route_entry *match = NULL; |
5f7a4718 | 461 | rib_dest_t *dest; |
718e3744 | 462 | |
d62a17ae | 463 | /* Lookup table. */ |
464 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
465 | if (!table) | |
466 | return 0; | |
718e3744 | 467 | |
d62a17ae | 468 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 469 | |
d62a17ae | 470 | /* No route for this prefix. */ |
471 | if (!rn) | |
472 | return NULL; | |
718e3744 | 473 | |
d62a17ae | 474 | /* Unlock node. */ |
475 | route_unlock_node(rn); | |
5f7a4718 | 476 | dest = rib_dest_from_rnode(rn); |
718e3744 | 477 | |
996c9314 LB |
478 | if (dest && dest->selected_fib |
479 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 480 | match = dest->selected_fib; |
718e3744 | 481 | |
d62a17ae | 482 | if (!match) |
483 | return NULL; | |
718e3744 | 484 | |
d62a17ae | 485 | if (match->type == ZEBRA_ROUTE_CONNECT) |
486 | return match; | |
f9e1b38e | 487 | |
677c1dd5 DS |
488 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_INSTALLED)) |
489 | return match; | |
718e3744 | 490 | |
d62a17ae | 491 | return NULL; |
718e3744 | 492 | } |
493 | ||
a64448ba DS |
494 | /* |
495 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
496 | * (nexthops) must have a label. | |
497 | */ | |
d62a17ae | 498 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 499 | { |
d62a17ae | 500 | struct nexthop *nexthop = NULL; |
a64448ba | 501 | |
d62a17ae | 502 | if (re->type != ZEBRA_ROUTE_BGP) |
503 | return 0; | |
a64448ba | 504 | |
7ee30f28 | 505 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d62a17ae | 506 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) |
507 | return 0; | |
6b0655a2 | 508 | |
d62a17ae | 509 | return 1; |
a64448ba | 510 | } |
718e3744 | 511 | |
6ae24471 DS |
512 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
513 | * is only used for IPv4. | |
514 | */ | |
0c555cc6 DS |
515 | void rib_install_kernel(struct route_node *rn, struct route_entry *re, |
516 | struct route_entry *old) | |
718e3744 | 517 | { |
d62a17ae | 518 | struct nexthop *nexthop; |
519 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 520 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
97f5b441 MS |
521 | const struct prefix *p, *src_p; |
522 | enum zebra_dplane_result ret; | |
523 | ||
524 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
718e3744 | 525 | |
d62a17ae | 526 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 527 | |
d62a17ae | 528 | if (info->safi != SAFI_UNICAST) { |
7ee30f28 | 529 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d62a17ae | 530 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 531 | return; |
25b9cb0c DL |
532 | } else { |
533 | struct nexthop *prev; | |
534 | ||
7ee30f28 | 535 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
25b9cb0c | 536 | UNSET_FLAG (nexthop->flags, NEXTHOP_FLAG_DUPLICATE); |
7ee30f28 | 537 | for (ALL_NEXTHOPS(re->ng, prev)) { |
25b9cb0c DL |
538 | if (prev == nexthop) |
539 | break; | |
996c9314 LB |
540 | if (nexthop_same_firsthop(nexthop, prev)) { |
541 | SET_FLAG(nexthop->flags, | |
542 | NEXTHOP_FLAG_DUPLICATE); | |
25b9cb0c DL |
543 | break; |
544 | } | |
545 | } | |
546 | } | |
d62a17ae | 547 | } |
718e3744 | 548 | |
2063a814 DS |
549 | /* |
550 | * If this is a replace to a new RE let the originator of the RE | |
551 | * know that they've lost | |
552 | */ | |
9a9f8926 | 553 | if (old && (old != re) && (old->type != re->type)) |
28610f7e | 554 | zsend_route_notify_owner(old, p, ZAPI_ROUTE_BETTER_ADMIN_WON); |
25b9cb0c | 555 | |
97f5b441 MS |
556 | /* Update fib selection */ |
557 | dest->selected_fib = re; | |
558 | ||
d62a17ae | 559 | /* |
560 | * Make sure we update the FPM any time we send new information to | |
561 | * the kernel. | |
562 | */ | |
563 | hook_call(rib_update, rn, "installing in kernel"); | |
97f5b441 MS |
564 | |
565 | /* Send add or update */ | |
8263d1d0 | 566 | if (old) |
97f5b441 | 567 | ret = dplane_route_update(rn, re, old); |
5709131c | 568 | else |
97f5b441 | 569 | ret = dplane_route_add(rn, re); |
97f5b441 MS |
570 | |
571 | switch (ret) { | |
ea1c14f6 | 572 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
60f98b23 | 573 | SET_FLAG(re->status, ROUTE_ENTRY_QUEUED); |
efe6c026 MS |
574 | |
575 | if (old) { | |
60f98b23 | 576 | SET_FLAG(old->status, ROUTE_ENTRY_QUEUED); |
efe6c026 MS |
577 | |
578 | /* Free old FIB nexthop group */ | |
579 | if (old->fib_ng.nexthop) { | |
580 | nexthops_free(old->fib_ng.nexthop); | |
581 | old->fib_ng.nexthop = NULL; | |
582 | } | |
583 | ||
584 | if (!RIB_SYSTEM_ROUTE(old)) { | |
585 | /* Clear old route's FIB flags */ | |
586 | for (ALL_NEXTHOPS(old->ng, nexthop)) { | |
587 | UNSET_FLAG(nexthop->flags, | |
588 | NEXTHOP_FLAG_FIB); | |
589 | } | |
590 | } | |
591 | } | |
592 | ||
97f5b441 MS |
593 | if (zvrf) |
594 | zvrf->installs_queued++; | |
1e885672 | 595 | break; |
ea1c14f6 | 596 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
597 | { |
598 | char str[SRCDEST2STR_BUFFER]; | |
599 | ||
600 | srcdest_rnode2str(rn, str, sizeof(str)); | |
601 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
602 | "%u:%s: Failed to enqueue dataplane install", | |
603 | re->vrf_id, str); | |
1e885672 | 604 | break; |
97f5b441 | 605 | } |
ea1c14f6 | 606 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
97f5b441 MS |
607 | if (zvrf) |
608 | zvrf->installs++; | |
1e885672 DS |
609 | break; |
610 | } | |
d62a17ae | 611 | |
0c555cc6 | 612 | return; |
718e3744 | 613 | } |
614 | ||
615 | /* Uninstall the route from kernel. */ | |
0c555cc6 | 616 | void rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 617 | { |
d62a17ae | 618 | struct nexthop *nexthop; |
619 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 620 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
05737783 | 621 | |
d62a17ae | 622 | if (info->safi != SAFI_UNICAST) { |
677c1dd5 | 623 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
7ee30f28 | 624 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d6792f9d | 625 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 626 | return; |
d62a17ae | 627 | } |
416ec78d | 628 | |
d62a17ae | 629 | /* |
630 | * Make sure we update the FPM any time we send new information to | |
97f5b441 | 631 | * the dataplane. |
d62a17ae | 632 | */ |
633 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
97f5b441 MS |
634 | |
635 | switch (dplane_route_delete(rn, re)) { | |
ea1c14f6 | 636 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
97f5b441 MS |
637 | if (zvrf) |
638 | zvrf->removals_queued++; | |
1e885672 | 639 | break; |
ea1c14f6 | 640 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
641 | { |
642 | char str[SRCDEST2STR_BUFFER]; | |
643 | ||
644 | srcdest_rnode2str(rn, str, sizeof(str)); | |
645 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
646 | "%u:%s: Failed to enqueue dataplane uninstall", | |
647 | re->vrf_id, str); | |
1e885672 | 648 | break; |
97f5b441 | 649 | } |
ea1c14f6 | 650 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
1e885672 DS |
651 | if (zvrf) |
652 | zvrf->removals++; | |
653 | break; | |
654 | } | |
718e3744 | 655 | |
0c555cc6 | 656 | return; |
718e3744 | 657 | } |
658 | ||
659 | /* Uninstall the route from kernel. */ | |
d62a17ae | 660 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 661 | { |
d62a17ae | 662 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
5f7a4718 | 663 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
97f5b441 | 664 | struct nexthop *nexthop; |
416ec78d | 665 | |
5f7a4718 | 666 | if (dest && dest->selected_fib == re) { |
d62a17ae | 667 | if (info->safi == SAFI_UNICAST) |
668 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 669 | |
d62a17ae | 670 | /* If labeled-unicast route, uninstall transit LSP. */ |
671 | if (zebra_rib_labeled_unicast(re)) | |
672 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
7cdb1a84 | 673 | |
3cdba47a | 674 | rib_uninstall_kernel(rn, re); |
97f5b441 MS |
675 | |
676 | dest->selected_fib = NULL; | |
677 | ||
efe6c026 MS |
678 | /* Free FIB nexthop group, if present */ |
679 | if (re->fib_ng.nexthop) { | |
680 | nexthops_free(re->fib_ng.nexthop); | |
681 | re->fib_ng.nexthop = NULL; | |
682 | } | |
683 | ||
97f5b441 MS |
684 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
685 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
d62a17ae | 686 | } |
446bb95e | 687 | |
d62a17ae | 688 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
86391e56 MS |
689 | const struct prefix *p, *src_p; |
690 | ||
d62a17ae | 691 | srcdest_rnode_prefixes(rn, &p, &src_p); |
05737783 | 692 | |
d62a17ae | 693 | redistribute_delete(p, src_p, re); |
694 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); | |
695 | } | |
718e3744 | 696 | } |
697 | ||
9fd92e3c AS |
698 | /* |
699 | * rib_can_delete_dest | |
700 | * | |
2951a7a4 | 701 | * Returns true if the given dest can be deleted from the table. |
9fd92e3c | 702 | */ |
d62a17ae | 703 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 704 | { |
045207e2 | 705 | if (re_list_first(&dest->routes)) { |
d62a17ae | 706 | return 0; |
707 | } | |
9fd92e3c | 708 | |
c86ba6c2 DS |
709 | /* |
710 | * Unresolved rnh's are stored on the default route's list | |
711 | * | |
712 | * dest->rnode can also be the source prefix node in an | |
713 | * ipv6 sourcedest table. Fortunately the prefix of a | |
714 | * source prefix node can never be the default prefix. | |
715 | */ | |
716 | if (is_default_prefix(&dest->rnode->p)) | |
717 | return 0; | |
718 | ||
d62a17ae | 719 | /* |
720 | * Don't delete the dest if we have to update the FPM about this | |
721 | * prefix. | |
722 | */ | |
723 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
724 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
725 | return 0; | |
5adc2528 | 726 | |
d62a17ae | 727 | return 1; |
9fd92e3c AS |
728 | } |
729 | ||
699dae23 DS |
730 | void zebra_rib_evaluate_rn_nexthops(struct route_node *rn, uint32_t seq) |
731 | { | |
732 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
699dae23 DS |
733 | struct rnh *rnh; |
734 | ||
735 | /* | |
50872b08 | 736 | * We are storing the rnh's associated withb |
699dae23 DS |
737 | * the tracked nexthop as a list of the rn's. |
738 | * Unresolved rnh's are placed at the top | |
739 | * of the tree list.( 0.0.0.0/0 for v4 and 0::0/0 for v6 ) | |
740 | * As such for each rn we need to walk up the tree | |
741 | * and see if any rnh's need to see if they | |
742 | * would match a more specific route | |
743 | */ | |
744 | while (rn) { | |
50872b08 DS |
745 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
746 | char buf[PREFIX_STRLEN]; | |
747 | ||
0a7be328 | 748 | zlog_debug("%s: %s Being examined for Nexthop Tracking Count: %zd", |
50872b08 | 749 | __PRETTY_FUNCTION__, |
0a7be328 DS |
750 | srcdest_rnode2str(rn, buf, sizeof(buf)), |
751 | dest ? rnh_list_count(&dest->nht) : 0); | |
50872b08 | 752 | } |
699dae23 DS |
753 | if (!dest) { |
754 | rn = rn->parent; | |
755 | if (rn) | |
756 | dest = rib_dest_from_rnode(rn); | |
757 | continue; | |
758 | } | |
759 | /* | |
760 | * If we have any rnh's stored in the nht list | |
761 | * then we know that this route node was used for | |
762 | * nht resolution and as such we need to call the | |
763 | * nexthop tracking evaluation code | |
764 | */ | |
739c9c90 | 765 | frr_each_safe(rnh_list, &dest->nht, rnh) { |
699dae23 DS |
766 | struct zebra_vrf *zvrf = |
767 | zebra_vrf_lookup_by_id(rnh->vrf_id); | |
768 | struct prefix *p = &rnh->node->p; | |
769 | ||
50872b08 | 770 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
699dae23 DS |
771 | char buf1[PREFIX_STRLEN]; |
772 | char buf2[PREFIX_STRLEN]; | |
773 | ||
0a7be328 | 774 | zlog_debug("%u:%s has Nexthop(%s) Type: %s depending on it, evaluating %u:%u", |
699dae23 | 775 | zvrf->vrf->vrf_id, |
50872b08 | 776 | srcdest_rnode2str(rn, buf1, |
699dae23 DS |
777 | sizeof(buf1)), |
778 | prefix2str(p, buf2, sizeof(buf2)), | |
0a7be328 | 779 | rnh_type2str(rnh->type), |
699dae23 DS |
780 | seq, rnh->seqno); |
781 | } | |
782 | ||
783 | /* | |
784 | * If we have evaluated this node on this pass | |
785 | * already, due to following the tree up | |
786 | * then we know that we can move onto the next | |
787 | * rnh to process. | |
788 | * | |
789 | * Additionally we call zebra_evaluate_rnh | |
790 | * when we gc the dest. In this case we know | |
791 | * that there must be no other re's where | |
792 | * we were originally as such we know that | |
793 | * that sequence number is ok to respect. | |
794 | */ | |
50872b08 DS |
795 | if (rnh->seqno == seq) { |
796 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) | |
797 | zlog_debug( | |
798 | "\tNode processed and moved already"); | |
699dae23 | 799 | continue; |
50872b08 | 800 | } |
699dae23 DS |
801 | |
802 | rnh->seqno = seq; | |
803 | zebra_evaluate_rnh(zvrf, family2afi(p->family), 0, | |
804 | rnh->type, p); | |
805 | } | |
806 | ||
807 | rn = rn->parent; | |
808 | if (rn) | |
809 | dest = rib_dest_from_rnode(rn); | |
810 | } | |
811 | } | |
812 | ||
9fd92e3c AS |
813 | /* |
814 | * rib_gc_dest | |
815 | * | |
816 | * Garbage collect the rib dest corresponding to the given route node | |
817 | * if appropriate. | |
818 | * | |
2951a7a4 | 819 | * Returns true if the dest was deleted, false otherwise. |
9fd92e3c | 820 | */ |
d62a17ae | 821 | int rib_gc_dest(struct route_node *rn) |
9fd92e3c | 822 | { |
d62a17ae | 823 | rib_dest_t *dest; |
9fd92e3c | 824 | |
d62a17ae | 825 | dest = rib_dest_from_rnode(rn); |
826 | if (!dest) | |
827 | return 0; | |
9fd92e3c | 828 | |
d62a17ae | 829 | if (!rib_can_delete_dest(dest)) |
830 | return 0; | |
9fd92e3c | 831 | |
c9abf558 DS |
832 | if (IS_ZEBRA_DEBUG_RIB) { |
833 | struct zebra_vrf *zvrf; | |
834 | ||
835 | zvrf = rib_dest_vrf(dest); | |
d62a17ae | 836 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); |
c9abf558 | 837 | } |
9fd92e3c | 838 | |
699dae23 DS |
839 | zebra_rib_evaluate_rn_nexthops(rn, zebra_router_get_next_sequence()); |
840 | ||
d62a17ae | 841 | dest->rnode = NULL; |
aa57abfb | 842 | rnh_list_fini(&dest->nht); |
d62a17ae | 843 | XFREE(MTYPE_RIB_DEST, dest); |
844 | rn->info = NULL; | |
9fd92e3c | 845 | |
d62a17ae | 846 | /* |
847 | * Release the one reference that we keep on the route node. | |
848 | */ | |
849 | route_unlock_node(rn); | |
850 | return 1; | |
9fd92e3c AS |
851 | } |
852 | ||
d62a17ae | 853 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
854 | struct route_entry *new) | |
3e5c6e00 | 855 | { |
d62a17ae | 856 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 857 | |
d62a17ae | 858 | /* Update real nexthop. This may actually determine if nexthop is active |
859 | * or not. */ | |
6883bf8d | 860 | if (!nexthop_group_active_nexthop_num(&new->ng)) { |
d62a17ae | 861 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
862 | return; | |
863 | } | |
3e5c6e00 | 864 | |
d62a17ae | 865 | if (IS_ZEBRA_DEBUG_RIB) { |
866 | char buf[SRCDEST2STR_BUFFER]; | |
867 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
868 | zlog_debug("%u:%s: Adding route rn %p, re %p (%s)", |
869 | zvrf_id(zvrf), buf, rn, new, | |
870 | zebra_route_string(new->type)); | |
d62a17ae | 871 | } |
3e5c6e00 | 872 | |
d62a17ae | 873 | /* If labeled-unicast route, install transit LSP. */ |
874 | if (zebra_rib_labeled_unicast(new)) | |
875 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 876 | |
3cdba47a | 877 | rib_install_kernel(rn, new, NULL); |
3e5c6e00 | 878 | |
d62a17ae | 879 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 880 | } |
881 | ||
d62a17ae | 882 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
883 | struct route_entry *old) | |
3e5c6e00 | 884 | { |
d62a17ae | 885 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 886 | |
d62a17ae | 887 | /* Uninstall from kernel. */ |
888 | if (IS_ZEBRA_DEBUG_RIB) { | |
889 | char buf[SRCDEST2STR_BUFFER]; | |
890 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
891 | zlog_debug("%u:%s: Deleting route rn %p, re %p (%s)", |
892 | zvrf_id(zvrf), buf, rn, old, | |
893 | zebra_route_string(old->type)); | |
d62a17ae | 894 | } |
3e5c6e00 | 895 | |
d62a17ae | 896 | /* If labeled-unicast route, uninstall transit LSP. */ |
897 | if (zebra_rib_labeled_unicast(old)) | |
898 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
899 | ||
3cdba47a | 900 | rib_uninstall_kernel(rn, old); |
d62a17ae | 901 | |
902 | /* Update nexthop for route, reset changed flag. */ | |
ef57f35f DL |
903 | /* Note: this code also handles the Linux case when an interface goes |
904 | * down, causing the kernel to delete routes without sending DELROUTE | |
905 | * notifications | |
906 | */ | |
6883bf8d | 907 | if (RIB_KERNEL_ROUTE(old)) |
212df1de PG |
908 | SET_FLAG(old->status, ROUTE_ENTRY_REMOVED); |
909 | else | |
910 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 911 | } |
912 | ||
913 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
914 | struct route_node *rn, | |
915 | struct route_entry *old, | |
916 | struct route_entry *new) | |
917 | { | |
d62a17ae | 918 | int nh_active = 0; |
d62a17ae | 919 | |
920 | /* | |
921 | * We have to install or update if a new route has been selected or | |
922 | * something has changed. | |
923 | */ | |
924 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
925 | hook_call(rib_update, rn, "updating existing route"); | |
926 | ||
927 | /* Update the nexthop; we could determine here that nexthop is | |
928 | * inactive. */ | |
6883bf8d | 929 | if (nexthop_group_active_nexthop_num(&new->ng)) |
d62a17ae | 930 | nh_active = 1; |
931 | ||
932 | /* If nexthop is active, install the selected route, if | |
933 | * appropriate. If | |
934 | * the install succeeds, cleanup flags for prior route, if | |
935 | * different from | |
936 | * newly selected. | |
937 | */ | |
938 | if (nh_active) { | |
939 | if (IS_ZEBRA_DEBUG_RIB) { | |
940 | char buf[SRCDEST2STR_BUFFER]; | |
941 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
942 | if (new != old) | |
943 | zlog_debug( | |
2da33d6b | 944 | "%u:%s: Updating route rn %p, re %p (%s) old %p (%s)", |
d62a17ae | 945 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
946 | zebra_route_string(new->type), |
947 | old, | |
948 | zebra_route_string(old->type)); | |
d62a17ae | 949 | else |
950 | zlog_debug( | |
2da33d6b | 951 | "%u:%s: Updating route rn %p, re %p (%s)", |
d62a17ae | 952 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 953 | zebra_route_string(new->type)); |
d62a17ae | 954 | } |
955 | ||
956 | /* If labeled-unicast route, uninstall transit LSP. */ | |
957 | if (zebra_rib_labeled_unicast(old)) | |
958 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
959 | ||
3cdba47a DS |
960 | /* |
961 | * Non-system route should be installed. | |
962 | * If labeled-unicast route, install transit | |
963 | * LSP. | |
964 | */ | |
965 | if (zebra_rib_labeled_unicast(new)) | |
966 | zebra_mpls_lsp_install(zvrf, rn, new); | |
d62a17ae | 967 | |
3cdba47a | 968 | rib_install_kernel(rn, new, old); |
d62a17ae | 969 | } |
a64448ba | 970 | |
d62a17ae | 971 | /* |
972 | * If nexthop for selected route is not active or install | |
973 | * failed, we | |
974 | * may need to uninstall and delete for redistribution. | |
975 | */ | |
ed216282 | 976 | if (!nh_active) { |
d62a17ae | 977 | if (IS_ZEBRA_DEBUG_RIB) { |
978 | char buf[SRCDEST2STR_BUFFER]; | |
979 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
980 | if (new != old) | |
981 | zlog_debug( | |
2da33d6b | 982 | "%u:%s: Deleting route rn %p, re %p (%s) old %p (%s) - nexthop inactive", |
d62a17ae | 983 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
984 | zebra_route_string(new->type), |
985 | old, | |
986 | zebra_route_string(old->type)); | |
d62a17ae | 987 | else |
988 | zlog_debug( | |
2da33d6b | 989 | "%u:%s: Deleting route rn %p, re %p (%s) - nexthop inactive", |
d62a17ae | 990 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 991 | zebra_route_string(new->type)); |
d62a17ae | 992 | } |
993 | ||
994 | /* If labeled-unicast route, uninstall transit LSP. */ | |
995 | if (zebra_rib_labeled_unicast(old)) | |
996 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
997 | ||
3cdba47a | 998 | rib_uninstall_kernel(rn, old); |
d62a17ae | 999 | } |
1000 | } else { | |
1001 | /* | |
1002 | * Same route selected; check if in the FIB and if not, | |
3cdba47a DS |
1003 | * re-install. This is housekeeping code to deal with |
1004 | * race conditions in kernel with linux netlink reporting | |
1005 | * interface up before IPv4 or IPv6 protocol is ready | |
d62a17ae | 1006 | * to add routes. |
1007 | */ | |
3cdba47a DS |
1008 | if (!CHECK_FLAG(new->status, ROUTE_ENTRY_INSTALLED) || |
1009 | RIB_SYSTEM_ROUTE(new)) | |
677c1dd5 | 1010 | rib_install_kernel(rn, new, NULL); |
d62a17ae | 1011 | } |
1012 | ||
1013 | /* Update prior route. */ | |
6883bf8d | 1014 | if (new != old) |
d62a17ae | 1015 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1016 | |
d62a17ae | 1017 | /* Clear changed flag. */ |
1018 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 1019 | } |
1020 | ||
d62a17ae | 1021 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
1022 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1023 | struct route_entry *alternate) | |
1024 | { | |
1025 | if (current == NULL) | |
1026 | return alternate; | |
1027 | ||
1028 | /* filter route selection in following order: | |
1029 | * - connected beats other types | |
fec4ca19 | 1030 | * - if both connected, loopback or vrf wins |
d62a17ae | 1031 | * - lower distance beats higher |
1032 | * - lower metric beats higher for equal distance | |
1033 | * - last, hence oldest, route wins tie break. | |
1034 | */ | |
1035 | ||
fec4ca19 DS |
1036 | /* Connected routes. Check to see if either are a vrf |
1037 | * or loopback interface. If not, pick the last connected | |
d62a17ae | 1038 | * route of the set of lowest metric connected routes. |
1039 | */ | |
1040 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
fec4ca19 DS |
1041 | if (current->type != ZEBRA_ROUTE_CONNECT) |
1042 | return alternate; | |
1043 | ||
1044 | /* both are connected. are either loop or vrf? */ | |
1045 | struct nexthop *nexthop = NULL; | |
1046 | ||
1047 | for (ALL_NEXTHOPS(alternate->ng, nexthop)) { | |
1048 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1049 | nexthop->ifindex, alternate->vrf_id))) | |
1050 | return alternate; | |
1051 | } | |
1052 | ||
1053 | for (ALL_NEXTHOPS(current->ng, nexthop)) { | |
1054 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1055 | nexthop->ifindex, current->vrf_id))) | |
1056 | return current; | |
1057 | } | |
1058 | ||
1059 | /* Neither are loop or vrf so pick best metric */ | |
1060 | if (alternate->metric <= current->metric) | |
d62a17ae | 1061 | return alternate; |
1062 | ||
1063 | return current; | |
1064 | } | |
3e5c6e00 | 1065 | |
d62a17ae | 1066 | if (current->type == ZEBRA_ROUTE_CONNECT) |
1067 | return current; | |
3e5c6e00 | 1068 | |
d62a17ae | 1069 | /* higher distance loses */ |
1070 | if (alternate->distance < current->distance) | |
1071 | return alternate; | |
1072 | if (current->distance < alternate->distance) | |
1073 | return current; | |
3e5c6e00 | 1074 | |
d62a17ae | 1075 | /* metric tie-breaks equal distance */ |
1076 | if (alternate->metric <= current->metric) | |
1077 | return alternate; | |
3e5c6e00 | 1078 | |
d62a17ae | 1079 | return current; |
3e5c6e00 | 1080 | } |
1081 | ||
d62a17ae | 1082 | /* Core function for processing routing information base. */ |
1083 | static void rib_process(struct route_node *rn) | |
1084 | { | |
1085 | struct route_entry *re; | |
1086 | struct route_entry *next; | |
1087 | struct route_entry *old_selected = NULL; | |
1088 | struct route_entry *new_selected = NULL; | |
1089 | struct route_entry *old_fib = NULL; | |
1090 | struct route_entry *new_fib = NULL; | |
1091 | struct route_entry *best = NULL; | |
1092 | char buf[SRCDEST2STR_BUFFER]; | |
1093 | rib_dest_t *dest; | |
1094 | struct zebra_vrf *zvrf = NULL; | |
86391e56 MS |
1095 | const struct prefix *p, *src_p; |
1096 | ||
d62a17ae | 1097 | srcdest_rnode_prefixes(rn, &p, &src_p); |
1098 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1099 | ||
1100 | assert(rn); | |
1101 | ||
1102 | dest = rib_dest_from_rnode(rn); | |
1103 | if (dest) { | |
1104 | zvrf = rib_dest_vrf(dest); | |
1105 | vrf_id = zvrf_id(zvrf); | |
1106 | } | |
bab85d4f | 1107 | |
d62a17ae | 1108 | if (IS_ZEBRA_DEBUG_RIB) |
1109 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1110 | |
d62a17ae | 1111 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1112 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1113 | |
607425e5 DS |
1114 | /* |
1115 | * we can have rn's that have a NULL info pointer | |
1116 | * (dest). As such let's not let the deref happen | |
1117 | * additionally we know RNODE_FOREACH_RE_SAFE | |
1118 | * will not iterate so we are ok. | |
1119 | */ | |
1120 | if (dest) | |
1121 | old_fib = dest->selected_fib; | |
5f7a4718 | 1122 | |
a2addae8 | 1123 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1124 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1125 | zlog_debug( | |
2da33d6b DS |
1126 | "%u:%s: Examine re %p (%s) status %x flags %x dist %d metric %d", |
1127 | vrf_id, buf, re, zebra_route_string(re->type), | |
1128 | re->status, re->flags, re->distance, | |
1129 | re->metric); | |
d62a17ae | 1130 | |
1131 | UNSET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1132 | ||
1133 | /* Currently selected re. */ | |
1134 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1135 | assert(old_selected == NULL); | |
1136 | old_selected = re; | |
1137 | } | |
bab85d4f | 1138 | |
d62a17ae | 1139 | /* Skip deleted entries from selection */ |
1140 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1141 | continue; | |
1142 | ||
1143 | /* Skip unreachable nexthop. */ | |
1144 | /* This first call to nexthop_active_update is merely to | |
6883bf8d DS |
1145 | * determine if there's any change to nexthops associated |
1146 | * with this RIB entry. Now, rib_process() can be invoked due | |
1147 | * to an external event such as link down or due to | |
1148 | * next-hop-tracking evaluation. In the latter case, | |
d62a17ae | 1149 | * a decision has already been made that the NHs have changed. |
6883bf8d DS |
1150 | * So, no need to invoke a potentially expensive call again. |
1151 | * Further, since the change might be in a recursive NH which | |
1152 | * is not caught in the nexthop_active_update() code. Thus, we | |
1153 | * might miss changes to recursive NHs. | |
d62a17ae | 1154 | */ |
6883bf8d | 1155 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) |
99eabcec | 1156 | && !nexthop_active_update(rn, re)) { |
d62a17ae | 1157 | if (re->type == ZEBRA_ROUTE_TABLE) { |
1158 | /* XXX: HERE BE DRAGONS!!!!! | |
1159 | * In all honesty, I have not yet figured out | |
6883bf8d DS |
1160 | * what this part does or why the |
1161 | * ROUTE_ENTRY_CHANGED test above is correct | |
d62a17ae | 1162 | * or why we need to delete a route here, and |
6883bf8d DS |
1163 | * also not whether this concerns both selected |
1164 | * and fib route, or only selected | |
1165 | * or only fib | |
1166 | * | |
1167 | * This entry was denied by the 'ip protocol | |
1168 | * table' route-map, we need to delete it */ | |
d62a17ae | 1169 | if (re != old_selected) { |
1170 | if (IS_ZEBRA_DEBUG_RIB) | |
1171 | zlog_debug( | |
32391aff | 1172 | "%s: %u:%s: imported via import-table but denied " |
d62a17ae | 1173 | "by the ip protocol table route-map", |
32391aff | 1174 | __func__, vrf_id, buf); |
d62a17ae | 1175 | rib_unlink(rn, re); |
1176 | } else | |
1177 | SET_FLAG(re->status, | |
1178 | ROUTE_ENTRY_REMOVED); | |
1179 | } | |
1180 | ||
1181 | continue; | |
1182 | } | |
bab85d4f | 1183 | |
d62a17ae | 1184 | /* Infinite distance. */ |
1185 | if (re->distance == DISTANCE_INFINITY) { | |
1186 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1187 | continue; | |
1188 | } | |
bab85d4f | 1189 | |
d62a17ae | 1190 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1191 | best = rib_choose_best(new_fib, re); | |
1192 | if (new_fib && best != new_fib) | |
1193 | UNSET_FLAG(new_fib->status, | |
1194 | ROUTE_ENTRY_CHANGED); | |
1195 | new_fib = best; | |
1196 | } else { | |
1197 | best = rib_choose_best(new_selected, re); | |
1198 | if (new_selected && best != new_selected) | |
1199 | UNSET_FLAG(new_selected->status, | |
1200 | ROUTE_ENTRY_CHANGED); | |
1201 | new_selected = best; | |
1202 | } | |
1203 | if (best != re) | |
1204 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1205 | } /* RNODE_FOREACH_RE */ | |
1206 | ||
1207 | /* If no FIB override route, use the selected route also for FIB */ | |
1208 | if (new_fib == NULL) | |
1209 | new_fib = new_selected; | |
1210 | ||
1211 | /* After the cycle is finished, the following pointers will be set: | |
1212 | * old_selected --- RE entry currently having SELECTED | |
1213 | * new_selected --- RE entry that is newly SELECTED | |
1214 | * old_fib --- RE entry currently in kernel FIB | |
1215 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1216 | * | |
1217 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1218 | * the zclients. new_fib (which can be new_selected) will be installed | |
1219 | * in kernel. | |
1220 | */ | |
1221 | ||
1222 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1223 | zlog_debug( | |
1224 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1225 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1226 | (void *)old_fib, (void *)new_fib); | |
1227 | } | |
446bb95e | 1228 | |
d62a17ae | 1229 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1230 | * fib == selected */ | |
9d303b37 DL |
1231 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, |
1232 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1233 | |
1234 | /* Update fib according to selection results */ | |
1235 | if (new_fib && old_fib) | |
1236 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1237 | else if (new_fib) | |
1238 | rib_process_add_fib(zvrf, rn, new_fib); | |
1239 | else if (old_fib) | |
1240 | rib_process_del_fib(zvrf, rn, old_fib); | |
1241 | ||
8cb41cd6 | 1242 | /* Update SELECTED entry */ |
d62a17ae | 1243 | if (old_selected != new_selected || selected_changed) { |
93bdadae | 1244 | |
6883bf8d | 1245 | if (new_selected && new_selected != new_fib) |
d62a17ae | 1246 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); |
41ec9222 | 1247 | |
3cdba47a | 1248 | if (new_selected) |
5af4b346 MS |
1249 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); |
1250 | ||
d62a17ae | 1251 | if (old_selected) { |
1252 | if (!new_selected) | |
1253 | redistribute_delete(p, src_p, old_selected); | |
1254 | if (old_selected != new_selected) | |
1255 | UNSET_FLAG(old_selected->flags, | |
1256 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1257 | } |
d62a17ae | 1258 | } |
3e5c6e00 | 1259 | |
d62a17ae | 1260 | /* Remove all RE entries queued for removal */ |
a2addae8 | 1261 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1262 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { |
1263 | if (IS_ZEBRA_DEBUG_RIB) { | |
1264 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1265 | (void *)rn, (void *)re); | |
1266 | } | |
1267 | rib_unlink(rn, re); | |
1268 | } | |
1269 | } | |
4d38fdb4 | 1270 | |
d62a17ae | 1271 | /* |
1272 | * Check if the dest can be deleted now. | |
1273 | */ | |
1274 | rib_gc_dest(rn); | |
e96f9203 DO |
1275 | } |
1276 | ||
a1494c25 DS |
1277 | static void zebra_rib_evaluate_mpls(struct route_node *rn) |
1278 | { | |
1279 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
1280 | struct zebra_vrf *zvrf = vrf_info_lookup(VRF_DEFAULT); | |
1281 | ||
1282 | if (!dest) | |
1283 | return; | |
1284 | ||
1285 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_LSPS)) { | |
1286 | if (IS_ZEBRA_DEBUG_MPLS) | |
1287 | zlog_debug( | |
1288 | "%u: Scheduling all LSPs upon RIB completion", | |
1289 | zvrf_id(zvrf)); | |
1290 | zebra_mpls_lsp_schedule(zvrf); | |
1291 | mpls_unmark_lsps_for_processing(rn); | |
1292 | } | |
1293 | } | |
1294 | ||
e5ac2adf MS |
1295 | /* |
1296 | * Utility to match route with dplane context data | |
1297 | */ | |
1298 | static bool rib_route_match_ctx(const struct route_entry *re, | |
25779064 MS |
1299 | const struct zebra_dplane_ctx *ctx, |
1300 | bool is_update) | |
e5ac2adf MS |
1301 | { |
1302 | bool result = false; | |
1303 | ||
1304 | if (is_update) { | |
1305 | /* | |
1306 | * In 'update' case, we test info about the 'previous' or | |
1307 | * 'old' route | |
1308 | */ | |
1309 | if ((re->type == dplane_ctx_get_old_type(ctx)) && | |
1310 | (re->instance == dplane_ctx_get_old_instance(ctx))) { | |
1311 | result = true; | |
1312 | ||
1313 | /* TODO -- we're using this extra test, but it's not | |
1314 | * exactly clear why. | |
1315 | */ | |
1316 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1317 | (re->distance != dplane_ctx_get_old_distance(ctx) || | |
1318 | re->tag != dplane_ctx_get_old_tag(ctx))) { | |
1319 | result = false; | |
1320 | } | |
1321 | } | |
1322 | ||
1323 | } else { | |
1324 | /* | |
1325 | * Ordinary, single-route case using primary context info | |
1326 | */ | |
1327 | if ((dplane_ctx_get_op(ctx) != DPLANE_OP_ROUTE_DELETE) && | |
1328 | CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1329 | /* Skip route that's been deleted */ | |
1330 | goto done; | |
1331 | } | |
1332 | ||
1333 | if ((re->type == dplane_ctx_get_type(ctx)) && | |
1334 | (re->instance == dplane_ctx_get_instance(ctx))) { | |
1335 | result = true; | |
1336 | ||
1337 | /* TODO -- we're using this extra test, but it's not | |
1338 | * exactly clear why. | |
1339 | */ | |
1340 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1341 | (re->distance != dplane_ctx_get_distance(ctx) || | |
1342 | re->tag != dplane_ctx_get_tag(ctx))) { | |
1343 | result = false; | |
1344 | } | |
1345 | } | |
1346 | } | |
1347 | ||
1348 | done: | |
1349 | ||
1350 | return (result); | |
1351 | } | |
1352 | ||
3cdba47a DS |
1353 | static void zebra_rib_fixup_system(struct route_node *rn) |
1354 | { | |
1355 | struct route_entry *re; | |
1356 | ||
1357 | RNODE_FOREACH_RE(rn, re) { | |
1358 | struct nexthop *nhop; | |
1359 | ||
1360 | if (!RIB_SYSTEM_ROUTE(re)) | |
1361 | continue; | |
1362 | ||
1363 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1364 | continue; | |
1365 | ||
1366 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1367 | ||
1368 | for (ALL_NEXTHOPS(re->ng, nhop)) { | |
1369 | if (CHECK_FLAG(nhop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1370 | continue; | |
1371 | ||
1372 | SET_FLAG(nhop->flags, NEXTHOP_FLAG_FIB); | |
1373 | } | |
1374 | } | |
1375 | } | |
1376 | ||
e5ac2adf | 1377 | /* |
78bf56b0 MS |
1378 | * Update a route from a dplane context. This consolidates common code |
1379 | * that can be used in processing of results from FIB updates, and in | |
1380 | * async notification processing. | |
efe6c026 | 1381 | * The return is 'true' if the installed nexthops changed; 'false' otherwise. |
e5ac2adf | 1382 | */ |
efe6c026 MS |
1383 | static bool rib_update_re_from_ctx(struct route_entry *re, |
1384 | struct route_node *rn, | |
1385 | struct zebra_dplane_ctx *ctx) | |
78bf56b0 | 1386 | { |
941e261c | 1387 | char dest_str[PREFIX_STRLEN] = ""; |
efe6c026 | 1388 | char nh_str[NEXTHOP_STRLEN]; |
78bf56b0 | 1389 | struct nexthop *nexthop, *ctx_nexthop; |
941e261c MS |
1390 | bool matched; |
1391 | const struct nexthop_group *ctxnhg; | |
efe6c026 MS |
1392 | bool is_selected = false; /* Is 're' currently the selected re? */ |
1393 | bool changed_p = false; /* Change to nexthops? */ | |
1394 | rib_dest_t *dest; | |
78bf56b0 | 1395 | |
941e261c MS |
1396 | /* Note well: only capturing the prefix string if debug is enabled here; |
1397 | * unconditional log messages will have to generate the string. | |
1398 | */ | |
1399 | if (IS_ZEBRA_DEBUG_RIB) | |
1400 | prefix2str(&(rn->p), dest_str, sizeof(dest_str)); | |
78bf56b0 | 1401 | |
efe6c026 MS |
1402 | dest = rib_dest_from_rnode(rn); |
1403 | if (dest) | |
1404 | is_selected = (re == dest->selected_fib); | |
1405 | ||
1406 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1407 | zlog_debug("update_from_ctx: %u:%s: %sSELECTED", | |
1408 | re->vrf_id, dest_str, (is_selected ? "" : "NOT ")); | |
1409 | ||
1410 | /* Update zebra's nexthop FIB flag for each nexthop that was installed. | |
941e261c | 1411 | * If the installed set differs from the set requested by the rib/owner, |
efe6c026 MS |
1412 | * we use the fib-specific nexthop-group to record the actual FIB |
1413 | * status. | |
78bf56b0 | 1414 | */ |
941e261c MS |
1415 | |
1416 | /* | |
1417 | * First check the fib nexthop-group, if it's present. The comparison | |
efe6c026 | 1418 | * here is quite strict: we require that the fib sets match exactly. |
941e261c MS |
1419 | */ |
1420 | matched = false; | |
1421 | do { | |
1422 | if (re->fib_ng.nexthop == NULL) | |
1423 | break; | |
1424 | ||
1425 | matched = true; | |
1426 | ||
1427 | /* First check the route's fib nexthops */ | |
1428 | for (ALL_NEXTHOPS(re->fib_ng, nexthop)) { | |
1429 | ||
1430 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1431 | continue; | |
1432 | ||
efe6c026 | 1433 | ctx_nexthop = NULL; |
941e261c MS |
1434 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), |
1435 | ctx_nexthop)) { | |
1436 | if (nexthop_same(ctx_nexthop, nexthop)) | |
1437 | break; | |
1438 | } | |
1439 | ||
1440 | if (ctx_nexthop == NULL) { | |
1441 | /* Nexthop not in the new installed set */ | |
efe6c026 MS |
1442 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1443 | nexthop2str(nexthop, nh_str, | |
1444 | sizeof(nh_str)); | |
1445 | zlog_debug("update_from_ctx: no match for fib nh %s", | |
1446 | nh_str); | |
1447 | } | |
1448 | ||
941e261c MS |
1449 | matched = false; |
1450 | break; | |
1451 | } | |
1452 | } | |
1453 | ||
1454 | if (!matched) | |
1455 | break; | |
1456 | ||
1457 | /* Check the new installed set */ | |
efe6c026 | 1458 | ctx_nexthop = NULL; |
941e261c MS |
1459 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), ctx_nexthop)) { |
1460 | ||
1461 | if (CHECK_FLAG(ctx_nexthop->flags, | |
1462 | NEXTHOP_FLAG_RECURSIVE)) | |
1463 | continue; | |
1464 | ||
1465 | /* Compare with the current group's nexthops */ | |
efe6c026 MS |
1466 | nexthop = NULL; |
1467 | for (ALL_NEXTHOPS(re->fib_ng, nexthop)) { | |
941e261c MS |
1468 | if (nexthop_same(nexthop, ctx_nexthop)) |
1469 | break; | |
1470 | } | |
1471 | ||
1472 | if (nexthop == NULL) { | |
1473 | /* Nexthop not in the old installed set */ | |
efe6c026 MS |
1474 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1475 | nexthop2str(ctx_nexthop, nh_str, | |
1476 | sizeof(nh_str)); | |
1477 | zlog_debug("update_from_ctx: no fib match for notif nh %s", | |
1478 | nh_str); | |
1479 | } | |
941e261c MS |
1480 | matched = false; |
1481 | break; | |
1482 | } | |
1483 | } | |
1484 | ||
1485 | } while (0); | |
1486 | ||
1487 | /* If the new FIB set matches the existing FIB set, we're done. */ | |
1488 | if (matched) { | |
1489 | if (IS_ZEBRA_DEBUG_RIB) | |
1490 | zlog_debug("%u:%s update_from_ctx(): existing fib nhg, no change", | |
1491 | re->vrf_id, dest_str); | |
1492 | goto done; | |
1493 | ||
1494 | } else if (re->fib_ng.nexthop) { | |
1495 | /* | |
1496 | * Free stale fib list and move on to check the rib nhg. | |
1497 | */ | |
1498 | if (IS_ZEBRA_DEBUG_RIB) | |
1499 | zlog_debug("%u:%s update_from_ctx(): replacing fib nhg", | |
1500 | re->vrf_id, dest_str); | |
1501 | nexthops_free(re->fib_ng.nexthop); | |
1502 | re->fib_ng.nexthop = NULL; | |
efe6c026 MS |
1503 | |
1504 | /* Note that the installed nexthops have changed */ | |
1505 | changed_p = true; | |
941e261c MS |
1506 | } else { |
1507 | if (IS_ZEBRA_DEBUG_RIB) | |
1508 | zlog_debug("%u:%s update_from_ctx(): no fib nhg", | |
1509 | re->vrf_id, dest_str); | |
1510 | } | |
1511 | ||
1512 | /* | |
1513 | * Compare with the rib nexthop group. The comparison here is different: | |
1514 | * the RIB group may be a superset of the list installed in the FIB. We | |
1515 | * walk the RIB group, looking for the 'installable' candidate | |
1516 | * nexthops, and then check those against the set | |
1517 | * that is actually installed. | |
1518 | */ | |
1519 | matched = true; | |
5695d9ac | 1520 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
78bf56b0 | 1521 | |
5695d9ac | 1522 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) |
78bf56b0 MS |
1523 | continue; |
1524 | ||
941e261c MS |
1525 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) |
1526 | continue; | |
5695d9ac | 1527 | |
941e261c | 1528 | /* Check for a FIB nexthop corresponding to the RIB nexthop */ |
efe6c026 | 1529 | ctx_nexthop = NULL; |
941e261c | 1530 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), ctx_nexthop)) { |
78bf56b0 MS |
1531 | if (nexthop_same(ctx_nexthop, nexthop)) |
1532 | break; | |
1533 | } | |
1534 | ||
5695d9ac MS |
1535 | /* If the FIB doesn't know about the nexthop, |
1536 | * it's not installed | |
1537 | */ | |
1538 | if (ctx_nexthop == NULL) { | |
efe6c026 MS |
1539 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
1540 | nexthop2str(nexthop, nh_str, sizeof(nh_str)); | |
1541 | zlog_debug("update_from_ctx: no notif match for rib nh %s", | |
1542 | nh_str); | |
1543 | } | |
941e261c | 1544 | matched = false; |
efe6c026 MS |
1545 | |
1546 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1547 | changed_p = true; | |
1548 | ||
1549 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
941e261c | 1550 | break; |
5695d9ac | 1551 | } |
78bf56b0 | 1552 | |
efe6c026 MS |
1553 | if (CHECK_FLAG(ctx_nexthop->flags, NEXTHOP_FLAG_FIB)) { |
1554 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1555 | changed_p = true; | |
1556 | ||
5695d9ac | 1557 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
efe6c026 MS |
1558 | } else { |
1559 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1560 | changed_p = true; | |
1561 | ||
5695d9ac | 1562 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
efe6c026 | 1563 | } |
78bf56b0 MS |
1564 | } |
1565 | ||
941e261c MS |
1566 | /* If all nexthops were processed, we're done */ |
1567 | if (matched) { | |
1568 | if (IS_ZEBRA_DEBUG_RIB) | |
efe6c026 MS |
1569 | zlog_debug("%u:%s update_from_ctx(): rib nhg matched, changed '%s'", |
1570 | re->vrf_id, dest_str, | |
1571 | (changed_p ? "true" : "false")); | |
941e261c MS |
1572 | goto done; |
1573 | } | |
78bf56b0 | 1574 | |
941e261c MS |
1575 | /* FIB nexthop set differs from the RIB set: |
1576 | * create a fib-specific nexthop-group | |
1577 | */ | |
1578 | if (IS_ZEBRA_DEBUG_RIB) | |
efe6c026 MS |
1579 | zlog_debug("%u:%s update_from_ctx(): changed %s, adding new fib nhg", |
1580 | re->vrf_id, dest_str, | |
1581 | (changed_p ? "true" : "false")); | |
941e261c MS |
1582 | |
1583 | ctxnhg = dplane_ctx_get_ng(ctx); | |
188a00e0 MS |
1584 | |
1585 | if (ctxnhg->nexthop) | |
1586 | copy_nexthops(&(re->fib_ng.nexthop), ctxnhg->nexthop, NULL); | |
1587 | else { | |
1588 | /* Bit of a special case when the fib has _no_ installed | |
1589 | * nexthops. | |
1590 | */ | |
1591 | nexthop = nexthop_new(); | |
1592 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
50d89650 | 1593 | _nexthop_add(&(re->fib_ng.nexthop), nexthop); |
188a00e0 | 1594 | } |
941e261c MS |
1595 | |
1596 | done: | |
efe6c026 | 1597 | return changed_p; |
78bf56b0 MS |
1598 | } |
1599 | ||
e5ac2adf | 1600 | /* |
fad4d69c MS |
1601 | * Helper to locate a zebra route-node from a dplane context. This is used |
1602 | * when processing dplane results, e.g. Note well: the route-node is returned | |
1603 | * with a ref held - route_unlock_node() must be called eventually. | |
e5ac2adf | 1604 | */ |
fad4d69c MS |
1605 | static struct route_node * |
1606 | rib_find_rn_from_ctx(const struct zebra_dplane_ctx *ctx) | |
e5ac2adf MS |
1607 | { |
1608 | struct route_table *table = NULL; | |
1609 | struct route_node *rn = NULL; | |
e5ac2adf MS |
1610 | const struct prefix *dest_pfx, *src_pfx; |
1611 | ||
1612 | /* Locate rn and re(s) from ctx */ | |
1613 | ||
1614 | table = zebra_vrf_table_with_table_id(dplane_ctx_get_afi(ctx), | |
1615 | dplane_ctx_get_safi(ctx), | |
1616 | dplane_ctx_get_vrf(ctx), | |
1617 | dplane_ctx_get_table(ctx)); | |
1618 | if (table == NULL) { | |
1619 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fad4d69c | 1620 | zlog_debug("Failed to find route for ctx: no table for afi %d, safi %d, vrf %u", |
e5ac2adf MS |
1621 | dplane_ctx_get_afi(ctx), |
1622 | dplane_ctx_get_safi(ctx), | |
1623 | dplane_ctx_get_vrf(ctx)); | |
1624 | } | |
1625 | goto done; | |
1626 | } | |
1627 | ||
fad4d69c MS |
1628 | dest_pfx = dplane_ctx_get_dest(ctx); |
1629 | src_pfx = dplane_ctx_get_src(ctx); | |
1630 | ||
1631 | rn = srcdest_rnode_get(table, dest_pfx, | |
1632 | src_pfx ? (struct prefix_ipv6 *)src_pfx : NULL); | |
1633 | ||
1634 | done: | |
1635 | return rn; | |
1636 | } | |
1637 | ||
1638 | ||
97f5b441 | 1639 | |
fad4d69c MS |
1640 | /* |
1641 | * Route-update results processing after async dataplane update. | |
1642 | */ | |
1643 | static void rib_process_result(struct zebra_dplane_ctx *ctx) | |
1644 | { | |
1645 | struct zebra_vrf *zvrf = NULL; | |
1646 | struct route_node *rn = NULL; | |
1647 | struct route_entry *re = NULL, *old_re = NULL, *rib; | |
1648 | bool is_update = false; | |
1649 | char dest_str[PREFIX_STRLEN] = ""; | |
1650 | enum dplane_op_e op; | |
1651 | enum zebra_dplane_result status; | |
1652 | const struct prefix *dest_pfx, *src_pfx; | |
1653 | uint32_t seq; | |
efe6c026 | 1654 | bool fib_changed = false; |
97f5b441 | 1655 | |
fad4d69c | 1656 | zvrf = vrf_info_lookup(dplane_ctx_get_vrf(ctx)); |
e5ac2adf MS |
1657 | dest_pfx = dplane_ctx_get_dest(ctx); |
1658 | ||
1659 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1660 | * unconditional log messages will have to generate the string. | |
1661 | */ | |
5709131c | 1662 | if (IS_ZEBRA_DEBUG_DPLANE) |
e5ac2adf | 1663 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
e5ac2adf | 1664 | |
fad4d69c MS |
1665 | /* Locate rn and re(s) from ctx */ |
1666 | rn = rib_find_rn_from_ctx(ctx); | |
e5ac2adf MS |
1667 | if (rn == NULL) { |
1668 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fe2c53d4 | 1669 | zlog_debug("Failed to process dplane results: no route for %u:%s", |
e5ac2adf MS |
1670 | dplane_ctx_get_vrf(ctx), dest_str); |
1671 | } | |
1672 | goto done; | |
1673 | } | |
1674 | ||
1675 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); | |
1676 | ||
1677 | op = dplane_ctx_get_op(ctx); | |
1678 | status = dplane_ctx_get_status(ctx); | |
1679 | ||
c831033f | 1680 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
f183e380 | 1681 | zlog_debug("%u:%s Processing dplane ctx %p, op %s result %s", |
e5ac2adf | 1682 | dplane_ctx_get_vrf(ctx), dest_str, ctx, |
f183e380 | 1683 | dplane_op2str(op), dplane_res2str(status)); |
e5ac2adf | 1684 | |
e5ac2adf MS |
1685 | /* |
1686 | * Update is a bit of a special case, where we may have both old and new | |
1687 | * routes to post-process. | |
1688 | */ | |
1689 | is_update = dplane_ctx_is_update(ctx); | |
1690 | ||
1691 | /* | |
1692 | * Take a pass through the routes, look for matches with the context | |
1693 | * info. | |
1694 | */ | |
1695 | RNODE_FOREACH_RE(rn, rib) { | |
1696 | ||
1697 | if (re == NULL) { | |
5709131c | 1698 | if (rib_route_match_ctx(rib, ctx, false)) |
e5ac2adf | 1699 | re = rib; |
e5ac2adf MS |
1700 | } |
1701 | ||
1702 | /* Check for old route match */ | |
1703 | if (is_update && (old_re == NULL)) { | |
5709131c | 1704 | if (rib_route_match_ctx(rib, ctx, true /*is_update*/)) |
e5ac2adf | 1705 | old_re = rib; |
e5ac2adf MS |
1706 | } |
1707 | ||
1708 | /* Have we found the routes we need to work on? */ | |
5709131c | 1709 | if (re && ((!is_update || old_re))) |
e5ac2adf | 1710 | break; |
e5ac2adf MS |
1711 | } |
1712 | ||
699dae23 DS |
1713 | seq = dplane_ctx_get_seq(ctx); |
1714 | ||
e5ac2adf MS |
1715 | /* |
1716 | * Check sequence number(s) to detect stale results before continuing | |
1717 | */ | |
60f98b23 | 1718 | if (re) { |
699dae23 | 1719 | if (re->dplane_sequence != seq) { |
60f98b23 DS |
1720 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
1721 | zlog_debug("%u:%s Stale dplane result for re %p", | |
1722 | dplane_ctx_get_vrf(ctx), | |
1723 | dest_str, re); | |
1724 | } else | |
1725 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); | |
e5ac2adf MS |
1726 | } |
1727 | ||
60f98b23 DS |
1728 | if (old_re) { |
1729 | if (old_re->dplane_sequence != dplane_ctx_get_old_seq(ctx)) { | |
1730 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
1731 | zlog_debug("%u:%s Stale dplane result for old_re %p", | |
1732 | dplane_ctx_get_vrf(ctx), | |
1733 | dest_str, old_re); | |
1734 | } else | |
9ef0c6ba | 1735 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_QUEUED); |
e5ac2adf MS |
1736 | } |
1737 | ||
12e7fe3a | 1738 | switch (op) { |
12e7fe3a DS |
1739 | case DPLANE_OP_ROUTE_INSTALL: |
1740 | case DPLANE_OP_ROUTE_UPDATE: | |
1741 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
677c1dd5 DS |
1742 | if (re) { |
1743 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
1744 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1745 | } | |
b9f0e5ee DS |
1746 | /* |
1747 | * On an update operation from the same route type | |
1748 | * context retrieval currently has no way to know | |
1749 | * which was the old and which was the new. | |
1750 | * So don't unset our flags that we just set. | |
1751 | * We know redistribution is ok because the | |
1752 | * old_re in this case is used for nothing | |
1753 | * more than knowing whom to contact if necessary. | |
1754 | */ | |
1755 | if (old_re && old_re != re) { | |
677c1dd5 DS |
1756 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
1757 | UNSET_FLAG(old_re->status, | |
1758 | ROUTE_ENTRY_INSTALLED); | |
1759 | } | |
f183e380 | 1760 | |
78bf56b0 | 1761 | /* Update zebra route based on the results in |
941e261c | 1762 | * the context struct. |
78bf56b0 | 1763 | */ |
941e261c | 1764 | if (re) { |
efe6c026 MS |
1765 | fib_changed = |
1766 | rib_update_re_from_ctx(re, rn, ctx); | |
1767 | ||
1768 | if (!fib_changed) { | |
1769 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
1770 | zlog_debug("%u:%s no fib change for re", | |
1771 | dplane_ctx_get_vrf( | |
1772 | ctx), | |
1773 | dest_str); | |
12e7fe3a DS |
1774 | } |
1775 | ||
941e261c MS |
1776 | /* Redistribute */ |
1777 | redistribute_update(dest_pfx, src_pfx, | |
1778 | re, NULL); | |
f183e380 MS |
1779 | } |
1780 | ||
3cdba47a DS |
1781 | /* |
1782 | * System routes are weird in that they | |
1783 | * allow multiple to be installed that match | |
1784 | * to the same prefix, so after we get the | |
1785 | * result we need to clean them up so that | |
1786 | * we can actually use them. | |
1787 | */ | |
1788 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
1789 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
1790 | zebra_rib_fixup_system(rn); | |
1791 | ||
f4c6e2a8 | 1792 | if (zvrf) |
12e7fe3a | 1793 | zvrf->installs++; |
f183e380 | 1794 | |
12e7fe3a | 1795 | /* Notify route owner */ |
677c1dd5 | 1796 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_INSTALLED); |
e5ac2adf | 1797 | |
12e7fe3a | 1798 | } else { |
7a230a9d | 1799 | if (re) { |
677c1dd5 | 1800 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); |
7a230a9d DS |
1801 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
1802 | } if (old_re) | |
677c1dd5 DS |
1803 | SET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
1804 | if (re) | |
1805 | zsend_route_notify_owner(re, dest_pfx, | |
1806 | ZAPI_ROUTE_FAIL_INSTALL); | |
97f5b441 | 1807 | |
12e7fe3a DS |
1808 | zlog_warn("%u:%s: Route install failed", |
1809 | dplane_ctx_get_vrf(ctx), | |
1810 | prefix2str(dest_pfx, | |
1811 | dest_str, sizeof(dest_str))); | |
1812 | } | |
1813 | break; | |
1814 | case DPLANE_OP_ROUTE_DELETE: | |
677c1dd5 DS |
1815 | if (re) |
1816 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
1817 | /* |
1818 | * In the delete case, the zebra core datastructs were | |
1819 | * updated (or removed) at the time the delete was issued, | |
1820 | * so we're just notifying the route owner. | |
e5ac2adf | 1821 | */ |
12e7fe3a | 1822 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { |
677c1dd5 DS |
1823 | if (re) { |
1824 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1825 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
1826 | } | |
12e7fe3a | 1827 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_REMOVED); |
e5ac2adf | 1828 | |
12e7fe3a DS |
1829 | if (zvrf) |
1830 | zvrf->removals++; | |
1831 | } else { | |
677c1dd5 DS |
1832 | if (re) |
1833 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
1834 | zsend_route_notify_owner_ctx(ctx, |
1835 | ZAPI_ROUTE_REMOVE_FAIL); | |
e5ac2adf | 1836 | |
12e7fe3a DS |
1837 | zlog_warn("%u:%s: Route Deletion failure", |
1838 | dplane_ctx_get_vrf(ctx), | |
1839 | prefix2str(dest_pfx, | |
1840 | dest_str, sizeof(dest_str))); | |
1841 | } | |
3cdba47a DS |
1842 | |
1843 | /* | |
1844 | * System routes are weird in that they | |
1845 | * allow multiple to be installed that match | |
1846 | * to the same prefix, so after we get the | |
1847 | * result we need to clean them up so that | |
1848 | * we can actually use them. | |
1849 | */ | |
1850 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
1851 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
1852 | zebra_rib_fixup_system(rn); | |
12e7fe3a | 1853 | break; |
d37f4d6c MS |
1854 | default: |
1855 | break; | |
e5ac2adf | 1856 | } |
699dae23 DS |
1857 | |
1858 | zebra_rib_evaluate_rn_nexthops(rn, seq); | |
a1494c25 | 1859 | zebra_rib_evaluate_mpls(rn); |
e5ac2adf MS |
1860 | done: |
1861 | ||
5f27bcba DS |
1862 | if (rn) |
1863 | route_unlock_node(rn); | |
1864 | ||
e5ac2adf MS |
1865 | /* Return context to dataplane module */ |
1866 | dplane_ctx_fini(&ctx); | |
1867 | } | |
1868 | ||
54818e3b MS |
1869 | /* |
1870 | * Handle notification from async dataplane: the dataplane has detected | |
1871 | * some change to a route, and notifies zebra so that the control plane | |
1872 | * can reflect that change. | |
1873 | */ | |
1874 | static void rib_process_dplane_notify(struct zebra_dplane_ctx *ctx) | |
1875 | { | |
1876 | struct route_node *rn = NULL; | |
efe6c026 MS |
1877 | struct route_entry *re = NULL; |
1878 | struct nexthop *nexthop; | |
54818e3b MS |
1879 | char dest_str[PREFIX_STRLEN] = ""; |
1880 | const struct prefix *dest_pfx, *src_pfx; | |
941e261c | 1881 | rib_dest_t *dest; |
efe6c026 | 1882 | bool fib_changed = false; |
941e261c | 1883 | bool debug_p = IS_ZEBRA_DEBUG_DPLANE | IS_ZEBRA_DEBUG_RIB; |
efe6c026 | 1884 | int start_count, end_count; |
54818e3b MS |
1885 | dest_pfx = dplane_ctx_get_dest(ctx); |
1886 | ||
1887 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1888 | * unconditional log messages will have to generate the string. | |
1889 | */ | |
941e261c | 1890 | if (debug_p) |
54818e3b MS |
1891 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
1892 | ||
1893 | /* Locate rn and re(s) from ctx */ | |
1894 | rn = rib_find_rn_from_ctx(ctx); | |
1895 | if (rn == NULL) { | |
941e261c MS |
1896 | if (debug_p) { |
1897 | zlog_debug("Failed to process dplane notification: no routes for %u:%s", | |
54818e3b MS |
1898 | dplane_ctx_get_vrf(ctx), dest_str); |
1899 | } | |
1900 | goto done; | |
1901 | } | |
1902 | ||
941e261c | 1903 | dest = rib_dest_from_rnode(rn); |
54818e3b MS |
1904 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); |
1905 | ||
efe6c026 | 1906 | if (debug_p) |
54818e3b MS |
1907 | zlog_debug("%u:%s Processing dplane notif ctx %p", |
1908 | dplane_ctx_get_vrf(ctx), dest_str, ctx); | |
1909 | ||
1910 | /* | |
1911 | * Take a pass through the routes, look for matches with the context | |
1912 | * info. | |
1913 | */ | |
efe6c026 MS |
1914 | RNODE_FOREACH_RE(rn, re) { |
1915 | if (rib_route_match_ctx(re, ctx, false /*!update*/)) | |
54818e3b MS |
1916 | break; |
1917 | } | |
1918 | ||
1919 | /* No match? Nothing we can do */ | |
1920 | if (re == NULL) { | |
941e261c MS |
1921 | if (debug_p) |
1922 | zlog_debug("%u:%s Unable to process dplane notification: no entry for type %s", | |
54818e3b | 1923 | dplane_ctx_get_vrf(ctx), dest_str, |
941e261c MS |
1924 | zebra_route_string( |
1925 | dplane_ctx_get_type(ctx))); | |
54818e3b MS |
1926 | |
1927 | goto done; | |
1928 | } | |
1929 | ||
941e261c MS |
1930 | /* Is this a notification that ... matters? We only really care about |
1931 | * the route that is currently selected for installation. | |
1932 | */ | |
1933 | if (re != dest->selected_fib) { | |
1934 | /* TODO -- don't skip processing entirely? We might like to | |
1935 | * at least report on the event. | |
1936 | */ | |
1937 | if (debug_p) | |
efe6c026 | 1938 | zlog_debug("%u:%s dplane notif, but type %s not selected_fib", |
941e261c MS |
1939 | dplane_ctx_get_vrf(ctx), dest_str, |
1940 | zebra_route_string( | |
1941 | dplane_ctx_get_type(ctx))); | |
1942 | goto done; | |
1943 | } | |
1944 | ||
efe6c026 MS |
1945 | /* We'll want to determine whether the installation status of the |
1946 | * route has changed: we'll check the status before processing, | |
1947 | * and then again if there's been a change. | |
1948 | */ | |
1949 | start_count = 0; | |
1950 | for (ALL_NEXTHOPS_PTR(rib_active_nhg(re), nexthop)) { | |
1951 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1952 | start_count++; | |
1953 | } | |
1954 | ||
54818e3b MS |
1955 | /* Update zebra's nexthop FIB flags based on the context struct's |
1956 | * nexthops. | |
1957 | */ | |
efe6c026 | 1958 | fib_changed = rib_update_re_from_ctx(re, rn, ctx); |
54818e3b | 1959 | |
efe6c026 MS |
1960 | if (!fib_changed) { |
1961 | if (debug_p) | |
54818e3b MS |
1962 | zlog_debug("%u:%s No change from dplane notification", |
1963 | dplane_ctx_get_vrf(ctx), dest_str); | |
efe6c026 MS |
1964 | |
1965 | goto done; | |
54818e3b MS |
1966 | } |
1967 | ||
188a00e0 MS |
1968 | /* |
1969 | * Perform follow-up work if the actual status of the prefix | |
efe6c026 | 1970 | * changed. |
941e261c MS |
1971 | */ |
1972 | ||
efe6c026 MS |
1973 | end_count = 0; |
1974 | for (ALL_NEXTHOPS_PTR(rib_active_nhg(re), nexthop)) { | |
1975 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) | |
1976 | end_count++; | |
1977 | } | |
941e261c | 1978 | |
188a00e0 MS |
1979 | /* Various fib transitions: changed nexthops; from installed to |
1980 | * not-installed; or not-installed to installed. | |
efe6c026 | 1981 | */ |
188a00e0 MS |
1982 | if (start_count > 0 && end_count > 0) { |
1983 | ||
1984 | /* Changed nexthops - update kernel/others */ | |
1985 | dplane_route_notif_update(rn, re, | |
1986 | DPLANE_OP_ROUTE_UPDATE, ctx); | |
1987 | ||
1988 | } else if (start_count == 0 && end_count > 0) { | |
efe6c026 MS |
1989 | if (debug_p) |
1990 | zlog_debug("%u:%s installed transition from dplane notification", | |
1991 | dplane_ctx_get_vrf(ctx), dest_str); | |
1992 | ||
1993 | /* We expect this to be the selected route, so we want | |
1994 | * to tell others about this transistion. | |
1995 | */ | |
1996 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1997 | ||
188a00e0 MS |
1998 | /* Changed nexthops - update kernel/others */ |
1999 | dplane_route_notif_update(rn, re, DPLANE_OP_ROUTE_INSTALL, ctx); | |
2000 | ||
efe6c026 MS |
2001 | /* Redistribute, lsp, and nht update */ |
2002 | redistribute_update(dest_pfx, src_pfx, re, NULL); | |
2003 | ||
2004 | zebra_rib_evaluate_rn_nexthops( | |
2005 | rn, zebra_router_get_next_sequence()); | |
2006 | ||
2007 | zebra_rib_evaluate_mpls(rn); | |
2008 | ||
2009 | } else if (start_count > 0 && end_count == 0) { | |
2010 | if (debug_p) | |
2011 | zlog_debug("%u:%s un-installed transition from dplane notification", | |
2012 | dplane_ctx_get_vrf(ctx), dest_str); | |
2013 | ||
2014 | /* Transition from _something_ installed to _nothing_ | |
2015 | * installed. | |
2016 | */ | |
2017 | /* We expect this to be the selected route, so we want | |
2018 | * to tell others about this transistion. | |
2019 | */ | |
2020 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
2021 | ||
188a00e0 MS |
2022 | /* Changed nexthops - update kernel/others */ |
2023 | dplane_route_notif_update(rn, re, DPLANE_OP_ROUTE_DELETE, ctx); | |
2024 | ||
efe6c026 MS |
2025 | /* Redistribute, lsp, and nht update */ |
2026 | redistribute_delete(dest_pfx, src_pfx, re); | |
2027 | ||
2028 | zebra_rib_evaluate_rn_nexthops( | |
2029 | rn, zebra_router_get_next_sequence()); | |
2030 | ||
2031 | zebra_rib_evaluate_mpls(rn); | |
2032 | } | |
941e261c | 2033 | |
54818e3b | 2034 | done: |
941e261c MS |
2035 | if (rn) |
2036 | route_unlock_node(rn); | |
2037 | ||
54818e3b MS |
2038 | /* Return context to dataplane module */ |
2039 | dplane_ctx_fini(&ctx); | |
2040 | } | |
2041 | ||
5110a0c6 | 2042 | /* Take a list of route_node structs and return 1, if there was a record |
d62a17ae | 2043 | * picked from it and processed by rib_process(). Don't process more, |
5110a0c6 | 2044 | * than one RN record; operate only in the specified sub-queue. |
e96f9203 | 2045 | */ |
d7c0a89a | 2046 | static unsigned int process_subq(struct list *subq, uint8_t qindex) |
e96f9203 | 2047 | { |
d62a17ae | 2048 | struct listnode *lnode = listhead(subq); |
2049 | struct route_node *rnode; | |
2050 | rib_dest_t *dest; | |
2051 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 2052 | |
d62a17ae | 2053 | if (!lnode) |
2054 | return 0; | |
5110a0c6 | 2055 | |
d62a17ae | 2056 | rnode = listgetdata(lnode); |
2057 | dest = rib_dest_from_rnode(rnode); | |
2058 | if (dest) | |
2059 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 2060 | |
d62a17ae | 2061 | rib_process(rnode); |
5110a0c6 | 2062 | |
d62a17ae | 2063 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
2064 | char buf[SRCDEST2STR_BUFFER]; | |
efe6c026 | 2065 | |
d62a17ae | 2066 | srcdest_rnode2str(rnode, buf, sizeof(buf)); |
2067 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
2068 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
2069 | } | |
41ec9222 | 2070 | |
d62a17ae | 2071 | if (rnode->info) |
2072 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
2073 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 2074 | |
67b9467f | 2075 | #if 0 |
5110a0c6 SH |
2076 | else |
2077 | { | |
2078 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
2079 | __func__, rnode, rnode->lock); | |
2080 | zlog_backtrace(LOG_DEBUG); | |
2081 | } | |
67b9467f | 2082 | #endif |
d62a17ae | 2083 | route_unlock_node(rnode); |
2084 | list_delete_node(subq, lnode); | |
2085 | return 1; | |
e96f9203 DO |
2086 | } |
2087 | ||
a1494c25 | 2088 | |
fb018d25 | 2089 | /* |
f183e380 | 2090 | * Perform next-hop tracking processing after RIB updates. |
fb018d25 | 2091 | */ |
f183e380 | 2092 | static void do_nht_processing(void) |
fb018d25 | 2093 | { |
fb018d25 DS |
2094 | } |
2095 | ||
e96f9203 | 2096 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from |
d62a17ae | 2097 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
2098 | * data | |
e96f9203 DO |
2099 | * is pointed to the meta queue structure. |
2100 | */ | |
d62a17ae | 2101 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 2102 | { |
d62a17ae | 2103 | struct meta_queue *mq = data; |
2104 | unsigned i; | |
91f16812 MS |
2105 | uint32_t queue_len, queue_limit; |
2106 | ||
2107 | /* Ensure there's room for more dataplane updates */ | |
2108 | queue_limit = dplane_get_in_queue_limit(); | |
2109 | queue_len = dplane_get_in_queue_len(); | |
2110 | if (queue_len > queue_limit) { | |
2111 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
8b962e77 MS |
2112 | zlog_debug("rib queue: dplane queue len %u, limit %u, retrying", |
2113 | queue_len, queue_limit); | |
91f16812 MS |
2114 | |
2115 | /* Ensure that the meta-queue is actually enqueued */ | |
489a9614 | 2116 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2117 | work_queue_add(zrouter.ribq, zrouter.mq); |
91f16812 MS |
2118 | |
2119 | return WQ_QUEUE_BLOCKED; | |
2120 | } | |
5110a0c6 | 2121 | |
d62a17ae | 2122 | for (i = 0; i < MQ_SIZE; i++) |
2123 | if (process_subq(mq->subq[i], i)) { | |
2124 | mq->size--; | |
2125 | break; | |
2126 | } | |
2127 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
2128 | } |
2129 | ||
f52ed677 DS |
2130 | |
2131 | /* | |
2132 | * Look into the RN and queue it into the highest priority queue | |
2133 | * at this point in time for processing. | |
2134 | * | |
2135 | * We will enqueue a route node only once per invocation. | |
2136 | * | |
2137 | * There are two possibilities here that should be kept in mind. | |
2138 | * If the original invocation has not been pulled off for processing | |
2139 | * yet, A subsuquent invocation can have a route entry with a better | |
2140 | * meta queue index value and we can have a situation where | |
2141 | * we might have the same node enqueued 2 times. Not necessarily | |
2142 | * an optimal situation but it should be ok. | |
2143 | * | |
2144 | * The other possibility is that the original invocation has not | |
2145 | * been pulled off for processing yet, A subsusquent invocation | |
2146 | * doesn't have a route_entry with a better meta-queue and the | |
2147 | * original metaqueue index value will win and we'll end up with | |
2148 | * the route node enqueued once. | |
e96f9203 | 2149 | */ |
d62a17ae | 2150 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) |
e96f9203 | 2151 | { |
f52ed677 DS |
2152 | struct route_entry *re = NULL, *curr_re = NULL; |
2153 | uint8_t qindex = MQ_SIZE, curr_qindex = MQ_SIZE; | |
5110a0c6 | 2154 | |
f52ed677 DS |
2155 | RNODE_FOREACH_RE (rn, curr_re) { |
2156 | curr_qindex = route_info[curr_re->type].meta_q_map; | |
d62a17ae | 2157 | |
f52ed677 DS |
2158 | if (curr_qindex <= qindex) { |
2159 | re = curr_re; | |
2160 | qindex = curr_qindex; | |
d62a17ae | 2161 | } |
f52ed677 | 2162 | } |
5110a0c6 | 2163 | |
f52ed677 DS |
2164 | if (!re) |
2165 | return; | |
5110a0c6 | 2166 | |
f52ed677 DS |
2167 | /* Invariant: at this point we always have rn->info set. */ |
2168 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
2169 | RIB_ROUTE_QUEUED(qindex))) { | |
d62a17ae | 2170 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2171 | rnode_debug(rn, re->vrf_id, | |
f52ed677 | 2172 | "rn %p is already queued in sub-queue %u", |
d62a17ae | 2173 | (void *)rn, qindex); |
f52ed677 | 2174 | return; |
d62a17ae | 2175 | } |
f52ed677 DS |
2176 | |
2177 | SET_FLAG(rib_dest_from_rnode(rn)->flags, RIB_ROUTE_QUEUED(qindex)); | |
2178 | listnode_add(mq->subq[qindex], rn); | |
2179 | route_lock_node(rn); | |
2180 | mq->size++; | |
2181 | ||
2182 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2183 | rnode_debug(rn, re->vrf_id, "queued rn %p into sub-queue %u", | |
2184 | (void *)rn, qindex); | |
4d38fdb4 | 2185 | } |
2186 | ||
6d691129 | 2187 | /* Add route_node to work queue and schedule processing */ |
d62a17ae | 2188 | void rib_queue_add(struct route_node *rn) |
4d38fdb4 | 2189 | { |
d62a17ae | 2190 | assert(rn); |
fc328ac9 | 2191 | |
d62a17ae | 2192 | /* Pointless to queue a route_node with no RIB entries to add or remove |
2193 | */ | |
2194 | if (!rnode_to_ribs(rn)) { | |
2195 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
2196 | __func__, (void *)rn, rn->lock); | |
2197 | zlog_backtrace(LOG_DEBUG); | |
2198 | return; | |
2199 | } | |
4d38fdb4 | 2200 | |
489a9614 | 2201 | if (zrouter.ribq == NULL) { |
e914ccbe | 2202 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2203 | "%s: work_queue does not exist!", __func__); |
d62a17ae | 2204 | return; |
2205 | } | |
2206 | ||
2207 | /* | |
2208 | * The RIB queue should normally be either empty or holding the only | |
2209 | * work_queue_item element. In the latter case this element would | |
2210 | * hold a pointer to the meta queue structure, which must be used to | |
2211 | * actually queue the route nodes to process. So create the MQ | |
2212 | * holder, if necessary, then push the work into it in any case. | |
2213 | * This semantics was introduced after 0.99.9 release. | |
2214 | */ | |
489a9614 | 2215 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2216 | work_queue_add(zrouter.ribq, zrouter.mq); |
e96f9203 | 2217 | |
ea45a4e7 | 2218 | rib_meta_queue_add(zrouter.mq, rn); |
fc328ac9 | 2219 | |
d62a17ae | 2220 | return; |
4d38fdb4 | 2221 | } |
2222 | ||
5110a0c6 SH |
2223 | /* Create new meta queue. |
2224 | A destructor function doesn't seem to be necessary here. | |
2225 | */ | |
d62a17ae | 2226 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 2227 | { |
d62a17ae | 2228 | struct meta_queue *new; |
2229 | unsigned i; | |
5110a0c6 | 2230 | |
d62a17ae | 2231 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
e96f9203 | 2232 | |
d62a17ae | 2233 | for (i = 0; i < MQ_SIZE; i++) { |
2234 | new->subq[i] = list_new(); | |
2235 | assert(new->subq[i]); | |
2236 | } | |
5110a0c6 | 2237 | |
d62a17ae | 2238 | return new; |
e96f9203 DO |
2239 | } |
2240 | ||
d62a17ae | 2241 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 2242 | { |
d62a17ae | 2243 | unsigned i; |
5a8dfcd8 | 2244 | |
d62a17ae | 2245 | for (i = 0; i < MQ_SIZE; i++) |
6a154c88 | 2246 | list_delete(&mq->subq[i]); |
5a8dfcd8 | 2247 | |
d62a17ae | 2248 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
2249 | } |
2250 | ||
4d38fdb4 | 2251 | /* initialise zebra rib work queue */ |
2561d12e | 2252 | static void rib_queue_init(void) |
4d38fdb4 | 2253 | { |
489a9614 DS |
2254 | if (!(zrouter.ribq = work_queue_new(zrouter.master, |
2255 | "route_node processing"))) { | |
e914ccbe | 2256 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2257 | "%s: could not initialise work queue!", __func__); |
d62a17ae | 2258 | return; |
2259 | } | |
4d38fdb4 | 2260 | |
d62a17ae | 2261 | /* fill in the work queue spec */ |
489a9614 DS |
2262 | zrouter.ribq->spec.workfunc = &meta_queue_process; |
2263 | zrouter.ribq->spec.errorfunc = NULL; | |
46a4e345 | 2264 | zrouter.ribq->spec.completion_func = NULL; |
d62a17ae | 2265 | /* XXX: TODO: These should be runtime configurable via vty */ |
489a9614 DS |
2266 | zrouter.ribq->spec.max_retries = 3; |
2267 | zrouter.ribq->spec.hold = ZEBRA_RIB_PROCESS_HOLD_TIME; | |
2268 | zrouter.ribq->spec.retry = ZEBRA_RIB_PROCESS_RETRY_TIME; | |
d62a17ae | 2269 | |
ea45a4e7 | 2270 | if (!(zrouter.mq = meta_queue_new())) { |
e914ccbe | 2271 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2272 | "%s: could not initialise meta queue!", __func__); |
d62a17ae | 2273 | return; |
2274 | } | |
2275 | return; | |
718e3744 | 2276 | } |
2277 | ||
434434f7 DS |
2278 | rib_dest_t *zebra_rib_create_dest(struct route_node *rn) |
2279 | { | |
2280 | rib_dest_t *dest; | |
2281 | ||
2282 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); | |
aa57abfb | 2283 | rnh_list_init(&dest->nht); |
434434f7 DS |
2284 | route_lock_node(rn); /* rn route table reference */ |
2285 | rn->info = dest; | |
2286 | dest->rnode = rn; | |
2287 | ||
2288 | return dest; | |
2289 | } | |
2290 | ||
6d691129 PJ |
2291 | /* RIB updates are processed via a queue of pointers to route_nodes. |
2292 | * | |
2293 | * The queue length is bounded by the maximal size of the routing table, | |
2294 | * as a route_node will not be requeued, if already queued. | |
2295 | * | |
f0f77c9a DS |
2296 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
2297 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 2298 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 2299 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 2300 | * |
f0f77c9a | 2301 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
2302 | * |
2303 | * rib_addnode | |
f0f77c9a DS |
2304 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
2305 | * |-------->| | best RE, if required | |
3c0755dc PJ |
2306 | * | | |
2307 | * static_install->|->rib_addqueue...... -> rib_process | |
2308 | * | | | |
2309 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
2310 | * |-> set ROUTE_ENTRY_REMOVE | |
2311 | * rib_delnode (RE freed) | |
6d691129 | 2312 | * |
9fd92e3c AS |
2313 | * The 'info' pointer of a route_node points to a rib_dest_t |
2314 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
2315 | * dest is created on-demand by rib_link() and is kept around at least | |
2316 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 2317 | * |
6d691129 PJ |
2318 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
2319 | * | |
2320 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
2321 | * - dest attached to route_node: |
2322 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
2323 | * - route_node processing queue |
2324 | * - managed by: rib_addqueue, rib_process. | |
2325 | * | |
2326 | */ | |
d62a17ae | 2327 | |
f0f77c9a | 2328 | /* Add RE to head of the route node. */ |
d62a17ae | 2329 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
2330 | { | |
d62a17ae | 2331 | rib_dest_t *dest; |
2332 | afi_t afi; | |
2333 | const char *rmap_name; | |
9fd92e3c | 2334 | |
d62a17ae | 2335 | assert(re && rn); |
9fd92e3c | 2336 | |
d62a17ae | 2337 | dest = rib_dest_from_rnode(rn); |
2338 | if (!dest) { | |
2339 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2340 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 2341 | |
434434f7 | 2342 | dest = zebra_rib_create_dest(rn); |
d62a17ae | 2343 | } |
2263a412 | 2344 | |
045207e2 | 2345 | re_list_add_head(&dest->routes, re); |
d62a17ae | 2346 | |
2347 | afi = (rn->p.family == AF_INET) | |
2348 | ? AFI_IP | |
2349 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
fe257ae7 DS |
2350 | if (is_zebra_import_table_enabled(afi, re->vrf_id, re->table)) { |
2351 | struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(re->vrf_id); | |
2352 | ||
d62a17ae | 2353 | rmap_name = zebra_get_import_table_route_map(afi, re->table); |
fe257ae7 | 2354 | zebra_add_import_table_entry(zvrf, rn, re, rmap_name); |
d62a17ae | 2355 | } else if (process) |
2356 | rib_queue_add(rn); | |
2357 | } | |
2358 | ||
7e24fdf3 DS |
2359 | static void rib_addnode(struct route_node *rn, |
2360 | struct route_entry *re, int process) | |
d62a17ae | 2361 | { |
2362 | /* RE node has been un-removed before route-node is processed. | |
2363 | * route_node must hence already be on the queue for processing.. | |
2364 | */ | |
2365 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
2366 | if (IS_ZEBRA_DEBUG_RIB) | |
2367 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
2368 | (void *)rn, (void *)re); | |
2369 | ||
2370 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2371 | return; | |
2372 | } | |
2373 | rib_link(rn, re, process); | |
6d691129 PJ |
2374 | } |
2375 | ||
9fd92e3c AS |
2376 | /* |
2377 | * rib_unlink | |
2378 | * | |
2379 | * Detach a rib structure from a route_node. | |
2380 | * | |
2381 | * Note that a call to rib_unlink() should be followed by a call to | |
2382 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
2383 | * longer required to be deleted. | |
2384 | */ | |
d62a17ae | 2385 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 2386 | { |
d62a17ae | 2387 | rib_dest_t *dest; |
9fd92e3c | 2388 | |
d62a17ae | 2389 | assert(rn && re); |
6d691129 | 2390 | |
d62a17ae | 2391 | if (IS_ZEBRA_DEBUG_RIB) |
2392 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
2393 | (void *)re); | |
6d691129 | 2394 | |
d62a17ae | 2395 | dest = rib_dest_from_rnode(rn); |
6d691129 | 2396 | |
045207e2 | 2397 | re_list_del(&dest->routes, re); |
7a4bb9c5 | 2398 | |
2eb07de3 DS |
2399 | if (dest->selected_fib == re) |
2400 | dest->selected_fib = NULL; | |
2401 | ||
7ee30f28 | 2402 | nexthops_free(re->ng.nexthop); |
ee5e8a48 MS |
2403 | nexthops_free(re->fib_ng.nexthop); |
2404 | ||
d62a17ae | 2405 | XFREE(MTYPE_RE, re); |
2406 | } | |
2407 | ||
2408 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2409 | { | |
2410 | afi_t afi; | |
2411 | ||
2412 | if (IS_ZEBRA_DEBUG_RIB) | |
2413 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2414 | (void *)rn, (void *)re); | |
2415 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2416 | ||
2417 | afi = (rn->p.family == AF_INET) | |
2418 | ? AFI_IP | |
2419 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
fe257ae7 DS |
2420 | if (is_zebra_import_table_enabled(afi, re->vrf_id, re->table)) { |
2421 | struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(re->vrf_id); | |
2422 | ||
2423 | zebra_del_import_table_entry(zvrf, rn, re); | |
d62a17ae | 2424 | /* Just clean up if non main table */ |
2425 | if (IS_ZEBRA_DEBUG_RIB) { | |
2426 | char buf[SRCDEST2STR_BUFFER]; | |
2427 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
2428 | zlog_debug("%u:%s: Freeing route rn %p, re %p (%s)", |
2429 | re->vrf_id, buf, rn, re, | |
2430 | zebra_route_string(re->type)); | |
d62a17ae | 2431 | } |
7a4bb9c5 | 2432 | |
d62a17ae | 2433 | rib_unlink(rn, re); |
2434 | } else { | |
2435 | rib_queue_add(rn); | |
2436 | } | |
718e3744 | 2437 | } |
2438 | ||
f0f77c9a | 2439 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2440 | * standard debug log. Calling function name and IP prefix in |
2441 | * question are passed as 1st and 2nd arguments. | |
2442 | */ | |
2443 | ||
d62a17ae | 2444 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2445 | union prefixconstptr src_pp, | |
2446 | const struct route_entry *re) | |
2447 | { | |
d62a17ae | 2448 | const struct prefix *src_p = src_pp.p; |
2449 | bool is_srcdst = src_p && src_p->prefixlen; | |
2450 | char straddr[PREFIX_STRLEN]; | |
2451 | char srcaddr[PREFIX_STRLEN]; | |
53c16fbe | 2452 | char nhname[PREFIX_STRLEN]; |
d62a17ae | 2453 | struct nexthop *nexthop; |
2454 | ||
2455 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2456 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2457 | is_srcdst ? " from " : "", | |
2458 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2459 | : "", | |
2460 | re->vrf_id); | |
cc54cfee | 2461 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", |
53c16fbe | 2462 | straddr, (unsigned long)re->uptime, re->type, re->instance, |
cc54cfee | 2463 | re->table); |
d62a17ae | 2464 | zlog_debug( |
2465 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
53c16fbe DS |
2466 | straddr, re->metric, re->mtu, re->distance, re->flags, re->status); |
2467 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", straddr, | |
d62a17ae | 2468 | re->nexthop_num, re->nexthop_active_num); |
2469 | ||
7ee30f28 | 2470 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
2d68a0f2 DS |
2471 | struct interface *ifp; |
2472 | struct vrf *vrf = vrf_lookup_by_id(nexthop->vrf_id); | |
2473 | ||
2474 | switch (nexthop->type) { | |
2475 | case NEXTHOP_TYPE_BLACKHOLE: | |
53c16fbe | 2476 | sprintf(nhname, "Blackhole"); |
2d68a0f2 DS |
2477 | break; |
2478 | case NEXTHOP_TYPE_IFINDEX: | |
2479 | ifp = if_lookup_by_index(nexthop->ifindex, | |
2480 | nexthop->vrf_id); | |
53c16fbe | 2481 | sprintf(nhname, "%s", ifp ? ifp->name : "Unknown"); |
2d68a0f2 DS |
2482 | break; |
2483 | case NEXTHOP_TYPE_IPV4: | |
2484 | /* fallthrough */ | |
2485 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
53c16fbe | 2486 | inet_ntop(AF_INET, &nexthop->gate, nhname, |
2d68a0f2 DS |
2487 | INET6_ADDRSTRLEN); |
2488 | break; | |
2489 | case NEXTHOP_TYPE_IPV6: | |
2490 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
53c16fbe | 2491 | inet_ntop(AF_INET6, &nexthop->gate, nhname, |
2d68a0f2 DS |
2492 | INET6_ADDRSTRLEN); |
2493 | break; | |
2494 | } | |
df38b099 | 2495 | zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s%s%s%s", |
53c16fbe | 2496 | straddr, (nexthop->rparent ? " NH" : "NH"), nhname, |
2d68a0f2 DS |
2497 | nexthop->ifindex, vrf ? vrf->name : "Unknown", |
2498 | nexthop->vrf_id, | |
d62a17ae | 2499 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) |
2500 | ? "ACTIVE " | |
2501 | : ""), | |
677c1dd5 | 2502 | (CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED) |
d62a17ae | 2503 | ? "FIB " |
2504 | : ""), | |
2505 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
df38b099 DS |
2506 | ? "RECURSIVE " |
2507 | : ""), | |
2508 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK) | |
2509 | ? "ONLINK " | |
2510 | : ""), | |
2511 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_MATCHED) | |
2512 | ? "MATCHED " | |
2513 | : ""), | |
2514 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_DUPLICATE) | |
2515 | ? "DUPLICATE " | |
d62a17ae | 2516 | : "")); |
2517 | } | |
53c16fbe | 2518 | zlog_debug("%s: dump complete", straddr); |
dc95824a DO |
2519 | } |
2520 | ||
2521 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2522 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2523 | */ |
2524 | ||
d62a17ae | 2525 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2526 | { | |
2527 | struct route_table *table; | |
2528 | struct route_node *rn; | |
2529 | struct route_entry *re; | |
2530 | char prefix_buf[INET_ADDRSTRLEN]; | |
2531 | ||
2532 | /* Lookup table. */ | |
2533 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2534 | if (!table) { | |
e914ccbe | 2535 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2536 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2537 | vrf_id); | |
d62a17ae | 2538 | return; |
2539 | } | |
2540 | ||
2541 | /* Scan the RIB table for exactly matching RE entry. */ | |
2542 | rn = route_node_lookup(table, (struct prefix *)p); | |
2543 | ||
2544 | /* No route for this prefix. */ | |
2545 | if (!rn) { | |
32391aff | 2546 | zlog_debug("%s:%u lookup failed for %s", __func__, vrf_id, |
d62a17ae | 2547 | prefix2str((struct prefix *)p, prefix_buf, |
2548 | sizeof(prefix_buf))); | |
2549 | return; | |
2550 | } | |
2551 | ||
2552 | /* Unlock node. */ | |
2553 | route_unlock_node(rn); | |
2554 | ||
2555 | /* let's go */ | |
a2addae8 | 2556 | RNODE_FOREACH_RE (rn, re) { |
32391aff DS |
2557 | zlog_debug("%s:%u rn %p, re %p: %s, %s", |
2558 | __func__, vrf_id, | |
2559 | (void *)rn, (void *)re, | |
d62a17ae | 2560 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) |
2561 | ? "removed" | |
2562 | : "NOT removed"), | |
2563 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2564 | ? "selected" | |
2565 | : "NOT selected")); | |
2566 | route_entry_dump(p, NULL, re); | |
2567 | } | |
dc95824a DO |
2568 | } |
2569 | ||
20e5ff0a DO |
2570 | /* Check if requested address assignment will fail due to another |
2571 | * route being installed by zebra in FIB already. Take necessary | |
2572 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2573 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2574 | */ |
d62a17ae | 2575 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2576 | { | |
2577 | struct route_table *table; | |
2578 | struct route_node *rn; | |
5f7a4718 | 2579 | rib_dest_t *dest; |
d62a17ae | 2580 | |
2581 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
e914ccbe | 2582 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2583 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2584 | vrf_id); | |
d62a17ae | 2585 | return; |
2586 | } | |
2587 | ||
2588 | /* No matches would be the simplest case. */ | |
2589 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2590 | return; | |
2591 | ||
2592 | /* Unlock node. */ | |
2593 | route_unlock_node(rn); | |
2594 | ||
5f7a4718 | 2595 | dest = rib_dest_from_rnode(rn); |
d62a17ae | 2596 | /* Check all RE entries. In case any changes have to be done, requeue |
2597 | * the RN into RIBQ head. If the routing message about the new connected | |
2598 | * route (generated by the IP address we are going to assign very soon) | |
2599 | * comes before the RIBQ is processed, the new RE entry will join | |
2600 | * RIBQ record already on head. This is necessary for proper | |
2601 | * revalidation | |
2602 | * of the rest of the RE. | |
2603 | */ | |
3cdba47a | 2604 | if (dest->selected_fib) { |
5f7a4718 DS |
2605 | if (IS_ZEBRA_DEBUG_RIB) { |
2606 | char buf[PREFIX_STRLEN]; | |
2607 | ||
2608 | zlog_debug("%u:%s: freeing way for connected prefix", | |
2609 | dest->selected_fib->vrf_id, | |
2610 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2611 | route_entry_dump(&rn->p, NULL, dest->selected_fib); | |
d62a17ae | 2612 | } |
5f7a4718 | 2613 | rib_uninstall(rn, dest->selected_fib); |
d62a17ae | 2614 | rib_queue_add(rn); |
ade4a886 | 2615 | } |
20e5ff0a DO |
2616 | } |
2617 | ||
d62a17ae | 2618 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
2619 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
718e3744 | 2620 | { |
d62a17ae | 2621 | struct route_table *table; |
2622 | struct route_node *rn; | |
40ecd8e4 | 2623 | struct route_entry *same = NULL; |
d62a17ae | 2624 | int ret = 0; |
b4c034b0 | 2625 | |
d62a17ae | 2626 | if (!re) |
2627 | return 0; | |
b4c034b0 | 2628 | |
1f610a1f | 2629 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
05737783 | 2630 | |
d62a17ae | 2631 | /* Lookup table. */ |
7865c65d RW |
2632 | table = zebra_vrf_table_with_table_id(afi, safi, re->vrf_id, re->table); |
2633 | if (!table) { | |
2634 | XFREE(MTYPE_RE, re); | |
d62a17ae | 2635 | return 0; |
7865c65d | 2636 | } |
cddf391b | 2637 | |
d62a17ae | 2638 | /* Make it sure prefixlen is applied to the prefix. */ |
2639 | apply_mask(p); | |
2640 | if (src_p) | |
2641 | apply_mask_ipv6(src_p); | |
718e3744 | 2642 | |
d62a17ae | 2643 | /* Set default distance by route type. */ |
eea2ef07 | 2644 | if (re->distance == 0) |
0492eea0 | 2645 | re->distance = route_distance(re->type); |
718e3744 | 2646 | |
d62a17ae | 2647 | /* Lookup route node.*/ |
2648 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2649 | |
40ecd8e4 DS |
2650 | /* |
2651 | * If same type of route are installed, treat it as a implicit | |
2652 | * withdraw. | |
2653 | * If the user has specified the No route replace semantics | |
2654 | * for the install don't do a route replace. | |
2655 | */ | |
a2addae8 | 2656 | RNODE_FOREACH_RE (rn, same) { |
d62a17ae | 2657 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) |
2658 | continue; | |
41ec9222 | 2659 | |
eb327fa5 RW |
2660 | if (same->type != re->type) |
2661 | continue; | |
2662 | if (same->instance != re->instance) | |
2663 | continue; | |
996c9314 LB |
2664 | if (same->type == ZEBRA_ROUTE_KERNEL |
2665 | && same->metric != re->metric) | |
eb327fa5 | 2666 | continue; |
40ecd8e4 DS |
2667 | |
2668 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && | |
2669 | same->distance != re->distance) | |
2670 | continue; | |
2671 | ||
844b3a87 | 2672 | /* |
40ecd8e4 DS |
2673 | * We should allow duplicate connected routes |
2674 | * because of IPv6 link-local routes and unnumbered | |
2675 | * interfaces on Linux. | |
844b3a87 RW |
2676 | */ |
2677 | if (same->type != ZEBRA_ROUTE_CONNECT) | |
d62a17ae | 2678 | break; |
2679 | } | |
718e3744 | 2680 | |
cf363e1b MS |
2681 | /* If this route is kernel/connected route, notify the dataplane. */ |
2682 | if (RIB_SYSTEM_ROUTE(re)) { | |
2683 | /* Notify dataplane */ | |
2684 | dplane_sys_route_add(rn, re); | |
2685 | } | |
2686 | ||
d62a17ae | 2687 | /* Link new re to node.*/ |
2688 | if (IS_ZEBRA_DEBUG_RIB) { | |
2da33d6b DS |
2689 | rnode_debug(rn, re->vrf_id, |
2690 | "Inserting route rn %p, re %p (%s) existing %p", | |
2691 | rn, re, zebra_route_string(re->type), same); | |
718e3744 | 2692 | |
d62a17ae | 2693 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2694 | route_entry_dump(p, src_p, re); | |
718e3744 | 2695 | } |
6883bf8d DS |
2696 | |
2697 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 2698 | rib_addnode(rn, re, 1); |
2699 | ret = 1; | |
6b0655a2 | 2700 | |
d62a17ae | 2701 | /* Free implicit route.*/ |
2702 | if (same) { | |
2703 | rib_delnode(rn, same); | |
2704 | ret = -1; | |
2705 | } | |
718e3744 | 2706 | |
d62a17ae | 2707 | route_unlock_node(rn); |
2708 | return ret; | |
2709 | } | |
2710 | ||
2711 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
d7c0a89a | 2712 | unsigned short instance, int flags, struct prefix *p, |
fd36be7e | 2713 | struct prefix_ipv6 *src_p, const struct nexthop *nh, |
40ecd8e4 DS |
2714 | uint32_t table_id, uint32_t metric, uint8_t distance, |
2715 | bool fromkernel) | |
d62a17ae | 2716 | { |
2717 | struct route_table *table; | |
2718 | struct route_node *rn; | |
2719 | struct route_entry *re; | |
2720 | struct route_entry *fib = NULL; | |
2721 | struct route_entry *same = NULL; | |
fd36be7e | 2722 | struct nexthop *rtnh; |
d62a17ae | 2723 | char buf2[INET6_ADDRSTRLEN]; |
5f7a4718 | 2724 | rib_dest_t *dest; |
d62a17ae | 2725 | |
1f610a1f | 2726 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
d62a17ae | 2727 | |
2728 | /* Lookup table. */ | |
2729 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2730 | if (!table) | |
2731 | return; | |
2732 | ||
2733 | /* Apply mask. */ | |
2734 | apply_mask(p); | |
2735 | if (src_p) | |
2736 | apply_mask_ipv6(src_p); | |
2737 | ||
2738 | /* Lookup route node. */ | |
2739 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2740 | if (!rn) { | |
2741 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2742 | ||
2743 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2744 | if (src_p && src_p->prefixlen) | |
2745 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2746 | else | |
2747 | src_buf[0] = '\0'; | |
2748 | ||
82d7d0e2 DS |
2749 | if (IS_ZEBRA_DEBUG_RIB) { |
2750 | struct vrf *vrf = vrf_lookup_by_id(vrf_id); | |
2751 | ||
2752 | zlog_debug("%s[%d]:%s%s%s doesn't exist in rib", | |
2753 | vrf->name, table_id, dst_buf, | |
d62a17ae | 2754 | (src_buf[0] != '\0') ? " from " : "", |
2755 | src_buf); | |
82d7d0e2 | 2756 | } |
d62a17ae | 2757 | return; |
2758 | } | |
718e3744 | 2759 | |
5f7a4718 DS |
2760 | dest = rib_dest_from_rnode(rn); |
2761 | fib = dest->selected_fib; | |
2762 | ||
d62a17ae | 2763 | /* Lookup same type route. */ |
a2addae8 | 2764 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2765 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2766 | continue; | |
2767 | ||
d62a17ae | 2768 | if (re->type != type) |
2769 | continue; | |
2770 | if (re->instance != instance) | |
2771 | continue; | |
40ecd8e4 DS |
2772 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && |
2773 | distance != re->distance) | |
2774 | continue; | |
2775 | ||
996c9314 | 2776 | if (re->type == ZEBRA_ROUTE_KERNEL && re->metric != metric) |
f19435a8 | 2777 | continue; |
7ee30f28 | 2778 | if (re->type == ZEBRA_ROUTE_CONNECT && (rtnh = re->ng.nexthop) |
fd36be7e DL |
2779 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { |
2780 | if (rtnh->ifindex != nh->ifindex) | |
d62a17ae | 2781 | continue; |
d62a17ae | 2782 | same = re; |
2783 | break; | |
2784 | } | |
2785 | /* Make sure that the route found has the same gateway. */ | |
2786 | else { | |
fd36be7e | 2787 | if (nh == NULL) { |
d62a17ae | 2788 | same = re; |
2789 | break; | |
2790 | } | |
7ee30f28 | 2791 | for (ALL_NEXTHOPS(re->ng, rtnh)) |
a5a2d802 SW |
2792 | /* |
2793 | * No guarantee all kernel send nh with labels | |
2794 | * on delete. | |
2795 | */ | |
2796 | if (nexthop_same_no_labels(rtnh, nh)) { | |
d62a17ae | 2797 | same = re; |
2798 | break; | |
2799 | } | |
2800 | if (same) | |
2801 | break; | |
2802 | } | |
2803 | } | |
2804 | /* If same type of route can't be found and this message is from | |
2805 | kernel. */ | |
2806 | if (!same) { | |
5dfeba19 DS |
2807 | /* |
2808 | * In the past(HA!) we could get here because | |
2809 | * we were receiving a route delete from the | |
2810 | * kernel and we're not marking the proto | |
2811 | * as coming from it's appropriate originator. | |
2812 | * Now that we are properly noticing the fact | |
2813 | * that the kernel has deleted our route we | |
2814 | * are not going to get called in this path | |
2815 | * I am going to leave this here because | |
2816 | * this might still work this way on non-linux | |
2817 | * platforms as well as some weird state I have | |
2818 | * not properly thought of yet. | |
2819 | * If we can show that this code path is | |
2820 | * dead then we can remove it. | |
2821 | */ | |
b8faa875 | 2822 | if (fib && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { |
d62a17ae | 2823 | if (IS_ZEBRA_DEBUG_RIB) { |
2da33d6b DS |
2824 | rnode_debug(rn, vrf_id, |
2825 | "rn %p, re %p (%s) was deleted from kernel, adding", | |
2826 | rn, fib, | |
2827 | zebra_route_string(fib->type)); | |
d62a17ae | 2828 | } |
2829 | if (allow_delete) { | |
677c1dd5 | 2830 | UNSET_FLAG(fib->status, ROUTE_ENTRY_INSTALLED); |
d62a17ae | 2831 | /* Unset flags. */ |
7ee30f28 | 2832 | for (rtnh = fib->ng.nexthop; rtnh; |
fd36be7e DL |
2833 | rtnh = rtnh->next) |
2834 | UNSET_FLAG(rtnh->flags, | |
d62a17ae | 2835 | NEXTHOP_FLAG_FIB); |
2836 | ||
ed216282 DS |
2837 | /* |
2838 | * This is a non FRR route | |
2839 | * as such we should mark | |
2840 | * it as deleted | |
2841 | */ | |
5f7a4718 | 2842 | dest->selected_fib = NULL; |
d62a17ae | 2843 | } else { |
2844 | /* This means someone else, other than Zebra, | |
2845 | * has deleted | |
2846 | * a Zebra router from the kernel. We will add | |
2847 | * it back */ | |
2848 | rib_install_kernel(rn, fib, NULL); | |
2849 | } | |
2850 | } else { | |
2851 | if (IS_ZEBRA_DEBUG_RIB) { | |
fd36be7e | 2852 | if (nh) |
d62a17ae | 2853 | rnode_debug( |
2854 | rn, vrf_id, | |
2855 | "via %s ifindex %d type %d " | |
2856 | "doesn't exist in rib", | |
36228974 | 2857 | inet_ntop(afi2family(afi), |
2858 | &nh->gate, buf2, | |
2859 | sizeof(buf2)), | |
2860 | nh->ifindex, type); | |
d62a17ae | 2861 | else |
2862 | rnode_debug( | |
2863 | rn, vrf_id, | |
fd36be7e DL |
2864 | "type %d doesn't exist in rib", |
2865 | type); | |
d62a17ae | 2866 | } |
2867 | route_unlock_node(rn); | |
2868 | return; | |
2869 | } | |
2870 | } | |
718e3744 | 2871 | |
5dfeba19 | 2872 | if (same) { |
996c9314 LB |
2873 | if (fromkernel && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE) |
2874 | && !allow_delete) { | |
5dfeba19 DS |
2875 | rib_install_kernel(rn, same, NULL); |
2876 | route_unlock_node(rn); | |
2877 | ||
2878 | return; | |
2879 | } | |
6134fd82 | 2880 | |
2b83602b | 2881 | /* Special handling for IPv4 or IPv6 routes sourced from |
2882 | * EVPN - the nexthop (and associated MAC) need to be | |
2883 | * uninstalled if no more refs. | |
2884 | */ | |
90264d64 | 2885 | if (CHECK_FLAG(flags, ZEBRA_FLAG_EVPN_ROUTE)) { |
6134fd82 | 2886 | struct nexthop *tmp_nh; |
2887 | ||
7ee30f28 | 2888 | for (ALL_NEXTHOPS(re->ng, tmp_nh)) { |
6134fd82 | 2889 | struct ipaddr vtep_ip; |
2890 | ||
2891 | memset(&vtep_ip, 0, sizeof(struct ipaddr)); | |
1ec31309 | 2892 | if (afi == AFI_IP) { |
2893 | vtep_ip.ipa_type = IPADDR_V4; | |
2894 | memcpy(&(vtep_ip.ipaddr_v4), | |
2895 | &(tmp_nh->gate.ipv4), | |
2896 | sizeof(struct in_addr)); | |
2897 | } else { | |
2898 | vtep_ip.ipa_type = IPADDR_V6; | |
2899 | memcpy(&(vtep_ip.ipaddr_v6), | |
2900 | &(tmp_nh->gate.ipv6), | |
2901 | sizeof(struct in6_addr)); | |
2902 | } | |
a317a9b9 | 2903 | zebra_vxlan_evpn_vrf_route_del(re->vrf_id, |
6134fd82 | 2904 | &vtep_ip, p); |
2905 | } | |
2906 | } | |
cf363e1b MS |
2907 | |
2908 | /* Notify dplane if system route changes */ | |
2909 | if (RIB_SYSTEM_ROUTE(re)) | |
2910 | dplane_sys_route_del(rn, same); | |
2911 | ||
d62a17ae | 2912 | rib_delnode(rn, same); |
5dfeba19 | 2913 | } |
05737783 | 2914 | |
d62a17ae | 2915 | route_unlock_node(rn); |
2916 | return; | |
2917 | } | |
718e3744 | 2918 | |
718e3744 | 2919 | |
d7c0a89a QY |
2920 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, |
2921 | unsigned short instance, int flags, struct prefix *p, | |
2922 | struct prefix_ipv6 *src_p, const struct nexthop *nh, | |
2923 | uint32_t table_id, uint32_t metric, uint32_t mtu, uint8_t distance, | |
2924 | route_tag_t tag) | |
d62a17ae | 2925 | { |
2926 | struct route_entry *re; | |
66af6845 | 2927 | struct nexthop *nexthop; |
718e3744 | 2928 | |
66af6845 | 2929 | /* Allocate new route_entry structure. */ |
d62a17ae | 2930 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); |
d62a17ae | 2931 | re->type = type; |
2932 | re->instance = instance; | |
2933 | re->distance = distance; | |
2934 | re->flags = flags; | |
2935 | re->metric = metric; | |
2936 | re->mtu = mtu; | |
2937 | re->table = table_id; | |
2938 | re->vrf_id = vrf_id; | |
2939 | re->nexthop_num = 0; | |
98572489 | 2940 | re->uptime = monotime(NULL); |
4e40b6d6 | 2941 | re->tag = tag; |
d62a17ae | 2942 | |
66af6845 RW |
2943 | /* Add nexthop. */ |
2944 | nexthop = nexthop_new(); | |
2945 | *nexthop = *nh; | |
2946 | route_entry_nexthop_add(re, nexthop); | |
718e3744 | 2947 | |
66af6845 | 2948 | return rib_add_multipath(afi, safi, p, src_p, re); |
718e3744 | 2949 | } |
2950 | ||
1c848137 | 2951 | /* Schedule routes of a particular table (address-family) based on event. */ |
d5b8c216 | 2952 | void rib_update_table(struct route_table *table, rib_update_event_t event) |
d62a17ae | 2953 | { |
2954 | struct route_node *rn; | |
2955 | struct route_entry *re, *next; | |
2956 | ||
2957 | /* Walk all routes and queue for processing, if appropriate for | |
2958 | * the trigger event. | |
2959 | */ | |
2960 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
1ca60f2c DS |
2961 | /* |
2962 | * If we are looking at a route node and the node | |
2963 | * has already been queued we don't | |
2964 | * need to queue it up again | |
2965 | */ | |
996c9314 LB |
2966 | if (rn->info && CHECK_FLAG(rib_dest_from_rnode(rn)->flags, |
2967 | RIB_ROUTE_ANY_QUEUED)) | |
1ca60f2c | 2968 | continue; |
d62a17ae | 2969 | switch (event) { |
2970 | case RIB_UPDATE_IF_CHANGE: | |
2971 | /* Examine all routes that won't get processed by the | |
2972 | * protocol or | |
2973 | * triggered by nexthop evaluation (NHT). This would be | |
2974 | * system, | |
2975 | * kernel and certain static routes. Note that NHT will | |
2976 | * get | |
2977 | * triggered upon an interface event as connected routes | |
2978 | * always | |
2979 | * get queued for processing. | |
2980 | */ | |
a2addae8 | 2981 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
0a16efff DS |
2982 | struct nexthop *nh; |
2983 | ||
996c9314 LB |
2984 | if (re->type != ZEBRA_ROUTE_SYSTEM |
2985 | && re->type != ZEBRA_ROUTE_KERNEL | |
2986 | && re->type != ZEBRA_ROUTE_CONNECT | |
2987 | && re->type != ZEBRA_ROUTE_STATIC) | |
0a16efff DS |
2988 | continue; |
2989 | ||
2990 | if (re->type != ZEBRA_ROUTE_STATIC) { | |
ade4a886 DS |
2991 | SET_FLAG(re->status, |
2992 | ROUTE_ENTRY_CHANGED); | |
0a16efff DS |
2993 | rib_queue_add(rn); |
2994 | continue; | |
2995 | } | |
2996 | ||
7ee30f28 | 2997 | for (nh = re->ng.nexthop; nh; nh = nh->next) |
0a16efff DS |
2998 | if (!(nh->type == NEXTHOP_TYPE_IPV4 |
2999 | || nh->type == NEXTHOP_TYPE_IPV6)) | |
3000 | break; | |
3001 | ||
3002 | /* If we only have nexthops to a | |
3003 | * gateway, NHT will | |
3004 | * take care. | |
3005 | */ | |
ade4a886 DS |
3006 | if (nh) { |
3007 | SET_FLAG(re->status, | |
3008 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 3009 | rib_queue_add(rn); |
ade4a886 | 3010 | } |
d62a17ae | 3011 | } |
3012 | break; | |
3013 | ||
3014 | case RIB_UPDATE_RMAP_CHANGE: | |
3015 | case RIB_UPDATE_OTHER: | |
3016 | /* Right now, examine all routes. Can restrict to a | |
3017 | * protocol in | |
3018 | * some cases (TODO). | |
3019 | */ | |
ade4a886 DS |
3020 | if (rnode_to_ribs(rn)) { |
3021 | RNODE_FOREACH_RE_SAFE (rn, re, next) | |
3022 | SET_FLAG(re->status, | |
3023 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 3024 | rib_queue_add(rn); |
ade4a886 | 3025 | } |
d62a17ae | 3026 | break; |
3027 | ||
3028 | default: | |
3029 | break; | |
3030 | } | |
3031 | } | |
b84c7253 | 3032 | } |
3033 | ||
718e3744 | 3034 | /* RIB update function. */ |
d62a17ae | 3035 | void rib_update(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 3036 | { |
d62a17ae | 3037 | struct route_table *table; |
1c848137 | 3038 | |
d62a17ae | 3039 | /* Process routes of interested address-families. */ |
3040 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
d5b8c216 | 3041 | if (table) { |
3042 | if (IS_ZEBRA_DEBUG_EVENT) | |
3043 | zlog_debug("%s : AFI_IP event %d", __func__, event); | |
d62a17ae | 3044 | rib_update_table(table, event); |
d5b8c216 | 3045 | } |
718e3744 | 3046 | |
d62a17ae | 3047 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
d5b8c216 | 3048 | if (table) { |
3049 | if (IS_ZEBRA_DEBUG_EVENT) | |
3050 | zlog_debug("%s : AFI_IP6 event %d", __func__, event); | |
d62a17ae | 3051 | rib_update_table(table, event); |
d5b8c216 | 3052 | } |
718e3744 | 3053 | } |
3054 | ||
718e3744 | 3055 | /* Delete self installed routes after zebra is relaunched. */ |
95a29032 | 3056 | void rib_sweep_table(struct route_table *table) |
d62a17ae | 3057 | { |
3058 | struct route_node *rn; | |
3059 | struct route_entry *re; | |
3060 | struct route_entry *next; | |
915902cb | 3061 | struct nexthop *nexthop; |
d62a17ae | 3062 | |
915902cb DS |
3063 | if (!table) |
3064 | return; | |
d62a17ae | 3065 | |
915902cb | 3066 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
a2addae8 | 3067 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d4644d41 | 3068 | |
915902cb DS |
3069 | if (IS_ZEBRA_DEBUG_RIB) |
3070 | route_entry_dump(&rn->p, NULL, re); | |
3071 | ||
3072 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
3073 | continue; | |
3074 | ||
3075 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
3076 | continue; | |
3077 | ||
d4644d41 DS |
3078 | /* |
3079 | * If routes are older than startup_time then | |
3080 | * we know we read them in from the kernel. | |
3081 | * As such we can safely remove them. | |
3082 | */ | |
3083 | if (zrouter.startup_time < re->uptime) | |
3084 | continue; | |
3085 | ||
915902cb DS |
3086 | /* |
3087 | * So we are starting up and have received | |
3088 | * routes from the kernel that we have installed | |
3089 | * from a previous run of zebra but not cleaned | |
3090 | * up ( say a kill -9 ) | |
3091 | * But since we haven't actually installed | |
3092 | * them yet( we received them from the kernel ) | |
3093 | * we don't think they are active. | |
3094 | * So let's pretend they are active to actually | |
3095 | * remove them. | |
3096 | * In all honesty I'm not sure if we should | |
3097 | * mark them as active when we receive them | |
3098 | * This is startup only so probably ok. | |
3099 | * | |
3100 | * If we ever decide to move rib_sweep_table | |
3101 | * to a different spot (ie startup ) | |
3102 | * this decision needs to be revisited | |
3103 | */ | |
677c1dd5 | 3104 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
7ee30f28 | 3105 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
915902cb DS |
3106 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
3107 | ||
0c555cc6 DS |
3108 | rib_uninstall_kernel(rn, re); |
3109 | rib_delnode(rn, re); | |
915902cb DS |
3110 | } |
3111 | } | |
718e3744 | 3112 | } |
3113 | ||
3114 | /* Sweep all RIB tables. */ | |
d4644d41 | 3115 | int rib_sweep_route(struct thread *t) |
718e3744 | 3116 | { |
d62a17ae | 3117 | struct vrf *vrf; |
3118 | struct zebra_vrf *zvrf; | |
78104b9b | 3119 | |
a2addae8 | 3120 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
915902cb DS |
3121 | if ((zvrf = vrf->info) == NULL) |
3122 | continue; | |
3123 | ||
d62a17ae | 3124 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); |
3125 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3126 | } | |
95a29032 | 3127 | |
89272910 | 3128 | zebra_router_sweep_route(); |
d4644d41 DS |
3129 | |
3130 | return 0; | |
718e3744 | 3131 | } |
2ea1ab1c VT |
3132 | |
3133 | /* Remove specific by protocol routes from 'table'. */ | |
d7c0a89a | 3134 | unsigned long rib_score_proto_table(uint8_t proto, unsigned short instance, |
47a08aa9 | 3135 | struct route_table *table) |
d62a17ae | 3136 | { |
3137 | struct route_node *rn; | |
3138 | struct route_entry *re; | |
3139 | struct route_entry *next; | |
3140 | unsigned long n = 0; | |
3141 | ||
3142 | if (table) | |
3143 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
a2addae8 | 3144 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 3145 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
3146 | continue; | |
3147 | if (re->type == proto | |
3148 | && re->instance == instance) { | |
3149 | rib_delnode(rn, re); | |
3150 | n++; | |
3151 | } | |
3152 | } | |
3153 | return n; | |
2ea1ab1c VT |
3154 | } |
3155 | ||
3156 | /* Remove specific by protocol routes. */ | |
d7c0a89a | 3157 | unsigned long rib_score_proto(uint8_t proto, unsigned short instance) |
2ea1ab1c | 3158 | { |
d62a17ae | 3159 | struct vrf *vrf; |
3160 | struct zebra_vrf *zvrf; | |
d8612e65 | 3161 | struct other_route_table *ort; |
d62a17ae | 3162 | unsigned long cnt = 0; |
78104b9b | 3163 | |
d8612e65 DS |
3164 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
3165 | zvrf = vrf->info; | |
3166 | if (!zvrf) | |
3167 | continue; | |
78104b9b | 3168 | |
d8612e65 DS |
3169 | cnt += rib_score_proto_table(proto, instance, |
3170 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
3171 | + rib_score_proto_table( | |
3172 | proto, instance, | |
3173 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3174 | ||
81fddbe7 | 3175 | frr_each(otable, &zvrf->other_tables, ort) cnt += |
d8612e65 DS |
3176 | rib_score_proto_table(proto, instance, ort->table); |
3177 | } | |
47a08aa9 | 3178 | |
d62a17ae | 3179 | return cnt; |
2ea1ab1c VT |
3180 | } |
3181 | ||
718e3744 | 3182 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 3183 | void rib_close_table(struct route_table *table) |
718e3744 | 3184 | { |
d62a17ae | 3185 | struct route_node *rn; |
1e9f448f | 3186 | rib_table_info_t *info; |
5f7a4718 | 3187 | rib_dest_t *dest; |
718e3744 | 3188 | |
1e9f448f DS |
3189 | if (!table) |
3190 | return; | |
9fd92e3c | 3191 | |
6ca30e9e | 3192 | info = route_table_get_info(table); |
5adc2528 | 3193 | |
5f7a4718 DS |
3194 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
3195 | dest = rib_dest_from_rnode(rn); | |
1e9f448f | 3196 | |
5f7a4718 | 3197 | if (dest && dest->selected_fib) { |
1e9f448f DS |
3198 | if (info->safi == SAFI_UNICAST) |
3199 | hook_call(rib_update, rn, NULL); | |
3200 | ||
3cdba47a DS |
3201 | rib_uninstall_kernel(rn, dest->selected_fib); |
3202 | dest->selected_fib = NULL; | |
1e9f448f | 3203 | } |
5f7a4718 | 3204 | } |
718e3744 | 3205 | } |
3206 | ||
9bd9717b MS |
3207 | /* |
3208 | * Handler for async dataplane results after a pseudowire installation | |
3209 | */ | |
3210 | static int handle_pw_result(struct zebra_dplane_ctx *ctx) | |
3211 | { | |
9bd9717b MS |
3212 | struct zebra_pw *pw; |
3213 | struct zebra_vrf *vrf; | |
3214 | ||
3215 | /* The pseudowire code assumes success - we act on an error | |
3216 | * result for installation attempts here. | |
3217 | */ | |
3218 | if (dplane_ctx_get_op(ctx) != DPLANE_OP_PW_INSTALL) | |
3219 | goto done; | |
3220 | ||
3221 | if (dplane_ctx_get_status(ctx) != ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
3222 | vrf = zebra_vrf_lookup_by_id(dplane_ctx_get_vrf(ctx)); | |
3223 | pw = zebra_pw_find(vrf, dplane_ctx_get_pw_ifname(ctx)); | |
3224 | if (pw) | |
3225 | zebra_pw_install_failure(pw); | |
3226 | } | |
3227 | ||
3228 | done: | |
3229 | ||
9f2d0354 | 3230 | return 0; |
9bd9717b MS |
3231 | } |
3232 | ||
3233 | ||
7cdb1a84 | 3234 | /* |
d37f4d6c MS |
3235 | * Handle results from the dataplane system. Dequeue update context |
3236 | * structs, dispatch to appropriate internal handlers. | |
7cdb1a84 MS |
3237 | */ |
3238 | static int rib_process_dplane_results(struct thread *thread) | |
3239 | { | |
25779064 | 3240 | struct zebra_dplane_ctx *ctx; |
4c206c8f MS |
3241 | struct dplane_ctx_q ctxlist; |
3242 | ||
3243 | /* Dequeue a list of completed updates with one lock/unlock cycle */ | |
7cdb1a84 MS |
3244 | |
3245 | do { | |
4c206c8f MS |
3246 | TAILQ_INIT(&ctxlist); |
3247 | ||
7cdb1a84 MS |
3248 | /* Take lock controlling queue of results */ |
3249 | pthread_mutex_lock(&dplane_mutex); | |
3250 | { | |
d37f4d6c | 3251 | /* Dequeue list of context structs */ |
4c206c8f | 3252 | dplane_ctx_list_append(&ctxlist, &rib_dplane_q); |
7cdb1a84 MS |
3253 | } |
3254 | pthread_mutex_unlock(&dplane_mutex); | |
3255 | ||
4c206c8f MS |
3256 | /* Dequeue context block */ |
3257 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3258 | ||
3259 | /* If we've emptied the results queue, we're done */ | |
3260 | if (ctx == NULL) | |
7cdb1a84 | 3261 | break; |
7cdb1a84 | 3262 | |
4c206c8f | 3263 | while (ctx) { |
d37f4d6c MS |
3264 | switch (dplane_ctx_get_op(ctx)) { |
3265 | case DPLANE_OP_ROUTE_INSTALL: | |
3266 | case DPLANE_OP_ROUTE_UPDATE: | |
3267 | case DPLANE_OP_ROUTE_DELETE: | |
188a00e0 MS |
3268 | { |
3269 | /* Bit of special case for route updates | |
3270 | * that were generated by async notifications: | |
3271 | * we don't want to continue processing these | |
3272 | * in the rib. | |
3273 | */ | |
3274 | if (dplane_ctx_get_notif_provider(ctx) == 0) | |
3275 | rib_process_result(ctx); | |
3276 | else | |
3277 | dplane_ctx_fini(&ctx); | |
3278 | } | |
3279 | break; | |
d37f4d6c | 3280 | |
54818e3b MS |
3281 | case DPLANE_OP_ROUTE_NOTIFY: |
3282 | rib_process_dplane_notify(ctx); | |
d37f4d6c MS |
3283 | break; |
3284 | ||
3285 | case DPLANE_OP_LSP_INSTALL: | |
3286 | case DPLANE_OP_LSP_UPDATE: | |
3287 | case DPLANE_OP_LSP_DELETE: | |
188a00e0 MS |
3288 | { |
3289 | /* Bit of special case for LSP updates | |
3290 | * that were generated by async notifications: | |
3291 | * we don't want to continue processing these. | |
3292 | */ | |
3293 | if (dplane_ctx_get_notif_provider(ctx) == 0) | |
3294 | zebra_mpls_lsp_dplane_result(ctx); | |
3295 | else | |
3296 | dplane_ctx_fini(&ctx); | |
3297 | } | |
3298 | break; | |
d37f4d6c | 3299 | |
104e3ad9 MS |
3300 | case DPLANE_OP_LSP_NOTIFY: |
3301 | zebra_mpls_process_dplane_notify(ctx); | |
d37f4d6c MS |
3302 | break; |
3303 | ||
9bd9717b MS |
3304 | case DPLANE_OP_PW_INSTALL: |
3305 | case DPLANE_OP_PW_UNINSTALL: | |
3306 | handle_pw_result(ctx); | |
3307 | break; | |
3308 | ||
cf363e1b MS |
3309 | case DPLANE_OP_SYS_ROUTE_ADD: |
3310 | case DPLANE_OP_SYS_ROUTE_DELETE: | |
3311 | /* No further processing in zebra for these. */ | |
3312 | dplane_ctx_fini(&ctx); | |
3313 | break; | |
3314 | ||
d37f4d6c MS |
3315 | default: |
3316 | /* Don't expect this: just return the struct? */ | |
3317 | dplane_ctx_fini(&ctx); | |
3318 | break; | |
3319 | } /* Dispatch by op code */ | |
4c206c8f MS |
3320 | |
3321 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3322 | } | |
3323 | ||
5709131c | 3324 | } while (1); |
7cdb1a84 | 3325 | |
f183e380 MS |
3326 | /* Check for nexthop tracking processing after finishing with results */ |
3327 | do_nht_processing(); | |
3328 | ||
5709131c | 3329 | return 0; |
7cdb1a84 MS |
3330 | } |
3331 | ||
3332 | /* | |
3333 | * Results are returned from the dataplane subsystem, in the context of | |
1bcea841 | 3334 | * the dataplane pthread. We enqueue the results here for processing by |
7cdb1a84 MS |
3335 | * the main thread later. |
3336 | */ | |
4c206c8f | 3337 | static int rib_dplane_results(struct dplane_ctx_q *ctxlist) |
7cdb1a84 MS |
3338 | { |
3339 | /* Take lock controlling queue of results */ | |
3340 | pthread_mutex_lock(&dplane_mutex); | |
3341 | { | |
4c206c8f MS |
3342 | /* Enqueue context blocks */ |
3343 | dplane_ctx_list_append(&rib_dplane_q, ctxlist); | |
7cdb1a84 MS |
3344 | } |
3345 | pthread_mutex_unlock(&dplane_mutex); | |
3346 | ||
4c206c8f | 3347 | /* Ensure event is signalled to zebra main pthread */ |
3801e764 | 3348 | thread_add_event(zrouter.master, rib_process_dplane_results, NULL, 0, |
7cdb1a84 MS |
3349 | &t_dplane); |
3350 | ||
5709131c | 3351 | return 0; |
7cdb1a84 MS |
3352 | } |
3353 | ||
eaa2716d SW |
3354 | /* |
3355 | * Ensure there are no empty slots in the route_info array. | |
3356 | * Every route type in zebra should be present there. | |
3357 | */ | |
3358 | static void check_route_info(void) | |
3359 | { | |
3360 | int len = array_size(route_info); | |
3361 | ||
3362 | /* | |
3363 | * ZEBRA_ROUTE_SYSTEM is special cased since | |
3364 | * its key is 0 anyway. | |
3365 | * | |
3366 | * ZEBRA_ROUTE_ALL is also ignored. | |
3367 | */ | |
3368 | for (int i = 0; i < len; i++) { | |
3369 | if (i == ZEBRA_ROUTE_SYSTEM || i == ZEBRA_ROUTE_ALL) | |
3370 | continue; | |
3371 | assert(route_info[i].key); | |
8dc7a759 | 3372 | assert(route_info[i].meta_q_map < MQ_SIZE); |
eaa2716d SW |
3373 | } |
3374 | } | |
3375 | ||
718e3744 | 3376 | /* Routing information base initialize. */ |
d62a17ae | 3377 | void rib_init(void) |
718e3744 | 3378 | { |
eaa2716d SW |
3379 | check_route_info(); |
3380 | ||
2561d12e | 3381 | rib_queue_init(); |
7cdb1a84 MS |
3382 | |
3383 | /* Init dataplane, and register for results */ | |
3384 | pthread_mutex_init(&dplane_mutex, NULL); | |
3385 | TAILQ_INIT(&rib_dplane_q); | |
4c206c8f | 3386 | zebra_dplane_init(rib_dplane_results); |
718e3744 | 3387 | } |
0915bb0c AS |
3388 | |
3389 | /* | |
3390 | * vrf_id_get_next | |
3391 | * | |
3392 | * Get the first vrf id that is greater than the given vrf id if any. | |
3393 | * | |
2951a7a4 | 3394 | * Returns true if a vrf id was found, false otherwise. |
0915bb0c | 3395 | */ |
d62a17ae | 3396 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 3397 | { |
d62a17ae | 3398 | struct vrf *vrf; |
b72ede27 | 3399 | |
d62a17ae | 3400 | vrf = vrf_lookup_by_id(vrf_id); |
3401 | if (vrf) { | |
3402 | vrf = RB_NEXT(vrf_id_head, vrf); | |
3403 | if (vrf) { | |
3404 | *next_id_p = vrf->vrf_id; | |
3405 | return 1; | |
3406 | } | |
3407 | } | |
0915bb0c | 3408 | |
d62a17ae | 3409 | return 0; |
0915bb0c AS |
3410 | } |
3411 | ||
3412 | /* | |
3413 | * rib_tables_iter_next | |
3414 | * | |
3415 | * Returns the next table in the iteration. | |
3416 | */ | |
d62a17ae | 3417 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
3418 | { | |
3419 | struct route_table *table; | |
3420 | ||
3421 | /* | |
3422 | * Array that helps us go over all AFI/SAFI combinations via one | |
3423 | * index. | |
3424 | */ | |
3425 | static struct { | |
3426 | afi_t afi; | |
3427 | safi_t safi; | |
3428 | } afi_safis[] = { | |
3429 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
3430 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
3431 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
3432 | }; | |
3433 | ||
3434 | table = NULL; | |
3435 | ||
3436 | switch (iter->state) { | |
3437 | ||
3438 | case RIB_TABLES_ITER_S_INIT: | |
3439 | iter->vrf_id = VRF_DEFAULT; | |
3440 | iter->afi_safi_ix = -1; | |
3441 | ||
3442 | /* Fall through */ | |
3443 | ||
3444 | case RIB_TABLES_ITER_S_ITERATING: | |
3445 | iter->afi_safi_ix++; | |
3446 | while (1) { | |
3447 | ||
3448 | while (iter->afi_safi_ix | |
7e3a1ec7 | 3449 | < (int)array_size(afi_safis)) { |
d62a17ae | 3450 | table = zebra_vrf_table( |
3451 | afi_safis[iter->afi_safi_ix].afi, | |
3452 | afi_safis[iter->afi_safi_ix].safi, | |
3453 | iter->vrf_id); | |
3454 | if (table) | |
3455 | break; | |
3456 | ||
3457 | iter->afi_safi_ix++; | |
3458 | } | |
3459 | ||
3460 | /* | |
3461 | * Found another table in this vrf. | |
3462 | */ | |
3463 | if (table) | |
3464 | break; | |
3465 | ||
3466 | /* | |
3467 | * Done with all tables in the current vrf, go to the | |
3468 | * next | |
3469 | * one. | |
3470 | */ | |
3471 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
3472 | break; | |
3473 | ||
3474 | iter->afi_safi_ix = 0; | |
3475 | } | |
0915bb0c | 3476 | |
0915bb0c AS |
3477 | break; |
3478 | ||
d62a17ae | 3479 | case RIB_TABLES_ITER_S_DONE: |
3480 | return NULL; | |
0915bb0c AS |
3481 | } |
3482 | ||
d62a17ae | 3483 | if (table) |
3484 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
3485 | else | |
3486 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 3487 | |
d62a17ae | 3488 | return table; |
0915bb0c | 3489 | } |