]>
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" |
deaa50db | 27 | #include "log_int.h" |
43e52561 QY |
28 | #include "memory.h" |
29 | #include "mpls.h" | |
30 | #include "nexthop.h" | |
31 | #include "prefix.h" | |
7514fb77 PJ |
32 | #include "prefix.h" |
33 | #include "routemap.h" | |
43e52561 | 34 | #include "sockunion.h" |
05737783 | 35 | #include "srcdest_table.h" |
43e52561 QY |
36 | #include "table.h" |
37 | #include "thread.h" | |
38 | #include "vrf.h" | |
39 | #include "workqueue.h" | |
718e3744 | 40 | |
43e52561 QY |
41 | #include "zebra/connected.h" |
42 | #include "zebra/debug.h" | |
43 | #include "zebra/interface.h" | |
44 | #include "zebra/redistribute.h" | |
718e3744 | 45 | #include "zebra/rib.h" |
46 | #include "zebra/rt.h" | |
43e52561 QY |
47 | #include "zebra/zapi_msg.h" |
48 | #include "zebra/zebra_errors.h" | |
49 | #include "zebra/zebra_memory.h" | |
7c551956 | 50 | #include "zebra/zebra_ns.h" |
fb018d25 | 51 | #include "zebra/zebra_rnh.h" |
43e52561 QY |
52 | #include "zebra/zebra_routemap.h" |
53 | #include "zebra/zebra_vrf.h" | |
6134fd82 | 54 | #include "zebra/zebra_vxlan.h" |
7cdb1a84 MS |
55 | #include "zebra/zapi_msg.h" |
56 | #include "zebra/zebra_dplane.h" | |
57 | ||
58 | /* | |
59 | * Event, list, and mutex for delivery of dataplane results | |
60 | */ | |
61 | static pthread_mutex_t dplane_mutex; | |
62 | static struct thread *t_dplane; | |
63 | static struct dplane_ctx_q_s rib_dplane_q; | |
718e3744 | 64 | |
d62a17ae | 65 | DEFINE_HOOK(rib_update, (struct route_node * rn, const char *reason), |
66 | (rn, reason)) | |
4f8ea50c | 67 | |
6baf7bb8 DS |
68 | /* Should we allow non Quagga processes to delete our routes */ |
69 | extern int allow_delete; | |
70 | ||
718e3744 | 71 | /* Each route type's string and default distance value. */ |
d62a17ae | 72 | static const struct { |
73 | int key; | |
74 | int distance; | |
75 | } route_info[ZEBRA_ROUTE_MAX] = { | |
9d303b37 DL |
76 | [ZEBRA_ROUTE_SYSTEM] = {ZEBRA_ROUTE_SYSTEM, 0}, |
77 | [ZEBRA_ROUTE_KERNEL] = {ZEBRA_ROUTE_KERNEL, 0}, | |
78 | [ZEBRA_ROUTE_CONNECT] = {ZEBRA_ROUTE_CONNECT, 0}, | |
79 | [ZEBRA_ROUTE_STATIC] = {ZEBRA_ROUTE_STATIC, 1}, | |
80 | [ZEBRA_ROUTE_RIP] = {ZEBRA_ROUTE_RIP, 120}, | |
81 | [ZEBRA_ROUTE_RIPNG] = {ZEBRA_ROUTE_RIPNG, 120}, | |
82 | [ZEBRA_ROUTE_OSPF] = {ZEBRA_ROUTE_OSPF, 110}, | |
83 | [ZEBRA_ROUTE_OSPF6] = {ZEBRA_ROUTE_OSPF6, 110}, | |
84 | [ZEBRA_ROUTE_ISIS] = {ZEBRA_ROUTE_ISIS, 115}, | |
85 | [ZEBRA_ROUTE_BGP] = {ZEBRA_ROUTE_BGP, 20 /* IBGP is 200. */}, | |
c710b277 DS |
86 | [ZEBRA_ROUTE_PIM] = {ZEBRA_ROUTE_PIM, 255}, |
87 | [ZEBRA_ROUTE_EIGRP] = {ZEBRA_ROUTE_EIGRP, 90}, | |
9d303b37 | 88 | [ZEBRA_ROUTE_NHRP] = {ZEBRA_ROUTE_NHRP, 10}, |
c710b277 DS |
89 | [ZEBRA_ROUTE_HSLS] = {ZEBRA_ROUTE_HSLS, 255}, |
90 | [ZEBRA_ROUTE_OLSR] = {ZEBRA_ROUTE_OLSR, 255}, | |
91 | [ZEBRA_ROUTE_TABLE] = {ZEBRA_ROUTE_TABLE, 150}, | |
92 | [ZEBRA_ROUTE_LDP] = {ZEBRA_ROUTE_LDP, 150}, | |
93 | [ZEBRA_ROUTE_VNC] = {ZEBRA_ROUTE_VNC, 20}, | |
94 | [ZEBRA_ROUTE_VNC_DIRECT] = {ZEBRA_ROUTE_VNC_DIRECT, 20}, | |
95 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = {ZEBRA_ROUTE_VNC_DIRECT_RH, 20}, | |
96 | [ZEBRA_ROUTE_BGP_DIRECT] = {ZEBRA_ROUTE_BGP_DIRECT, 20}, | |
97 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = {ZEBRA_ROUTE_BGP_DIRECT_EXT, 20}, | |
98 | [ZEBRA_ROUTE_BABEL] = {ZEBRA_ROUTE_BABEL, 100}, | |
400a663b | 99 | [ZEBRA_ROUTE_SHARP] = {ZEBRA_ROUTE_SHARP, 150}, |
c710b277 | 100 | |
d62a17ae | 101 | /* no entry/default: 150 */ |
718e3744 | 102 | }; |
6b0655a2 | 103 | |
4623d897 DL |
104 | /* RPF lookup behaviour */ |
105 | static enum multicast_mode ipv4_multicast_mode = MCAST_NO_CONFIG; | |
106 | ||
6c4f4e6e | 107 | |
d62a17ae | 108 | static void __attribute__((format(printf, 5, 6))) |
109 | _rnode_zlog(const char *_func, vrf_id_t vrf_id, struct route_node *rn, | |
110 | int priority, const char *msgfmt, ...) | |
2263a412 | 111 | { |
d62a17ae | 112 | char buf[SRCDEST2STR_BUFFER + sizeof(" (MRIB)")]; |
113 | char msgbuf[512]; | |
114 | va_list ap; | |
2263a412 | 115 | |
d62a17ae | 116 | va_start(ap, msgfmt); |
117 | vsnprintf(msgbuf, sizeof(msgbuf), msgfmt, ap); | |
118 | va_end(ap); | |
2263a412 | 119 | |
d62a17ae | 120 | if (rn) { |
121 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
122 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
cb653491 | 123 | |
d62a17ae | 124 | if (info->safi == SAFI_MULTICAST) |
125 | strcat(buf, " (MRIB)"); | |
126 | } else { | |
127 | snprintf(buf, sizeof(buf), "{(route_node *) NULL}"); | |
128 | } | |
2263a412 | 129 | |
d62a17ae | 130 | zlog(priority, "%s: %d:%s: %s", _func, vrf_id, buf, msgbuf); |
2263a412 DL |
131 | } |
132 | ||
d62a17ae | 133 | #define rnode_debug(node, vrf_id, ...) \ |
2263a412 | 134 | _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__) |
d62a17ae | 135 | #define rnode_info(node, ...) \ |
2263a412 DL |
136 | _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__) |
137 | ||
fd289fc8 | 138 | uint8_t route_distance(int type) |
40c7bdb0 | 139 | { |
fd289fc8 | 140 | uint8_t distance; |
40c7bdb0 | 141 | |
d62a17ae | 142 | if ((unsigned)type >= array_size(route_info)) |
143 | distance = 150; | |
144 | else | |
145 | distance = route_info[type].distance; | |
40c7bdb0 | 146 | |
d62a17ae | 147 | return distance; |
40c7bdb0 | 148 | } |
149 | ||
d7c0a89a | 150 | int is_zebra_valid_kernel_table(uint32_t table_id) |
7a4bb9c5 | 151 | { |
8f500a1c | 152 | #ifdef linux |
d62a17ae | 153 | if ((table_id == RT_TABLE_UNSPEC) || (table_id == RT_TABLE_LOCAL) |
154 | || (table_id == RT_TABLE_COMPAT)) | |
155 | return 0; | |
8f500a1c RW |
156 | #endif |
157 | ||
d62a17ae | 158 | return 1; |
7a4bb9c5 DS |
159 | } |
160 | ||
d7c0a89a | 161 | int is_zebra_main_routing_table(uint32_t table_id) |
7a4bb9c5 | 162 | { |
d62a17ae | 163 | if ((table_id == RT_TABLE_MAIN) |
164 | || (table_id == zebrad.rtm_table_default)) | |
165 | return 1; | |
166 | return 0; | |
7a4bb9c5 DS |
167 | } |
168 | ||
86391e56 | 169 | int zebra_check_addr(const struct prefix *p) |
0aabccc0 | 170 | { |
d62a17ae | 171 | if (p->family == AF_INET) { |
d7c0a89a | 172 | uint32_t addr; |
0aabccc0 | 173 | |
d62a17ae | 174 | addr = p->u.prefix4.s_addr; |
175 | addr = ntohl(addr); | |
0aabccc0 | 176 | |
d62a17ae | 177 | if (IPV4_NET127(addr) || IN_CLASSD(addr) |
178 | || IPV4_LINKLOCAL(addr)) | |
179 | return 0; | |
180 | } | |
181 | if (p->family == AF_INET6) { | |
182 | if (IN6_IS_ADDR_LOOPBACK(&p->u.prefix6)) | |
183 | return 0; | |
184 | if (IN6_IS_ADDR_LINKLOCAL(&p->u.prefix6)) | |
185 | return 0; | |
186 | } | |
187 | return 1; | |
0aabccc0 DD |
188 | } |
189 | ||
fa713d9e | 190 | /* Add nexthop to the end of a rib node's nexthop list */ |
d62a17ae | 191 | void route_entry_nexthop_add(struct route_entry *re, struct nexthop *nexthop) |
fa713d9e | 192 | { |
7ee30f28 | 193 | nexthop_add(&re->ng.nexthop, nexthop); |
d62a17ae | 194 | re->nexthop_num++; |
718e3744 | 195 | } |
196 | ||
6e26278c | 197 | |
6e26278c DS |
198 | /** |
199 | * copy_nexthop - copy a nexthop to the rib structure. | |
200 | */ | |
d62a17ae | 201 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) |
6e26278c | 202 | { |
7ee30f28 DS |
203 | assert(!re->ng.nexthop); |
204 | copy_nexthops(&re->ng.nexthop, nh, NULL); | |
d62a17ae | 205 | for (struct nexthop *nexthop = nh; nexthop; nexthop = nexthop->next) |
206 | re->nexthop_num++; | |
6e26278c DS |
207 | } |
208 | ||
718e3744 | 209 | /* Delete specified nexthop from the list. */ |
d62a17ae | 210 | void route_entry_nexthop_delete(struct route_entry *re, struct nexthop *nexthop) |
718e3744 | 211 | { |
d62a17ae | 212 | if (nexthop->next) |
213 | nexthop->next->prev = nexthop->prev; | |
214 | if (nexthop->prev) | |
215 | nexthop->prev->next = nexthop->next; | |
216 | else | |
7ee30f28 | 217 | re->ng.nexthop = nexthop->next; |
d62a17ae | 218 | re->nexthop_num--; |
718e3744 | 219 | } |
220 | ||
fa713d9e | 221 | |
d62a17ae | 222 | struct nexthop *route_entry_nexthop_ifindex_add(struct route_entry *re, |
4a7371e9 DS |
223 | ifindex_t ifindex, |
224 | vrf_id_t nh_vrf_id) | |
718e3744 | 225 | { |
d62a17ae | 226 | struct nexthop *nexthop; |
718e3744 | 227 | |
d62a17ae | 228 | nexthop = nexthop_new(); |
229 | nexthop->type = NEXTHOP_TYPE_IFINDEX; | |
230 | nexthop->ifindex = ifindex; | |
4a7371e9 | 231 | nexthop->vrf_id = nh_vrf_id; |
718e3744 | 232 | |
d62a17ae | 233 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 234 | |
d62a17ae | 235 | return nexthop; |
718e3744 | 236 | } |
237 | ||
d62a17ae | 238 | struct nexthop *route_entry_nexthop_ipv4_add(struct route_entry *re, |
239 | struct in_addr *ipv4, | |
4a7371e9 DS |
240 | struct in_addr *src, |
241 | vrf_id_t nh_vrf_id) | |
718e3744 | 242 | { |
d62a17ae | 243 | struct nexthop *nexthop; |
718e3744 | 244 | |
d62a17ae | 245 | nexthop = nexthop_new(); |
246 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
4a7371e9 | 247 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 248 | nexthop->gate.ipv4 = *ipv4; |
249 | if (src) | |
250 | nexthop->src.ipv4 = *src; | |
718e3744 | 251 | |
d62a17ae | 252 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 253 | |
d62a17ae | 254 | return nexthop; |
718e3744 | 255 | } |
256 | ||
d62a17ae | 257 | struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re, |
258 | struct in_addr *ipv4, | |
259 | struct in_addr *src, | |
4a7371e9 DS |
260 | ifindex_t ifindex, |
261 | vrf_id_t nh_vrf_id) | |
718e3744 | 262 | { |
d62a17ae | 263 | struct nexthop *nexthop; |
264 | struct interface *ifp; | |
718e3744 | 265 | |
d62a17ae | 266 | nexthop = nexthop_new(); |
4a7371e9 | 267 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 268 | nexthop->type = NEXTHOP_TYPE_IPV4_IFINDEX; |
269 | nexthop->gate.ipv4 = *ipv4; | |
270 | if (src) | |
271 | nexthop->src.ipv4 = *src; | |
272 | nexthop->ifindex = ifindex; | |
4a7371e9 | 273 | ifp = if_lookup_by_index(nexthop->ifindex, nh_vrf_id); |
d62a17ae | 274 | /*Pending: need to think if null ifp here is ok during bootup? |
275 | There was a crash because ifp here was coming to be NULL */ | |
276 | if (ifp) | |
996c9314 | 277 | if (connected_is_unnumbered(ifp) |
103e4a71 CF |
278 | || CHECK_FLAG(re->flags, ZEBRA_FLAG_EVPN_ROUTE) |
279 | || CHECK_FLAG(re->flags, ZEBRA_FLAG_ONLINK)) { | |
d62a17ae | 280 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); |
281 | } | |
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; | |
103e4a71 CF |
316 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_EVPN_ROUTE) |
317 | || CHECK_FLAG(re->flags, ZEBRA_FLAG_ONLINK)) { | |
1ec31309 | 318 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); |
103e4a71 | 319 | } |
718e3744 | 320 | |
d62a17ae | 321 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 322 | |
d62a17ae | 323 | return nexthop; |
718e3744 | 324 | } |
718e3744 | 325 | |
a8309422 | 326 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re, |
60466a63 | 327 | enum blackhole_type bh_type) |
595db7f1 | 328 | { |
d62a17ae | 329 | struct nexthop *nexthop; |
595db7f1 | 330 | |
d62a17ae | 331 | nexthop = nexthop_new(); |
4a7371e9 | 332 | nexthop->vrf_id = VRF_DEFAULT; |
d62a17ae | 333 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; |
a8309422 | 334 | nexthop->bh_type = bh_type; |
595db7f1 | 335 | |
d62a17ae | 336 | route_entry_nexthop_add(re, nexthop); |
595db7f1 | 337 | |
d62a17ae | 338 | return nexthop; |
595db7f1 | 339 | } |
340 | ||
86391e56 | 341 | static void nexthop_set_resolved(afi_t afi, const struct nexthop *newhop, |
d62a17ae | 342 | struct nexthop *nexthop) |
4491a88f | 343 | { |
d62a17ae | 344 | struct nexthop *resolved_hop; |
4491a88f | 345 | |
d62a17ae | 346 | resolved_hop = nexthop_new(); |
347 | SET_FLAG(resolved_hop->flags, NEXTHOP_FLAG_ACTIVE); | |
d855d11f | 348 | |
4a7371e9 | 349 | resolved_hop->vrf_id = nexthop->vrf_id; |
d855d11f RW |
350 | switch (newhop->type) { |
351 | case NEXTHOP_TYPE_IPV4: | |
352 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
353 | /* If the resolving route specifies a gateway, use it */ | |
d62a17ae | 354 | resolved_hop->type = newhop->type; |
355 | resolved_hop->gate.ipv4 = newhop->gate.ipv4; | |
4491a88f | 356 | |
d62a17ae | 357 | if (newhop->ifindex) { |
358 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
359 | resolved_hop->ifindex = newhop->ifindex; | |
360 | if (newhop->flags & NEXTHOP_FLAG_ONLINK) | |
361 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
362 | } | |
d855d11f RW |
363 | break; |
364 | case NEXTHOP_TYPE_IPV6: | |
365 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
d62a17ae | 366 | resolved_hop->type = newhop->type; |
367 | resolved_hop->gate.ipv6 = newhop->gate.ipv6; | |
368 | ||
369 | if (newhop->ifindex) { | |
370 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
371 | resolved_hop->ifindex = newhop->ifindex; | |
372 | } | |
d855d11f RW |
373 | break; |
374 | case NEXTHOP_TYPE_IFINDEX: | |
375 | /* If the resolving route is an interface route, | |
376 | * it means the gateway we are looking up is connected | |
377 | * to that interface. (The actual network is _not_ onlink). | |
378 | * Therefore, the resolved route should have the original | |
379 | * gateway as nexthop as it is directly connected. | |
380 | * | |
381 | * On Linux, we have to set the onlink netlink flag because | |
382 | * otherwise, the kernel won't accept the route. | |
383 | */ | |
d62a17ae | 384 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; |
385 | if (afi == AFI_IP) { | |
386 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
387 | resolved_hop->gate.ipv4 = nexthop->gate.ipv4; | |
388 | } else if (afi == AFI_IP6) { | |
389 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
390 | resolved_hop->gate.ipv6 = nexthop->gate.ipv6; | |
391 | } | |
392 | resolved_hop->ifindex = newhop->ifindex; | |
d855d11f RW |
393 | break; |
394 | case NEXTHOP_TYPE_BLACKHOLE: | |
59693377 DS |
395 | resolved_hop->type = NEXTHOP_TYPE_BLACKHOLE; |
396 | resolved_hop->bh_type = nexthop->bh_type; | |
d855d11f | 397 | break; |
59693377 | 398 | } |
d855d11f | 399 | |
f674dfe2 RW |
400 | /* Copy labels of the resolved route */ |
401 | if (newhop->nh_label) | |
402 | nexthop_add_labels(resolved_hop, newhop->nh_label_type, | |
403 | newhop->nh_label->num_labels, | |
404 | &newhop->nh_label->label[0]); | |
405 | ||
d62a17ae | 406 | resolved_hop->rparent = nexthop; |
407 | nexthop_add(&nexthop->resolved, resolved_hop); | |
4491a88f DS |
408 | } |
409 | ||
718e3744 | 410 | /* If force flag is not set, do not modify falgs at all for uninstall |
411 | the route from FIB. */ | |
d62a17ae | 412 | static int nexthop_active(afi_t afi, struct route_entry *re, |
413 | struct nexthop *nexthop, int set, | |
414 | struct route_node *top) | |
415 | { | |
416 | struct prefix p; | |
417 | struct route_table *table; | |
418 | struct route_node *rn; | |
5f7a4718 | 419 | struct route_entry *match = NULL; |
d62a17ae | 420 | int resolved; |
421 | struct nexthop *newhop; | |
422 | struct interface *ifp; | |
5f7a4718 | 423 | rib_dest_t *dest; |
d62a17ae | 424 | |
425 | if ((nexthop->type == NEXTHOP_TYPE_IPV4) | |
426 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
427 | nexthop->ifindex = 0; | |
428 | ||
429 | if (set) { | |
430 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE); | |
d62a17ae | 431 | nexthops_free(nexthop->resolved); |
432 | nexthop->resolved = NULL; | |
433 | re->nexthop_mtu = 0; | |
d44ca835 | 434 | } |
18ff3edd | 435 | |
1ec31309 | 436 | /* Next hops (remote VTEPs) for EVPN routes are fully resolved. */ |
437 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_EVPN_RVTEP)) | |
438 | return 1; | |
439 | ||
d62a17ae | 440 | /* Skip nexthops that have been filtered out due to route-map */ |
441 | /* The nexthops are specific to this route and so the same */ | |
442 | /* nexthop for a different route may not have this flag set */ | |
34815ea3 DS |
443 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FILTERED)) { |
444 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
445 | zlog_debug("\t%s: Nexthop Filtered", | |
446 | __PRETTY_FUNCTION__); | |
d62a17ae | 447 | return 0; |
34815ea3 | 448 | } |
d62a17ae | 449 | |
450 | /* | |
451 | * Check to see if we should trust the passed in information | |
452 | * for UNNUMBERED interfaces as that we won't find the GW | |
453 | * address in the routing table. | |
454 | */ | |
455 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK)) { | |
4a7371e9 | 456 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); |
103e4a71 CF |
457 | if ((ifp && connected_is_unnumbered(ifp)) |
458 | || CHECK_FLAG(re->flags, ZEBRA_FLAG_ONLINK)) { | |
d62a17ae | 459 | if (if_is_operative(ifp)) |
460 | return 1; | |
34815ea3 DS |
461 | else { |
462 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
463 | zlog_debug( | |
464 | "\t%s: Onlink and interface %s is not operative", | |
465 | __PRETTY_FUNCTION__, ifp->name); | |
d62a17ae | 466 | return 0; |
34815ea3 DS |
467 | } |
468 | } else { | |
469 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
470 | zlog_debug( | |
471 | "\t%s: Interface %s is not unnumbered", | |
472 | __PRETTY_FUNCTION__, | |
473 | ifp ? ifp->name : "Unknown"); | |
d62a17ae | 474 | return 0; |
34815ea3 | 475 | } |
16814f96 | 476 | } |
718e3744 | 477 | |
d62a17ae | 478 | /* Make lookup prefix. */ |
479 | memset(&p, 0, sizeof(struct prefix)); | |
480 | switch (afi) { | |
481 | case AFI_IP: | |
482 | p.family = AF_INET; | |
483 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
484 | p.u.prefix4 = nexthop->gate.ipv4; | |
485 | break; | |
486 | case AFI_IP6: | |
487 | p.family = AF_INET6; | |
488 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
489 | p.u.prefix6 = nexthop->gate.ipv6; | |
490 | break; | |
491 | default: | |
492 | assert(afi != AFI_IP && afi != AFI_IP6); | |
493 | break; | |
718e3744 | 494 | } |
d62a17ae | 495 | /* Lookup table. */ |
4a7371e9 | 496 | table = zebra_vrf_table(afi, SAFI_UNICAST, nexthop->vrf_id); |
34815ea3 DS |
497 | if (!table) { |
498 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
499 | zlog_debug("\t%s: Table not found", | |
500 | __PRETTY_FUNCTION__); | |
d62a17ae | 501 | return 0; |
34815ea3 | 502 | } |
d62a17ae | 503 | |
504 | rn = route_node_match(table, (struct prefix *)&p); | |
505 | while (rn) { | |
506 | route_unlock_node(rn); | |
507 | ||
fd7fd9e5 DS |
508 | /* Lookup should halt if we've matched against ourselves ('top', |
509 | * if specified) - i.e., we cannot have a nexthop NH1 is | |
510 | * resolved by a route NH1. The exception is if the route is a | |
511 | * host route. | |
512 | */ | |
513 | if (top && rn == top) | |
996c9314 | 514 | if (((afi == AFI_IP) && (rn->p.prefixlen != 32)) |
34815ea3 DS |
515 | || ((afi == AFI_IP6) && (rn->p.prefixlen != 128))) { |
516 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
517 | zlog_debug( | |
518 | "\t%s: Matched against ourself and prefix length is not max bit length", | |
519 | __PRETTY_FUNCTION__); | |
fd7fd9e5 | 520 | return 0; |
34815ea3 | 521 | } |
d62a17ae | 522 | |
523 | /* Pick up selected route. */ | |
524 | /* However, do not resolve over default route unless explicitly | |
525 | * allowed. */ | |
526 | if (is_default_prefix(&rn->p) | |
34815ea3 DS |
527 | && !rnh_resolve_via_default(p.family)) { |
528 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
529 | zlog_debug( | |
530 | "\t:%s: Resolved against default route", | |
531 | __PRETTY_FUNCTION__); | |
d62a17ae | 532 | return 0; |
34815ea3 | 533 | } |
d62a17ae | 534 | |
5f7a4718 | 535 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
536 | if (dest && dest->selected_fib |
537 | && !CHECK_FLAG(dest->selected_fib->status, | |
538 | ROUTE_ENTRY_REMOVED) | |
539 | && dest->selected_fib->type != ZEBRA_ROUTE_TABLE) | |
5f7a4718 | 540 | match = dest->selected_fib; |
48a53dc7 | 541 | |
d62a17ae | 542 | /* If there is no selected route or matched route is EGP, go up |
543 | tree. */ | |
544 | if (!match) { | |
545 | do { | |
546 | rn = rn->parent; | |
547 | } while (rn && rn->info == NULL); | |
548 | if (rn) | |
549 | route_lock_node(rn); | |
c87bdd2b | 550 | |
d62a17ae | 551 | continue; |
552 | } | |
553 | ||
d62a17ae | 554 | if (match->type == ZEBRA_ROUTE_CONNECT) { |
555 | /* Directly point connected route. */ | |
7ee30f28 | 556 | newhop = match->ng.nexthop; |
d62a17ae | 557 | if (newhop) { |
558 | if (nexthop->type == NEXTHOP_TYPE_IPV4 | |
559 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
560 | nexthop->ifindex = newhop->ifindex; | |
561 | } | |
562 | return 1; | |
4e8b02f4 | 563 | } else if (CHECK_FLAG(re->flags, ZEBRA_FLAG_ALLOW_RECURSION)) { |
d62a17ae | 564 | resolved = 0; |
7ee30f28 | 565 | for (ALL_NEXTHOPS(match->ng, newhop)) { |
a8309422 DL |
566 | if (!CHECK_FLAG(newhop->flags, |
567 | NEXTHOP_FLAG_FIB)) | |
568 | continue; | |
569 | if (CHECK_FLAG(newhop->flags, | |
570 | NEXTHOP_FLAG_RECURSIVE)) | |
571 | continue; | |
572 | ||
573 | if (set) { | |
574 | SET_FLAG(nexthop->flags, | |
575 | NEXTHOP_FLAG_RECURSIVE); | |
576 | SET_FLAG(re->status, | |
577 | ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
578 | nexthop_set_resolved(afi, newhop, | |
579 | nexthop); | |
d62a17ae | 580 | } |
a8309422 DL |
581 | resolved = 1; |
582 | } | |
43e31305 JB |
583 | if (resolved && set) |
584 | re->nexthop_mtu = match->mtu; | |
34815ea3 DS |
585 | if (!resolved && IS_ZEBRA_DEBUG_RIB_DETAILED) |
586 | zlog_debug("\t%s: Recursion failed to find", | |
587 | __PRETTY_FUNCTION__); | |
d62a17ae | 588 | return resolved; |
589 | } else if (re->type == ZEBRA_ROUTE_STATIC) { | |
590 | resolved = 0; | |
7ee30f28 | 591 | for (ALL_NEXTHOPS(match->ng, newhop)) { |
a8309422 DL |
592 | if (!CHECK_FLAG(newhop->flags, |
593 | NEXTHOP_FLAG_FIB)) | |
594 | continue; | |
595 | ||
596 | if (set) { | |
597 | SET_FLAG(nexthop->flags, | |
598 | NEXTHOP_FLAG_RECURSIVE); | |
599 | nexthop_set_resolved(afi, newhop, | |
600 | nexthop); | |
d62a17ae | 601 | } |
a8309422 DL |
602 | resolved = 1; |
603 | } | |
d62a17ae | 604 | if (resolved && set) |
605 | re->nexthop_mtu = match->mtu; | |
34815ea3 DS |
606 | |
607 | if (!resolved && IS_ZEBRA_DEBUG_RIB_DETAILED) | |
608 | zlog_debug( | |
609 | "\t%s: Static route unable to resolve", | |
610 | __PRETTY_FUNCTION__); | |
d62a17ae | 611 | return resolved; |
612 | } else { | |
7939ff76 DS |
613 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
614 | zlog_debug("\t%s: Route Type %s has not turned on recursion", | |
615 | __PRETTY_FUNCTION__, | |
616 | zebra_route_string(re->type)); | |
617 | if (re->type == ZEBRA_ROUTE_BGP && | |
618 | !CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
619 | zlog_debug("\tEBGP: see \"disable-ebgp-connected-route-check\" or \"disable-connected-check\""); | |
620 | } | |
d62a17ae | 621 | return 0; |
622 | } | |
718e3744 | 623 | } |
7939ff76 DS |
624 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
625 | zlog_debug("\t%s: Nexthop did not lookup in table", | |
626 | __PRETTY_FUNCTION__); | |
d62a17ae | 627 | return 0; |
718e3744 | 628 | } |
718e3744 | 629 | |
d62a17ae | 630 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
631 | union g_addr *addr, struct route_node **rn_out) | |
632 | { | |
633 | struct prefix p; | |
634 | struct route_table *table; | |
635 | struct route_node *rn; | |
5f7a4718 | 636 | struct route_entry *match = NULL; |
d62a17ae | 637 | struct nexthop *newhop; |
638 | ||
639 | /* Lookup table. */ | |
640 | table = zebra_vrf_table(afi, safi, vrf_id); | |
641 | if (!table) | |
642 | return 0; | |
643 | ||
644 | memset(&p, 0, sizeof(struct prefix)); | |
645 | p.family = afi; | |
646 | if (afi == AFI_IP) { | |
647 | p.u.prefix4 = addr->ipv4; | |
648 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
649 | } else { | |
650 | p.u.prefix6 = addr->ipv6; | |
651 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
652 | } | |
718e3744 | 653 | |
d62a17ae | 654 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 655 | |
d62a17ae | 656 | while (rn) { |
5f7a4718 DS |
657 | rib_dest_t *dest; |
658 | ||
d62a17ae | 659 | route_unlock_node(rn); |
718e3744 | 660 | |
5f7a4718 | 661 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
662 | if (dest && dest->selected_fib |
663 | && !CHECK_FLAG(dest->selected_fib->status, | |
664 | ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 665 | match = dest->selected_fib; |
718e3744 | 666 | |
d62a17ae | 667 | /* If there is no selected route or matched route is EGP, go up |
668 | tree. */ | |
669 | if (!match) { | |
670 | do { | |
671 | rn = rn->parent; | |
672 | } while (rn && rn->info == NULL); | |
673 | if (rn) | |
674 | route_lock_node(rn); | |
675 | } else { | |
676 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
677 | int found = 0; | |
7ee30f28 | 678 | for (ALL_NEXTHOPS(match->ng, newhop)) |
d62a17ae | 679 | if (CHECK_FLAG(newhop->flags, |
680 | NEXTHOP_FLAG_FIB)) { | |
681 | found = 1; | |
682 | break; | |
683 | } | |
684 | if (!found) | |
685 | return NULL; | |
686 | } | |
687 | ||
688 | if (rn_out) | |
689 | *rn_out = rn; | |
690 | return match; | |
691 | } | |
16814f96 | 692 | } |
d62a17ae | 693 | return NULL; |
694 | } | |
695 | ||
696 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
697 | struct in_addr addr, | |
698 | struct route_node **rn_out) | |
699 | { | |
700 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
701 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
702 | union g_addr gaddr = {.ipv4 = addr}; | |
703 | ||
704 | switch (ipv4_multicast_mode) { | |
705 | case MCAST_MRIB_ONLY: | |
706 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
707 | rn_out); | |
708 | case MCAST_URIB_ONLY: | |
709 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
710 | case MCAST_NO_CONFIG: | |
711 | case MCAST_MIX_MRIB_FIRST: | |
712 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
713 | &m_rn); | |
714 | if (!mre) | |
715 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
716 | &gaddr, &u_rn); | |
717 | break; | |
718 | case MCAST_MIX_DISTANCE: | |
719 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
720 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
721 | if (mre && ure) | |
722 | re = ure->distance < mre->distance ? ure : mre; | |
723 | else if (mre) | |
724 | re = mre; | |
725 | else if (ure) | |
726 | re = ure; | |
727 | break; | |
728 | case MCAST_MIX_PFXLEN: | |
729 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
730 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
731 | if (mre && ure) | |
732 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
733 | else if (mre) | |
734 | re = mre; | |
735 | else if (ure) | |
736 | re = ure; | |
737 | break; | |
718e3744 | 738 | } |
33550aa8 | 739 | |
d62a17ae | 740 | if (rn_out) |
741 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 742 | |
d62a17ae | 743 | if (IS_ZEBRA_DEBUG_RIB) { |
744 | char buf[BUFSIZ]; | |
745 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 746 | |
32391aff DS |
747 | zlog_debug("%s: %s: vrf: %u found %s, using %s", |
748 | __func__, buf, vrf_id, | |
d62a17ae | 749 | mre ? (ure ? "MRIB+URIB" : "MRIB") |
750 | : ure ? "URIB" : "nothing", | |
751 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
752 | } | |
753 | return re; | |
4623d897 DL |
754 | } |
755 | ||
d62a17ae | 756 | void multicast_mode_ipv4_set(enum multicast_mode mode) |
4623d897 | 757 | { |
d62a17ae | 758 | if (IS_ZEBRA_DEBUG_RIB) |
759 | zlog_debug("%s: multicast lookup mode set (%d)", __func__, | |
760 | mode); | |
761 | ipv4_multicast_mode = mode; | |
4623d897 DL |
762 | } |
763 | ||
d62a17ae | 764 | enum multicast_mode multicast_mode_ipv4_get(void) |
4623d897 | 765 | { |
d62a17ae | 766 | return ipv4_multicast_mode; |
4623d897 DL |
767 | } |
768 | ||
d62a17ae | 769 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 770 | { |
d62a17ae | 771 | struct route_table *table; |
772 | struct route_node *rn; | |
5f7a4718 | 773 | struct route_entry *match = NULL; |
d62a17ae | 774 | struct nexthop *nexthop; |
5f7a4718 | 775 | rib_dest_t *dest; |
718e3744 | 776 | |
d62a17ae | 777 | /* Lookup table. */ |
778 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
779 | if (!table) | |
780 | return 0; | |
718e3744 | 781 | |
d62a17ae | 782 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 783 | |
d62a17ae | 784 | /* No route for this prefix. */ |
785 | if (!rn) | |
786 | return NULL; | |
718e3744 | 787 | |
d62a17ae | 788 | /* Unlock node. */ |
789 | route_unlock_node(rn); | |
5f7a4718 | 790 | dest = rib_dest_from_rnode(rn); |
718e3744 | 791 | |
996c9314 LB |
792 | if (dest && dest->selected_fib |
793 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 794 | match = dest->selected_fib; |
718e3744 | 795 | |
d62a17ae | 796 | if (!match) |
797 | return NULL; | |
718e3744 | 798 | |
d62a17ae | 799 | if (match->type == ZEBRA_ROUTE_CONNECT) |
800 | return match; | |
f9e1b38e | 801 | |
7ee30f28 | 802 | for (ALL_NEXTHOPS(match->ng, nexthop)) |
d62a17ae | 803 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) |
804 | return match; | |
718e3744 | 805 | |
d62a17ae | 806 | return NULL; |
718e3744 | 807 | } |
808 | ||
dc95824a DO |
809 | /* |
810 | * This clone function, unlike its original rib_lookup_ipv4(), checks | |
811 | * if specified IPv4 route record (prefix/mask -> gate) exists in | |
f0f77c9a | 812 | * the whole RIB and has ROUTE_ENTRY_SELECTED_FIB set. |
dc95824a DO |
813 | * |
814 | * Return values: | |
815 | * -1: error | |
816 | * 0: exact match found | |
817 | * 1: a match was found with a different gate | |
818 | * 2: connected route found | |
819 | * 3: no matches found | |
820 | */ | |
d62a17ae | 821 | int rib_lookup_ipv4_route(struct prefix_ipv4 *p, union sockunion *qgate, |
822 | vrf_id_t vrf_id) | |
dc95824a | 823 | { |
d62a17ae | 824 | struct route_table *table; |
825 | struct route_node *rn; | |
5f7a4718 | 826 | struct route_entry *match = NULL; |
d62a17ae | 827 | struct nexthop *nexthop; |
828 | int nexthops_active; | |
5f7a4718 | 829 | rib_dest_t *dest; |
dc95824a | 830 | |
d62a17ae | 831 | /* Lookup table. */ |
832 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
833 | if (!table) | |
834 | return ZEBRA_RIB_LOOKUP_ERROR; | |
dc95824a | 835 | |
d62a17ae | 836 | /* Scan the RIB table for exactly matching RIB entry. */ |
837 | rn = route_node_lookup(table, (struct prefix *)p); | |
dc95824a | 838 | |
d62a17ae | 839 | /* No route for this prefix. */ |
840 | if (!rn) | |
841 | return ZEBRA_RIB_NOTFOUND; | |
dc95824a | 842 | |
d62a17ae | 843 | /* Unlock node. */ |
844 | route_unlock_node(rn); | |
5f7a4718 | 845 | dest = rib_dest_from_rnode(rn); |
dc95824a | 846 | |
d62a17ae | 847 | /* Find out if a "selected" RR for the discovered RIB entry exists ever. |
848 | */ | |
996c9314 LB |
849 | if (dest && dest->selected_fib |
850 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 851 | match = dest->selected_fib; |
dc95824a | 852 | |
d62a17ae | 853 | /* None such found :( */ |
854 | if (!match) | |
855 | return ZEBRA_RIB_NOTFOUND; | |
856 | ||
857 | if (match->type == ZEBRA_ROUTE_CONNECT) | |
858 | return ZEBRA_RIB_FOUND_CONNECTED; | |
859 | ||
860 | /* Ok, we have a cood candidate, let's check it's nexthop list... */ | |
861 | nexthops_active = 0; | |
7ee30f28 | 862 | for (ALL_NEXTHOPS(match->ng, nexthop)) |
d62a17ae | 863 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) { |
864 | nexthops_active = 1; | |
865 | if (nexthop->gate.ipv4.s_addr == sockunion2ip(qgate)) | |
866 | return ZEBRA_RIB_FOUND_EXACT; | |
867 | if (IS_ZEBRA_DEBUG_RIB) { | |
868 | char gate_buf[INET_ADDRSTRLEN], | |
869 | qgate_buf[INET_ADDRSTRLEN]; | |
870 | inet_ntop(AF_INET, &nexthop->gate.ipv4.s_addr, | |
871 | gate_buf, INET_ADDRSTRLEN); | |
872 | inet_ntop(AF_INET, &sockunion2ip(qgate), | |
873 | qgate_buf, INET_ADDRSTRLEN); | |
874 | zlog_debug("%s: qgate == %s, %s == %s", | |
875 | __func__, qgate_buf, | |
876 | nexthop->rparent ? "rgate" : "gate", | |
877 | gate_buf); | |
878 | } | |
879 | } | |
880 | ||
881 | if (nexthops_active) | |
882 | return ZEBRA_RIB_FOUND_NOGATE; | |
883 | ||
884 | return ZEBRA_RIB_NOTFOUND; | |
885 | } | |
886 | ||
887 | #define RIB_SYSTEM_ROUTE(R) \ | |
888 | ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT) | |
7514fb77 | 889 | |
212df1de PG |
890 | #define RIB_KERNEL_ROUTE(R) \ |
891 | ((R)->type == ZEBRA_ROUTE_KERNEL) | |
892 | ||
dc95824a DO |
893 | /* This function verifies reachability of one given nexthop, which can be |
894 | * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored | |
895 | * in nexthop->flags field. If the 4th parameter, 'set', is non-zero, | |
896 | * nexthop->ifindex will be updated appropriately as well. | |
897 | * An existing route map can turn (otherwise active) nexthop into inactive, but | |
898 | * not vice versa. | |
899 | * | |
900 | * The return value is the final value of 'ACTIVE' flag. | |
901 | */ | |
902 | ||
d62a17ae | 903 | static unsigned nexthop_active_check(struct route_node *rn, |
904 | struct route_entry *re, | |
905 | struct nexthop *nexthop, int set) | |
906 | { | |
907 | struct interface *ifp; | |
908 | route_map_result_t ret = RMAP_MATCH; | |
909 | int family; | |
910 | char buf[SRCDEST2STR_BUFFER]; | |
86391e56 | 911 | const struct prefix *p, *src_p; |
ac6eebce | 912 | struct zebra_vrf *zvrf; |
913 | ||
d62a17ae | 914 | srcdest_rnode_prefixes(rn, &p, &src_p); |
915 | ||
916 | if (rn->p.family == AF_INET) | |
917 | family = AFI_IP; | |
918 | else if (rn->p.family == AF_INET6) | |
919 | family = AFI_IP6; | |
920 | else | |
921 | family = 0; | |
922 | switch (nexthop->type) { | |
923 | case NEXTHOP_TYPE_IFINDEX: | |
4a7371e9 | 924 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); |
d62a17ae | 925 | if (ifp && if_is_operative(ifp)) |
926 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
927 | else | |
928 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
929 | break; | |
930 | case NEXTHOP_TYPE_IPV4: | |
931 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
932 | family = AFI_IP; | |
1ec31309 | 933 | if (nexthop_active(AFI_IP, re, nexthop, set, rn)) |
d62a17ae | 934 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
935 | else | |
936 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
937 | break; | |
938 | case NEXTHOP_TYPE_IPV6: | |
939 | family = AFI_IP6; | |
940 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
941 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
942 | else | |
943 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
944 | break; | |
945 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
946 | /* RFC 5549, v4 prefix with v6 NH */ | |
947 | if (rn->p.family != AF_INET) | |
948 | family = AFI_IP6; | |
949 | if (IN6_IS_ADDR_LINKLOCAL(&nexthop->gate.ipv6)) { | |
007dbee6 | 950 | ifp = if_lookup_by_index(nexthop->ifindex, |
4a7371e9 | 951 | nexthop->vrf_id); |
d62a17ae | 952 | if (ifp && if_is_operative(ifp)) |
953 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
954 | else | |
955 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
956 | } else { | |
957 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
958 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
959 | else | |
960 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
961 | } | |
962 | break; | |
963 | case NEXTHOP_TYPE_BLACKHOLE: | |
964 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
965 | break; | |
966 | default: | |
967 | break; | |
968 | } | |
34815ea3 DS |
969 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) { |
970 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
971 | zlog_debug("\t%s: Unable to find a active nexthop", | |
972 | __PRETTY_FUNCTION__); | |
d62a17ae | 973 | return 0; |
34815ea3 | 974 | } |
d62a17ae | 975 | |
976 | /* XXX: What exactly do those checks do? Do we support | |
ac6eebce | 977 | * e.g. IPv4 routes with IPv6 nexthops or vice versa? |
978 | */ | |
d62a17ae | 979 | if (RIB_SYSTEM_ROUTE(re) || (family == AFI_IP && p->family != AF_INET) |
980 | || (family == AFI_IP6 && p->family != AF_INET6)) | |
981 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
982 | ||
983 | /* The original code didn't determine the family correctly | |
984 | * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi | |
985 | * from the rib_table_info in those cases. | |
986 | * Possibly it may be better to use only the rib_table_info | |
987 | * in every case. | |
988 | */ | |
989 | if (!family) { | |
990 | rib_table_info_t *info; | |
991 | ||
992 | info = srcdest_rnode_table_info(rn); | |
993 | family = info->afi; | |
718e3744 | 994 | } |
c52ef59f | 995 | |
d62a17ae | 996 | memset(&nexthop->rmap_src.ipv6, 0, sizeof(union g_addr)); |
997 | ||
ac6eebce | 998 | zvrf = zebra_vrf_lookup_by_id(nexthop->vrf_id); |
999 | if (!zvrf) { | |
1000 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1001 | zlog_debug("\t%s: zvrf is NULL", __PRETTY_FUNCTION__); | |
1002 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
1003 | } | |
1004 | ||
d62a17ae | 1005 | /* It'll get set if required inside */ |
ac6eebce | 1006 | ret = zebra_route_map_check(family, re->type, re->instance, p, |
1007 | nexthop, zvrf, re->tag); | |
d62a17ae | 1008 | if (ret == RMAP_DENYMATCH) { |
1009 | if (IS_ZEBRA_DEBUG_RIB) { | |
1010 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1011 | zlog_debug( | |
1012 | "%u:%s: Filtering out with NH out %s due to route map", | |
1013 | re->vrf_id, buf, | |
99b9d960 | 1014 | ifindex2ifname(nexthop->ifindex, |
4a7371e9 | 1015 | nexthop->vrf_id)); |
d62a17ae | 1016 | } |
1017 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
518f0eb1 | 1018 | } |
d62a17ae | 1019 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
718e3744 | 1020 | } |
1021 | ||
03e232a4 | 1022 | /* Iterate over all nexthops of the given RIB entry and refresh their |
f0f77c9a DS |
1023 | * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any |
1024 | * nexthop is found to toggle the ACTIVE flag, the whole re structure | |
1025 | * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is | |
03e232a4 DO |
1026 | * transparently passed to nexthop_active_check(). |
1027 | * | |
1028 | * Return value is the new number of active nexthops. | |
1029 | */ | |
1030 | ||
d62a17ae | 1031 | static int nexthop_active_update(struct route_node *rn, struct route_entry *re, |
1032 | int set) | |
1033 | { | |
1034 | struct nexthop *nexthop; | |
1035 | union g_addr prev_src; | |
1036 | unsigned int prev_active, new_active, old_num_nh; | |
1037 | ifindex_t prev_index; | |
ac6eebce | 1038 | |
d62a17ae | 1039 | old_num_nh = re->nexthop_active_num; |
1040 | ||
1041 | re->nexthop_active_num = 0; | |
1042 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1043 | ||
7ee30f28 | 1044 | for (nexthop = re->ng.nexthop; nexthop; nexthop = nexthop->next) { |
d62a17ae | 1045 | /* No protocol daemon provides src and so we're skipping |
1046 | * tracking it */ | |
1047 | prev_src = nexthop->rmap_src; | |
1048 | prev_active = CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
1049 | prev_index = nexthop->ifindex; | |
1050 | if ((new_active = nexthop_active_check(rn, re, nexthop, set))) | |
1051 | re->nexthop_active_num++; | |
1052 | /* Don't allow src setting on IPv6 addr for now */ | |
1053 | if (prev_active != new_active || prev_index != nexthop->ifindex | |
1054 | || ((nexthop->type >= NEXTHOP_TYPE_IFINDEX | |
1055 | && nexthop->type < NEXTHOP_TYPE_IPV6) | |
1056 | && prev_src.ipv4.s_addr | |
1057 | != nexthop->rmap_src.ipv4.s_addr) | |
1058 | || ((nexthop->type >= NEXTHOP_TYPE_IPV6 | |
1059 | && nexthop->type < NEXTHOP_TYPE_BLACKHOLE) | |
1060 | && !(IPV6_ADDR_SAME(&prev_src.ipv6, | |
1061 | &nexthop->rmap_src.ipv6)))) { | |
1062 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1063 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1064 | } | |
1065 | } | |
1066 | ||
1067 | if (old_num_nh != re->nexthop_active_num) | |
1068 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
6e26278c | 1069 | |
d62a17ae | 1070 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED)) { |
1071 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1072 | } | |
1073 | ||
1074 | return re->nexthop_active_num; | |
718e3744 | 1075 | } |
6baeb988 | 1076 | |
a64448ba DS |
1077 | /* |
1078 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
1079 | * (nexthops) must have a label. | |
1080 | */ | |
d62a17ae | 1081 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 1082 | { |
d62a17ae | 1083 | struct nexthop *nexthop = NULL; |
a64448ba | 1084 | |
d62a17ae | 1085 | if (re->type != ZEBRA_ROUTE_BGP) |
1086 | return 0; | |
a64448ba | 1087 | |
7ee30f28 | 1088 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d62a17ae | 1089 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) |
1090 | return 0; | |
6b0655a2 | 1091 | |
d62a17ae | 1092 | return 1; |
a64448ba | 1093 | } |
718e3744 | 1094 | |
86391e56 | 1095 | void kernel_route_rib_pass_fail(struct route_node *rn, const struct prefix *p, |
7d974ba3 | 1096 | struct route_entry *re, |
ea1c14f6 | 1097 | enum zebra_dplane_status res) |
0c555cc6 DS |
1098 | { |
1099 | struct nexthop *nexthop; | |
1100 | char buf[PREFIX_STRLEN]; | |
ed216282 DS |
1101 | rib_dest_t *dest; |
1102 | ||
1103 | dest = rib_dest_from_rnode(rn); | |
0c555cc6 DS |
1104 | |
1105 | switch (res) { | |
ea1c14f6 | 1106 | case ZEBRA_DPLANE_INSTALL_SUCCESS: |
ed216282 | 1107 | dest->selected_fib = re; |
7ee30f28 | 1108 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
0c555cc6 DS |
1109 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) |
1110 | continue; | |
1111 | ||
1112 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
1113 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1114 | else | |
1115 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
1116 | } | |
28610f7e | 1117 | zsend_route_notify_owner(re, p, ZAPI_ROUTE_INSTALLED); |
0c555cc6 | 1118 | break; |
ea1c14f6 | 1119 | case ZEBRA_DPLANE_INSTALL_FAILURE: |
ed216282 DS |
1120 | /* |
1121 | * I am not sure this is the right thing to do here | |
1122 | * but the code always set selected_fib before | |
1123 | * this assignment was moved here. | |
1124 | */ | |
1125 | dest->selected_fib = re; | |
1126 | ||
28610f7e | 1127 | zsend_route_notify_owner(re, p, ZAPI_ROUTE_FAIL_INSTALL); |
e914ccbe | 1128 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, |
9df414fe QY |
1129 | "%u:%s: Route install failed", re->vrf_id, |
1130 | prefix2str(p, buf, sizeof(buf))); | |
0c555cc6 | 1131 | break; |
ea1c14f6 | 1132 | case ZEBRA_DPLANE_DELETE_SUCCESS: |
ed216282 DS |
1133 | /* |
1134 | * The case where selected_fib is not re is | |
1135 | * when we have received a system route | |
1136 | * that is overriding our installed route | |
1137 | * as such we should leave the selected_fib | |
1138 | * pointer alone | |
1139 | */ | |
1140 | if (dest->selected_fib == re) | |
1141 | dest->selected_fib = NULL; | |
7ee30f28 | 1142 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
0c555cc6 | 1143 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
5e54c602 DS |
1144 | |
1145 | zsend_route_notify_owner(re, p, ZAPI_ROUTE_REMOVED); | |
0c555cc6 | 1146 | break; |
ea1c14f6 | 1147 | case ZEBRA_DPLANE_DELETE_FAILURE: |
ed216282 DS |
1148 | /* |
1149 | * Should we set this to NULL if the | |
1150 | * delete fails? | |
1151 | */ | |
1152 | dest->selected_fib = NULL; | |
e914ccbe | 1153 | flog_err(EC_ZEBRA_DP_DELETE_FAIL, |
9df414fe QY |
1154 | "%u:%s: Route Deletion failure", re->vrf_id, |
1155 | prefix2str(p, buf, sizeof(buf))); | |
5e54c602 DS |
1156 | |
1157 | zsend_route_notify_owner(re, p, ZAPI_ROUTE_REMOVE_FAIL); | |
0c555cc6 | 1158 | break; |
ea1c14f6 MS |
1159 | case ZEBRA_DPLANE_STATUS_NONE: |
1160 | break; | |
0c555cc6 DS |
1161 | } |
1162 | } | |
1163 | ||
6ae24471 DS |
1164 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
1165 | * is only used for IPv4. | |
1166 | */ | |
0c555cc6 DS |
1167 | void rib_install_kernel(struct route_node *rn, struct route_entry *re, |
1168 | struct route_entry *old) | |
718e3744 | 1169 | { |
d62a17ae | 1170 | struct nexthop *nexthop; |
1171 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 1172 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
97f5b441 MS |
1173 | const struct prefix *p, *src_p; |
1174 | enum zebra_dplane_result ret; | |
1175 | ||
1176 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
718e3744 | 1177 | |
d62a17ae | 1178 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 1179 | |
d62a17ae | 1180 | if (info->safi != SAFI_UNICAST) { |
7ee30f28 | 1181 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d62a17ae | 1182 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 1183 | return; |
25b9cb0c DL |
1184 | } else { |
1185 | struct nexthop *prev; | |
1186 | ||
7ee30f28 | 1187 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
25b9cb0c | 1188 | UNSET_FLAG (nexthop->flags, NEXTHOP_FLAG_DUPLICATE); |
7ee30f28 | 1189 | for (ALL_NEXTHOPS(re->ng, prev)) { |
25b9cb0c DL |
1190 | if (prev == nexthop) |
1191 | break; | |
996c9314 LB |
1192 | if (nexthop_same_firsthop(nexthop, prev)) { |
1193 | SET_FLAG(nexthop->flags, | |
1194 | NEXTHOP_FLAG_DUPLICATE); | |
25b9cb0c DL |
1195 | break; |
1196 | } | |
1197 | } | |
1198 | } | |
d62a17ae | 1199 | } |
718e3744 | 1200 | |
2063a814 DS |
1201 | /* |
1202 | * If this is a replace to a new RE let the originator of the RE | |
1203 | * know that they've lost | |
1204 | */ | |
9a9f8926 | 1205 | if (old && (old != re) && (old->type != re->type)) |
28610f7e | 1206 | zsend_route_notify_owner(old, p, ZAPI_ROUTE_BETTER_ADMIN_WON); |
25b9cb0c | 1207 | |
97f5b441 MS |
1208 | /* Update fib selection */ |
1209 | dest->selected_fib = re; | |
1210 | ||
d62a17ae | 1211 | /* |
1212 | * Make sure we update the FPM any time we send new information to | |
1213 | * the kernel. | |
1214 | */ | |
1215 | hook_call(rib_update, rn, "installing in kernel"); | |
97f5b441 MS |
1216 | |
1217 | /* Send add or update */ | |
5709131c | 1218 | if (old && (old != re)) |
97f5b441 | 1219 | ret = dplane_route_update(rn, re, old); |
5709131c | 1220 | else |
97f5b441 | 1221 | ret = dplane_route_add(rn, re); |
97f5b441 MS |
1222 | |
1223 | switch (ret) { | |
ea1c14f6 | 1224 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
97f5b441 MS |
1225 | if (zvrf) |
1226 | zvrf->installs_queued++; | |
1e885672 | 1227 | break; |
ea1c14f6 | 1228 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
1229 | { |
1230 | char str[SRCDEST2STR_BUFFER]; | |
1231 | ||
1232 | srcdest_rnode2str(rn, str, sizeof(str)); | |
1233 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
1234 | "%u:%s: Failed to enqueue dataplane install", | |
1235 | re->vrf_id, str); | |
1e885672 | 1236 | break; |
97f5b441 | 1237 | } |
ea1c14f6 | 1238 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
97f5b441 MS |
1239 | if (zvrf) |
1240 | zvrf->installs++; | |
1e885672 DS |
1241 | break; |
1242 | } | |
d62a17ae | 1243 | |
0c555cc6 | 1244 | return; |
718e3744 | 1245 | } |
1246 | ||
1247 | /* Uninstall the route from kernel. */ | |
0c555cc6 | 1248 | void rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 1249 | { |
d62a17ae | 1250 | struct nexthop *nexthop; |
1251 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 1252 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
05737783 | 1253 | |
d62a17ae | 1254 | if (info->safi != SAFI_UNICAST) { |
7ee30f28 | 1255 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d6792f9d | 1256 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 1257 | return; |
d62a17ae | 1258 | } |
416ec78d | 1259 | |
d62a17ae | 1260 | /* |
1261 | * Make sure we update the FPM any time we send new information to | |
97f5b441 | 1262 | * the dataplane. |
d62a17ae | 1263 | */ |
1264 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
97f5b441 MS |
1265 | |
1266 | switch (dplane_route_delete(rn, re)) { | |
ea1c14f6 | 1267 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
97f5b441 MS |
1268 | if (zvrf) |
1269 | zvrf->removals_queued++; | |
1e885672 | 1270 | break; |
ea1c14f6 | 1271 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
1272 | { |
1273 | char str[SRCDEST2STR_BUFFER]; | |
1274 | ||
1275 | srcdest_rnode2str(rn, str, sizeof(str)); | |
1276 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
1277 | "%u:%s: Failed to enqueue dataplane uninstall", | |
1278 | re->vrf_id, str); | |
1e885672 | 1279 | break; |
97f5b441 | 1280 | } |
ea1c14f6 | 1281 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
1e885672 DS |
1282 | if (zvrf) |
1283 | zvrf->removals++; | |
1284 | break; | |
1285 | } | |
718e3744 | 1286 | |
0c555cc6 | 1287 | return; |
718e3744 | 1288 | } |
1289 | ||
1290 | /* Uninstall the route from kernel. */ | |
d62a17ae | 1291 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 1292 | { |
d62a17ae | 1293 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
5f7a4718 | 1294 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
97f5b441 | 1295 | struct nexthop *nexthop; |
416ec78d | 1296 | |
5f7a4718 | 1297 | if (dest && dest->selected_fib == re) { |
d62a17ae | 1298 | if (info->safi == SAFI_UNICAST) |
1299 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 1300 | |
d62a17ae | 1301 | /* If labeled-unicast route, uninstall transit LSP. */ |
1302 | if (zebra_rib_labeled_unicast(re)) | |
1303 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
7cdb1a84 MS |
1304 | |
1305 | if (!RIB_SYSTEM_ROUTE(re)) | |
1306 | rib_uninstall_kernel(rn, re); | |
97f5b441 MS |
1307 | |
1308 | dest->selected_fib = NULL; | |
1309 | ||
1310 | for (ALL_NEXTHOPS(re->ng, nexthop)) | |
1311 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
d62a17ae | 1312 | } |
446bb95e | 1313 | |
d62a17ae | 1314 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
86391e56 MS |
1315 | const struct prefix *p, *src_p; |
1316 | ||
d62a17ae | 1317 | srcdest_rnode_prefixes(rn, &p, &src_p); |
05737783 | 1318 | |
d62a17ae | 1319 | redistribute_delete(p, src_p, re); |
1320 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); | |
1321 | } | |
718e3744 | 1322 | } |
1323 | ||
9fd92e3c AS |
1324 | /* |
1325 | * rib_can_delete_dest | |
1326 | * | |
1327 | * Returns TRUE if the given dest can be deleted from the table. | |
1328 | */ | |
d62a17ae | 1329 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 1330 | { |
d62a17ae | 1331 | if (dest->routes) { |
1332 | return 0; | |
1333 | } | |
9fd92e3c | 1334 | |
d62a17ae | 1335 | /* |
1336 | * Don't delete the dest if we have to update the FPM about this | |
1337 | * prefix. | |
1338 | */ | |
1339 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
1340 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
1341 | return 0; | |
5adc2528 | 1342 | |
d62a17ae | 1343 | return 1; |
9fd92e3c AS |
1344 | } |
1345 | ||
1346 | /* | |
1347 | * rib_gc_dest | |
1348 | * | |
1349 | * Garbage collect the rib dest corresponding to the given route node | |
1350 | * if appropriate. | |
1351 | * | |
1352 | * Returns TRUE if the dest was deleted, FALSE otherwise. | |
1353 | */ | |
d62a17ae | 1354 | int rib_gc_dest(struct route_node *rn) |
9fd92e3c | 1355 | { |
d62a17ae | 1356 | rib_dest_t *dest; |
9fd92e3c | 1357 | |
d62a17ae | 1358 | dest = rib_dest_from_rnode(rn); |
1359 | if (!dest) | |
1360 | return 0; | |
9fd92e3c | 1361 | |
d62a17ae | 1362 | if (!rib_can_delete_dest(dest)) |
1363 | return 0; | |
9fd92e3c | 1364 | |
c9abf558 DS |
1365 | if (IS_ZEBRA_DEBUG_RIB) { |
1366 | struct zebra_vrf *zvrf; | |
1367 | ||
1368 | zvrf = rib_dest_vrf(dest); | |
d62a17ae | 1369 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); |
c9abf558 | 1370 | } |
9fd92e3c | 1371 | |
d62a17ae | 1372 | dest->rnode = NULL; |
1373 | XFREE(MTYPE_RIB_DEST, dest); | |
1374 | rn->info = NULL; | |
9fd92e3c | 1375 | |
d62a17ae | 1376 | /* |
1377 | * Release the one reference that we keep on the route node. | |
1378 | */ | |
1379 | route_unlock_node(rn); | |
1380 | return 1; | |
9fd92e3c AS |
1381 | } |
1382 | ||
d62a17ae | 1383 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1384 | struct route_entry *new) | |
3e5c6e00 | 1385 | { |
5f7a4718 DS |
1386 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
1387 | ||
d62a17ae | 1388 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 1389 | |
d62a17ae | 1390 | /* Update real nexthop. This may actually determine if nexthop is active |
1391 | * or not. */ | |
1392 | if (!nexthop_active_update(rn, new, 1)) { | |
1393 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
1394 | return; | |
1395 | } | |
3e5c6e00 | 1396 | |
d62a17ae | 1397 | if (IS_ZEBRA_DEBUG_RIB) { |
1398 | char buf[SRCDEST2STR_BUFFER]; | |
1399 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1400 | zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)", | |
1401 | zvrf_id(zvrf), buf, rn, new, new->type); | |
1402 | } | |
3e5c6e00 | 1403 | |
d62a17ae | 1404 | /* If labeled-unicast route, install transit LSP. */ |
1405 | if (zebra_rib_labeled_unicast(new)) | |
1406 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 1407 | |
0c555cc6 DS |
1408 | if (!RIB_SYSTEM_ROUTE(new)) |
1409 | rib_install_kernel(rn, new, NULL); | |
ed216282 DS |
1410 | else |
1411 | dest->selected_fib = new; | |
3e5c6e00 | 1412 | |
d62a17ae | 1413 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1414 | } |
1415 | ||
d62a17ae | 1416 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1417 | struct route_entry *old) | |
3e5c6e00 | 1418 | { |
5f7a4718 | 1419 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
d62a17ae | 1420 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 1421 | |
d62a17ae | 1422 | /* Uninstall from kernel. */ |
1423 | if (IS_ZEBRA_DEBUG_RIB) { | |
1424 | char buf[SRCDEST2STR_BUFFER]; | |
1425 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1426 | zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)", | |
1427 | zvrf_id(zvrf), buf, rn, old, old->type); | |
1428 | } | |
3e5c6e00 | 1429 | |
d62a17ae | 1430 | /* If labeled-unicast route, uninstall transit LSP. */ |
1431 | if (zebra_rib_labeled_unicast(old)) | |
1432 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1433 | ||
1434 | if (!RIB_SYSTEM_ROUTE(old)) | |
1435 | rib_uninstall_kernel(rn, old); | |
ed216282 DS |
1436 | else { |
1437 | /* | |
1438 | * We are setting this to NULL here | |
1439 | * because that is what we traditionally | |
1440 | * have been doing. I am not positive | |
1441 | * that this is the right thing to do | |
1442 | * but let's leave the code alone | |
1443 | * for the RIB_SYSTEM_ROUTE case | |
1444 | */ | |
1445 | dest->selected_fib = NULL; | |
1446 | } | |
d62a17ae | 1447 | |
1448 | /* Update nexthop for route, reset changed flag. */ | |
ef57f35f DL |
1449 | /* Note: this code also handles the Linux case when an interface goes |
1450 | * down, causing the kernel to delete routes without sending DELROUTE | |
1451 | * notifications | |
1452 | */ | |
212df1de PG |
1453 | if (!nexthop_active_update(rn, old, 1) && |
1454 | (RIB_KERNEL_ROUTE(old))) | |
1455 | SET_FLAG(old->status, ROUTE_ENTRY_REMOVED); | |
1456 | else | |
1457 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1458 | } |
1459 | ||
1460 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
1461 | struct route_node *rn, | |
1462 | struct route_entry *old, | |
1463 | struct route_entry *new) | |
1464 | { | |
1465 | struct nexthop *nexthop = NULL; | |
1466 | int nh_active = 0; | |
5f7a4718 | 1467 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
d62a17ae | 1468 | |
1469 | /* | |
1470 | * We have to install or update if a new route has been selected or | |
1471 | * something has changed. | |
1472 | */ | |
1473 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
1474 | hook_call(rib_update, rn, "updating existing route"); | |
1475 | ||
1476 | /* Update the nexthop; we could determine here that nexthop is | |
1477 | * inactive. */ | |
1478 | if (nexthop_active_update(rn, new, 1)) | |
1479 | nh_active = 1; | |
1480 | ||
1481 | /* If nexthop is active, install the selected route, if | |
1482 | * appropriate. If | |
1483 | * the install succeeds, cleanup flags for prior route, if | |
1484 | * different from | |
1485 | * newly selected. | |
1486 | */ | |
1487 | if (nh_active) { | |
1488 | if (IS_ZEBRA_DEBUG_RIB) { | |
1489 | char buf[SRCDEST2STR_BUFFER]; | |
1490 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1491 | if (new != old) | |
1492 | zlog_debug( | |
1493 | "%u:%s: Updating route rn %p, re %p (type %d) " | |
1494 | "old %p (type %d)", | |
1495 | zvrf_id(zvrf), buf, rn, new, | |
1496 | new->type, old, old->type); | |
1497 | else | |
1498 | zlog_debug( | |
1499 | "%u:%s: Updating route rn %p, re %p (type %d)", | |
1500 | zvrf_id(zvrf), buf, rn, new, | |
1501 | new->type); | |
1502 | } | |
1503 | ||
1504 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1505 | if (zebra_rib_labeled_unicast(old)) | |
1506 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1507 | ||
1508 | /* Non-system route should be installed. */ | |
1509 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1510 | /* If labeled-unicast route, install transit | |
1511 | * LSP. */ | |
1512 | if (zebra_rib_labeled_unicast(new)) | |
1513 | zebra_mpls_lsp_install(zvrf, rn, new); | |
1514 | ||
0c555cc6 | 1515 | rib_install_kernel(rn, new, old); |
ed216282 DS |
1516 | } else { |
1517 | /* | |
1518 | * We do not need to install the | |
1519 | * selected route because it | |
1520 | * is already isntalled by | |
1521 | * the system( ie not us ) | |
1522 | * so just mark it as winning | |
1523 | * we do need to ensure that | |
1524 | * if we uninstall a route | |
1525 | * from ourselves we don't | |
1526 | * over write this pointer | |
1527 | */ | |
1528 | dest->selected_fib = NULL; | |
d62a17ae | 1529 | } |
d62a17ae | 1530 | /* If install succeeded or system route, cleanup flags |
1531 | * for prior route. */ | |
ed216282 | 1532 | if (new != old) { |
d62a17ae | 1533 | if (RIB_SYSTEM_ROUTE(new)) { |
1534 | if (!RIB_SYSTEM_ROUTE(old)) | |
1535 | rib_uninstall_kernel(rn, old); | |
1536 | } else { | |
7ee30f28 | 1537 | for (nexthop = old->ng.nexthop; nexthop; |
d62a17ae | 1538 | nexthop = nexthop->next) |
1539 | UNSET_FLAG(nexthop->flags, | |
1540 | NEXTHOP_FLAG_FIB); | |
1541 | } | |
1542 | } | |
d62a17ae | 1543 | } |
a64448ba | 1544 | |
d62a17ae | 1545 | /* |
1546 | * If nexthop for selected route is not active or install | |
1547 | * failed, we | |
1548 | * may need to uninstall and delete for redistribution. | |
1549 | */ | |
ed216282 | 1550 | if (!nh_active) { |
d62a17ae | 1551 | if (IS_ZEBRA_DEBUG_RIB) { |
1552 | char buf[SRCDEST2STR_BUFFER]; | |
1553 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1554 | if (new != old) | |
1555 | zlog_debug( | |
1556 | "%u:%s: Deleting route rn %p, re %p (type %d) " | |
0cfbff74 | 1557 | "old %p (type %d) - nexthop inactive", |
d62a17ae | 1558 | zvrf_id(zvrf), buf, rn, new, |
0cfbff74 | 1559 | new->type, old, old->type); |
d62a17ae | 1560 | else |
1561 | zlog_debug( | |
0cfbff74 | 1562 | "%u:%s: Deleting route rn %p, re %p (type %d) - nexthop inactive", |
d62a17ae | 1563 | zvrf_id(zvrf), buf, rn, new, |
0cfbff74 | 1564 | new->type); |
d62a17ae | 1565 | } |
1566 | ||
1567 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1568 | if (zebra_rib_labeled_unicast(old)) | |
1569 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1570 | ||
1571 | if (!RIB_SYSTEM_ROUTE(old)) | |
1572 | rib_uninstall_kernel(rn, old); | |
ed216282 DS |
1573 | else |
1574 | dest->selected_fib = NULL; | |
d62a17ae | 1575 | } |
1576 | } else { | |
1577 | /* | |
1578 | * Same route selected; check if in the FIB and if not, | |
1579 | * re-install. This | |
1580 | * is housekeeping code to deal with race conditions in kernel | |
1581 | * with linux | |
1582 | * netlink reporting interface up before IPv4 or IPv6 protocol | |
1583 | * is ready | |
1584 | * to add routes. | |
1585 | */ | |
1586 | if (!RIB_SYSTEM_ROUTE(new)) { | |
ed216282 | 1587 | bool in_fib = false; |
d62a17ae | 1588 | |
7ee30f28 | 1589 | for (ALL_NEXTHOPS(new->ng, nexthop)) |
d62a17ae | 1590 | if (CHECK_FLAG(nexthop->flags, |
1591 | NEXTHOP_FLAG_FIB)) { | |
ed216282 | 1592 | in_fib = true; |
d62a17ae | 1593 | break; |
1594 | } | |
1595 | if (!in_fib) | |
1596 | rib_install_kernel(rn, new, NULL); | |
1597 | } | |
1598 | } | |
1599 | ||
1600 | /* Update prior route. */ | |
1601 | if (new != old) { | |
d62a17ae | 1602 | /* Set real nexthop. */ |
1603 | nexthop_active_update(rn, old, 1); | |
1604 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
1605 | } | |
3e5c6e00 | 1606 | |
d62a17ae | 1607 | /* Clear changed flag. */ |
1608 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 1609 | } |
1610 | ||
d62a17ae | 1611 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
1612 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1613 | struct route_entry *alternate) | |
1614 | { | |
1615 | if (current == NULL) | |
1616 | return alternate; | |
1617 | ||
1618 | /* filter route selection in following order: | |
1619 | * - connected beats other types | |
fec4ca19 | 1620 | * - if both connected, loopback or vrf wins |
d62a17ae | 1621 | * - lower distance beats higher |
1622 | * - lower metric beats higher for equal distance | |
1623 | * - last, hence oldest, route wins tie break. | |
1624 | */ | |
1625 | ||
fec4ca19 DS |
1626 | /* Connected routes. Check to see if either are a vrf |
1627 | * or loopback interface. If not, pick the last connected | |
d62a17ae | 1628 | * route of the set of lowest metric connected routes. |
1629 | */ | |
1630 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
fec4ca19 DS |
1631 | if (current->type != ZEBRA_ROUTE_CONNECT) |
1632 | return alternate; | |
1633 | ||
1634 | /* both are connected. are either loop or vrf? */ | |
1635 | struct nexthop *nexthop = NULL; | |
1636 | ||
1637 | for (ALL_NEXTHOPS(alternate->ng, nexthop)) { | |
1638 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1639 | nexthop->ifindex, alternate->vrf_id))) | |
1640 | return alternate; | |
1641 | } | |
1642 | ||
1643 | for (ALL_NEXTHOPS(current->ng, nexthop)) { | |
1644 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1645 | nexthop->ifindex, current->vrf_id))) | |
1646 | return current; | |
1647 | } | |
1648 | ||
1649 | /* Neither are loop or vrf so pick best metric */ | |
1650 | if (alternate->metric <= current->metric) | |
d62a17ae | 1651 | return alternate; |
1652 | ||
1653 | return current; | |
1654 | } | |
3e5c6e00 | 1655 | |
d62a17ae | 1656 | if (current->type == ZEBRA_ROUTE_CONNECT) |
1657 | return current; | |
3e5c6e00 | 1658 | |
d62a17ae | 1659 | /* higher distance loses */ |
1660 | if (alternate->distance < current->distance) | |
1661 | return alternate; | |
1662 | if (current->distance < alternate->distance) | |
1663 | return current; | |
3e5c6e00 | 1664 | |
d62a17ae | 1665 | /* metric tie-breaks equal distance */ |
1666 | if (alternate->metric <= current->metric) | |
1667 | return alternate; | |
3e5c6e00 | 1668 | |
d62a17ae | 1669 | return current; |
3e5c6e00 | 1670 | } |
1671 | ||
d62a17ae | 1672 | /* Core function for processing routing information base. */ |
1673 | static void rib_process(struct route_node *rn) | |
1674 | { | |
1675 | struct route_entry *re; | |
1676 | struct route_entry *next; | |
1677 | struct route_entry *old_selected = NULL; | |
1678 | struct route_entry *new_selected = NULL; | |
1679 | struct route_entry *old_fib = NULL; | |
1680 | struct route_entry *new_fib = NULL; | |
1681 | struct route_entry *best = NULL; | |
1682 | char buf[SRCDEST2STR_BUFFER]; | |
1683 | rib_dest_t *dest; | |
1684 | struct zebra_vrf *zvrf = NULL; | |
86391e56 MS |
1685 | const struct prefix *p, *src_p; |
1686 | ||
d62a17ae | 1687 | srcdest_rnode_prefixes(rn, &p, &src_p); |
1688 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1689 | ||
1690 | assert(rn); | |
1691 | ||
1692 | dest = rib_dest_from_rnode(rn); | |
1693 | if (dest) { | |
1694 | zvrf = rib_dest_vrf(dest); | |
1695 | vrf_id = zvrf_id(zvrf); | |
1696 | } | |
bab85d4f | 1697 | |
d62a17ae | 1698 | if (IS_ZEBRA_DEBUG_RIB) |
1699 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1700 | |
d62a17ae | 1701 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1702 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1703 | |
607425e5 DS |
1704 | /* |
1705 | * we can have rn's that have a NULL info pointer | |
1706 | * (dest). As such let's not let the deref happen | |
1707 | * additionally we know RNODE_FOREACH_RE_SAFE | |
1708 | * will not iterate so we are ok. | |
1709 | */ | |
1710 | if (dest) | |
1711 | old_fib = dest->selected_fib; | |
5f7a4718 | 1712 | |
a2addae8 | 1713 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1714 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1715 | zlog_debug( | |
1716 | "%u:%s: Examine re %p (type %d) status %x flags %x " | |
1717 | "dist %d metric %d", | |
1718 | vrf_id, buf, re, re->type, re->status, | |
1719 | re->flags, re->distance, re->metric); | |
1720 | ||
1721 | UNSET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1722 | ||
1723 | /* Currently selected re. */ | |
1724 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1725 | assert(old_selected == NULL); | |
1726 | old_selected = re; | |
1727 | } | |
bab85d4f | 1728 | |
d62a17ae | 1729 | /* Skip deleted entries from selection */ |
1730 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1731 | continue; | |
1732 | ||
1733 | /* Skip unreachable nexthop. */ | |
1734 | /* This first call to nexthop_active_update is merely to | |
1735 | * determine if | |
1736 | * there's any change to nexthops associated with this RIB | |
1737 | * entry. Now, | |
1738 | * rib_process() can be invoked due to an external event such as | |
1739 | * link | |
1740 | * down or due to next-hop-tracking evaluation. In the latter | |
1741 | * case, | |
1742 | * a decision has already been made that the NHs have changed. | |
1743 | * So, no | |
1744 | * need to invoke a potentially expensive call again. Further, | |
1745 | * since | |
1746 | * the change might be in a recursive NH which is not caught in | |
1747 | * the nexthop_active_update() code. Thus, we might miss changes | |
1748 | * to | |
1749 | * recursive NHs. | |
1750 | */ | |
1751 | if (!CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) | |
1752 | && !nexthop_active_update(rn, re, 0)) { | |
1753 | if (re->type == ZEBRA_ROUTE_TABLE) { | |
1754 | /* XXX: HERE BE DRAGONS!!!!! | |
1755 | * In all honesty, I have not yet figured out | |
1756 | * what this part | |
1757 | * does or why the ROUTE_ENTRY_CHANGED test | |
1758 | * above is correct | |
1759 | * or why we need to delete a route here, and | |
1760 | * also not whether | |
1761 | * this concerns both selected and fib route, or | |
1762 | * only selected | |
1763 | * or only fib */ | |
1764 | /* This entry was denied by the 'ip protocol | |
1765 | * table' route-map, we | |
1766 | * need to delete it */ | |
1767 | if (re != old_selected) { | |
1768 | if (IS_ZEBRA_DEBUG_RIB) | |
1769 | zlog_debug( | |
32391aff | 1770 | "%s: %u:%s: imported via import-table but denied " |
d62a17ae | 1771 | "by the ip protocol table route-map", |
32391aff | 1772 | __func__, vrf_id, buf); |
d62a17ae | 1773 | rib_unlink(rn, re); |
1774 | } else | |
1775 | SET_FLAG(re->status, | |
1776 | ROUTE_ENTRY_REMOVED); | |
1777 | } | |
1778 | ||
1779 | continue; | |
1780 | } | |
bab85d4f | 1781 | |
d62a17ae | 1782 | /* Infinite distance. */ |
1783 | if (re->distance == DISTANCE_INFINITY) { | |
1784 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1785 | continue; | |
1786 | } | |
bab85d4f | 1787 | |
d62a17ae | 1788 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1789 | best = rib_choose_best(new_fib, re); | |
1790 | if (new_fib && best != new_fib) | |
1791 | UNSET_FLAG(new_fib->status, | |
1792 | ROUTE_ENTRY_CHANGED); | |
1793 | new_fib = best; | |
1794 | } else { | |
1795 | best = rib_choose_best(new_selected, re); | |
1796 | if (new_selected && best != new_selected) | |
1797 | UNSET_FLAG(new_selected->status, | |
1798 | ROUTE_ENTRY_CHANGED); | |
1799 | new_selected = best; | |
1800 | } | |
1801 | if (best != re) | |
1802 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1803 | } /* RNODE_FOREACH_RE */ | |
1804 | ||
1805 | /* If no FIB override route, use the selected route also for FIB */ | |
1806 | if (new_fib == NULL) | |
1807 | new_fib = new_selected; | |
1808 | ||
1809 | /* After the cycle is finished, the following pointers will be set: | |
1810 | * old_selected --- RE entry currently having SELECTED | |
1811 | * new_selected --- RE entry that is newly SELECTED | |
1812 | * old_fib --- RE entry currently in kernel FIB | |
1813 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1814 | * | |
1815 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1816 | * the zclients. new_fib (which can be new_selected) will be installed | |
1817 | * in kernel. | |
1818 | */ | |
1819 | ||
1820 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1821 | zlog_debug( | |
fe2c53d4 | 1822 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", |
d62a17ae | 1823 | vrf_id, buf, (void *)old_selected, (void *)new_selected, |
1824 | (void *)old_fib, (void *)new_fib); | |
1825 | } | |
446bb95e | 1826 | |
d62a17ae | 1827 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1828 | * fib == selected */ | |
9d303b37 DL |
1829 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, |
1830 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1831 | |
1832 | /* Update fib according to selection results */ | |
1833 | if (new_fib && old_fib) | |
1834 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1835 | else if (new_fib) | |
1836 | rib_process_add_fib(zvrf, rn, new_fib); | |
1837 | else if (old_fib) | |
1838 | rib_process_del_fib(zvrf, rn, old_fib); | |
1839 | ||
8cb41cd6 | 1840 | /* Update SELECTED entry */ |
d62a17ae | 1841 | if (old_selected != new_selected || selected_changed) { |
d62a17ae | 1842 | |
d62a17ae | 1843 | if (new_selected && new_selected != new_fib) { |
1844 | nexthop_active_update(rn, new_selected, 1); | |
1845 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); | |
1846 | } | |
41ec9222 | 1847 | |
5af4b346 MS |
1848 | if (new_selected) { |
1849 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); | |
1850 | ||
1851 | /* Special case: new route is system route, so | |
1852 | * dataplane update will not be done - ensure we | |
1853 | * redistribute the route. | |
1854 | */ | |
1855 | if (RIB_SYSTEM_ROUTE(new_selected)) | |
1856 | redistribute_update(p, src_p, new_selected, | |
1857 | old_selected); | |
1858 | } | |
1859 | ||
d62a17ae | 1860 | if (old_selected) { |
1861 | if (!new_selected) | |
1862 | redistribute_delete(p, src_p, old_selected); | |
1863 | if (old_selected != new_selected) | |
1864 | UNSET_FLAG(old_selected->flags, | |
1865 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1866 | } |
d62a17ae | 1867 | } |
3e5c6e00 | 1868 | |
d62a17ae | 1869 | /* Remove all RE entries queued for removal */ |
a2addae8 | 1870 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1871 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { |
1872 | if (IS_ZEBRA_DEBUG_RIB) { | |
1873 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1874 | (void *)rn, (void *)re); | |
1875 | } | |
1876 | rib_unlink(rn, re); | |
1877 | } | |
1878 | } | |
4d38fdb4 | 1879 | |
d62a17ae | 1880 | /* |
1881 | * Check if the dest can be deleted now. | |
1882 | */ | |
1883 | rib_gc_dest(rn); | |
e96f9203 DO |
1884 | } |
1885 | ||
e5ac2adf MS |
1886 | /* |
1887 | * Utility to match route with dplane context data | |
1888 | */ | |
1889 | static bool rib_route_match_ctx(const struct route_entry *re, | |
1890 | const dplane_ctx_h ctx, bool is_update) | |
1891 | { | |
1892 | bool result = false; | |
1893 | ||
1894 | if (is_update) { | |
1895 | /* | |
1896 | * In 'update' case, we test info about the 'previous' or | |
1897 | * 'old' route | |
1898 | */ | |
1899 | if ((re->type == dplane_ctx_get_old_type(ctx)) && | |
1900 | (re->instance == dplane_ctx_get_old_instance(ctx))) { | |
1901 | result = true; | |
1902 | ||
1903 | /* TODO -- we're using this extra test, but it's not | |
1904 | * exactly clear why. | |
1905 | */ | |
1906 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1907 | (re->distance != dplane_ctx_get_old_distance(ctx) || | |
1908 | re->tag != dplane_ctx_get_old_tag(ctx))) { | |
1909 | result = false; | |
1910 | } | |
1911 | } | |
1912 | ||
1913 | } else { | |
1914 | /* | |
1915 | * Ordinary, single-route case using primary context info | |
1916 | */ | |
1917 | if ((dplane_ctx_get_op(ctx) != DPLANE_OP_ROUTE_DELETE) && | |
1918 | CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1919 | /* Skip route that's been deleted */ | |
1920 | goto done; | |
1921 | } | |
1922 | ||
1923 | if ((re->type == dplane_ctx_get_type(ctx)) && | |
1924 | (re->instance == dplane_ctx_get_instance(ctx))) { | |
1925 | result = true; | |
1926 | ||
1927 | /* TODO -- we're using this extra test, but it's not | |
1928 | * exactly clear why. | |
1929 | */ | |
1930 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1931 | (re->distance != dplane_ctx_get_distance(ctx) || | |
1932 | re->tag != dplane_ctx_get_tag(ctx))) { | |
1933 | result = false; | |
1934 | } | |
1935 | } | |
1936 | } | |
1937 | ||
1938 | done: | |
1939 | ||
1940 | return (result); | |
1941 | } | |
1942 | ||
1943 | /* | |
97f5b441 MS |
1944 | * TODO - WIP version of route-update processing after async dataplane |
1945 | * update. | |
e5ac2adf MS |
1946 | */ |
1947 | static void rib_process_after(dplane_ctx_h ctx) | |
1948 | { | |
1949 | struct route_table *table = NULL; | |
97f5b441 | 1950 | struct zebra_vrf *zvrf = NULL; |
e5ac2adf MS |
1951 | struct route_node *rn = NULL; |
1952 | struct route_entry *re = NULL, *old_re = NULL, *rib; | |
1953 | bool is_update = false; | |
1954 | struct nexthop *nexthop; | |
97f5b441 | 1955 | char dest_str[PREFIX_STRLEN] = ""; |
5709131c | 1956 | enum dplane_op_e op; |
e5ac2adf MS |
1957 | enum zebra_dplane_result status; |
1958 | const struct prefix *dest_pfx, *src_pfx; | |
1959 | ||
1960 | /* Locate rn and re(s) from ctx */ | |
1961 | ||
1962 | table = zebra_vrf_table_with_table_id(dplane_ctx_get_afi(ctx), | |
1963 | dplane_ctx_get_safi(ctx), | |
1964 | dplane_ctx_get_vrf(ctx), | |
1965 | dplane_ctx_get_table(ctx)); | |
1966 | if (table == NULL) { | |
1967 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fe2c53d4 | 1968 | zlog_debug("Failed to process dplane results: no table for afi %d, safi %d, vrf %u", |
e5ac2adf MS |
1969 | dplane_ctx_get_afi(ctx), |
1970 | dplane_ctx_get_safi(ctx), | |
1971 | dplane_ctx_get_vrf(ctx)); | |
1972 | } | |
1973 | goto done; | |
1974 | } | |
1975 | ||
97f5b441 MS |
1976 | zvrf = vrf_info_lookup(dplane_ctx_get_vrf(ctx)); |
1977 | ||
e5ac2adf MS |
1978 | dest_pfx = dplane_ctx_get_dest(ctx); |
1979 | ||
1980 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1981 | * unconditional log messages will have to generate the string. | |
1982 | */ | |
5709131c | 1983 | if (IS_ZEBRA_DEBUG_DPLANE) |
e5ac2adf | 1984 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
e5ac2adf MS |
1985 | |
1986 | src_pfx = dplane_ctx_get_src(ctx); | |
1987 | rn = srcdest_rnode_get(table, dplane_ctx_get_dest(ctx), | |
5709131c | 1988 | src_pfx ? (struct prefix_ipv6 *)src_pfx : NULL); |
e5ac2adf MS |
1989 | if (rn == NULL) { |
1990 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fe2c53d4 | 1991 | zlog_debug("Failed to process dplane results: no route for %u:%s", |
e5ac2adf MS |
1992 | dplane_ctx_get_vrf(ctx), dest_str); |
1993 | } | |
1994 | goto done; | |
1995 | } | |
1996 | ||
1997 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); | |
1998 | ||
1999 | op = dplane_ctx_get_op(ctx); | |
2000 | status = dplane_ctx_get_status(ctx); | |
2001 | ||
2002 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) { | |
2003 | zlog_debug("%u:%s Processing dplane ctx %p, op %s result %d", | |
2004 | dplane_ctx_get_vrf(ctx), dest_str, ctx, | |
2005 | dplane_op2str(op), status); | |
2006 | } | |
2007 | ||
2008 | if (op == DPLANE_OP_ROUTE_DELETE) { | |
2009 | /* | |
2010 | * In the delete case, the zebra core datastructs were | |
2011 | * updated (or removed) at the time the delete was issued, | |
2012 | * so we're just notifying the route owner. | |
2013 | */ | |
2014 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
2015 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_REMOVED); | |
97f5b441 | 2016 | |
5709131c | 2017 | if (zvrf) |
97f5b441 | 2018 | zvrf->removals++; |
e5ac2adf MS |
2019 | } else { |
2020 | zsend_route_notify_owner_ctx(ctx, | |
2021 | ZAPI_ROUTE_FAIL_INSTALL); | |
2022 | ||
2023 | zlog_warn("%u:%s: Route Deletion failure", | |
2024 | dplane_ctx_get_vrf(ctx), | |
2025 | prefix2str(dest_pfx, | |
2026 | dest_str, sizeof(dest_str))); | |
2027 | } | |
2028 | ||
2029 | /* Nothing more to do in delete case */ | |
2030 | goto done; | |
2031 | } | |
2032 | ||
2033 | /* | |
2034 | * Update is a bit of a special case, where we may have both old and new | |
2035 | * routes to post-process. | |
2036 | */ | |
2037 | is_update = dplane_ctx_is_update(ctx); | |
2038 | ||
2039 | /* | |
2040 | * Take a pass through the routes, look for matches with the context | |
2041 | * info. | |
2042 | */ | |
2043 | RNODE_FOREACH_RE(rn, rib) { | |
2044 | ||
2045 | if (re == NULL) { | |
5709131c | 2046 | if (rib_route_match_ctx(rib, ctx, false)) |
e5ac2adf | 2047 | re = rib; |
e5ac2adf MS |
2048 | } |
2049 | ||
2050 | /* Check for old route match */ | |
2051 | if (is_update && (old_re == NULL)) { | |
5709131c | 2052 | if (rib_route_match_ctx(rib, ctx, true /*is_update*/)) |
e5ac2adf | 2053 | old_re = rib; |
e5ac2adf MS |
2054 | } |
2055 | ||
2056 | /* Have we found the routes we need to work on? */ | |
5709131c | 2057 | if (re && ((!is_update || old_re))) |
e5ac2adf | 2058 | break; |
e5ac2adf MS |
2059 | } |
2060 | ||
2061 | /* | |
2062 | * Check sequence number(s) to detect stale results before continuing | |
2063 | */ | |
2064 | if (re && (re->dplane_sequence != dplane_ctx_get_seq(ctx))) { | |
2065 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) { | |
2066 | zlog_debug("%u:%s Stale dplane result for re %p", | |
2067 | dplane_ctx_get_vrf(ctx), dest_str, re); | |
2068 | } | |
2069 | re = NULL; | |
2070 | } | |
2071 | ||
2072 | if (old_re && | |
2073 | (old_re->dplane_sequence != dplane_ctx_get_old_seq(ctx))) { | |
2074 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) { | |
2075 | zlog_debug("%u:%s Stale dplane result for old_re %p", | |
2076 | dplane_ctx_get_vrf(ctx), dest_str, old_re); | |
2077 | } | |
2078 | old_re = NULL; | |
2079 | } | |
2080 | ||
2081 | /* | |
2082 | * Here's sort of a tough one: the route update result is stale. | |
2083 | * Is it better to use the context block info to generate | |
2084 | * redist and owner notification, or is it better to wait | |
2085 | * for the up-to-date result to arrive? | |
2086 | */ | |
2087 | if (re == NULL) { | |
2088 | /* TODO -- for now, only expose up-to-date results */ | |
2089 | goto done; | |
2090 | } | |
2091 | ||
2092 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
2093 | /* Set nexthop FIB flags */ | |
2094 | for (ALL_NEXTHOPS(re->ng, nexthop)) { | |
2095 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
2096 | continue; | |
2097 | ||
2098 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
2099 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
2100 | else | |
2101 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
2102 | } | |
2103 | ||
5709131c | 2104 | if (zvrf) |
97f5b441 | 2105 | zvrf->installs++; |
97f5b441 | 2106 | |
e5ac2adf MS |
2107 | /* Redistribute */ |
2108 | /* TODO -- still calling the redist api using the route_entries, | |
2109 | * and there's a corner-case here: if there's no client | |
2110 | * for the 'new' route, a redist deleting the 'old' route | |
2111 | * will be sent. But if the 'old' context info was stale, | |
2112 | * 'old_re' will be NULL here and that delete will not be sent. | |
2113 | */ | |
2114 | redistribute_update(dest_pfx, src_pfx, re, old_re); | |
2115 | ||
2116 | /* Notify route owner */ | |
2117 | zsend_route_notify_owner(re, | |
2118 | dest_pfx, ZAPI_ROUTE_INSTALLED); | |
2119 | ||
2120 | } else { | |
2121 | zsend_route_notify_owner(re, dest_pfx, | |
2122 | ZAPI_ROUTE_FAIL_INSTALL); | |
2123 | ||
2124 | zlog_warn("%u:%s: Route install failed", | |
2125 | dplane_ctx_get_vrf(ctx), | |
2126 | prefix2str(dest_pfx, | |
2127 | dest_str, sizeof(dest_str))); | |
2128 | } | |
2129 | ||
2130 | done: | |
2131 | ||
2132 | /* Return context to dataplane module */ | |
2133 | dplane_ctx_fini(&ctx); | |
2134 | } | |
2135 | ||
5110a0c6 | 2136 | /* Take a list of route_node structs and return 1, if there was a record |
d62a17ae | 2137 | * picked from it and processed by rib_process(). Don't process more, |
5110a0c6 | 2138 | * than one RN record; operate only in the specified sub-queue. |
e96f9203 | 2139 | */ |
d7c0a89a | 2140 | static unsigned int process_subq(struct list *subq, uint8_t qindex) |
e96f9203 | 2141 | { |
d62a17ae | 2142 | struct listnode *lnode = listhead(subq); |
2143 | struct route_node *rnode; | |
2144 | rib_dest_t *dest; | |
2145 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 2146 | |
d62a17ae | 2147 | if (!lnode) |
2148 | return 0; | |
5110a0c6 | 2149 | |
d62a17ae | 2150 | rnode = listgetdata(lnode); |
2151 | dest = rib_dest_from_rnode(rnode); | |
2152 | if (dest) | |
2153 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 2154 | |
d62a17ae | 2155 | rib_process(rnode); |
5110a0c6 | 2156 | |
d62a17ae | 2157 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
2158 | char buf[SRCDEST2STR_BUFFER]; | |
2159 | srcdest_rnode2str(rnode, buf, sizeof(buf)); | |
2160 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
2161 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
2162 | } | |
41ec9222 | 2163 | |
d62a17ae | 2164 | if (rnode->info) |
2165 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
2166 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 2167 | |
67b9467f | 2168 | #if 0 |
5110a0c6 SH |
2169 | else |
2170 | { | |
2171 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
2172 | __func__, rnode, rnode->lock); | |
2173 | zlog_backtrace(LOG_DEBUG); | |
2174 | } | |
67b9467f | 2175 | #endif |
d62a17ae | 2176 | route_unlock_node(rnode); |
2177 | list_delete_node(subq, lnode); | |
2178 | return 1; | |
e96f9203 DO |
2179 | } |
2180 | ||
fb018d25 DS |
2181 | /* |
2182 | * All meta queues have been processed. Trigger next-hop evaluation. | |
2183 | */ | |
d62a17ae | 2184 | static void meta_queue_process_complete(struct work_queue *dummy) |
fb018d25 | 2185 | { |
d62a17ae | 2186 | struct vrf *vrf; |
2187 | struct zebra_vrf *zvrf; | |
9ec6b0bb | 2188 | |
d62a17ae | 2189 | /* Evaluate nexthops for those VRFs which underwent route processing. |
2190 | * This | |
2191 | * should limit the evaluation to the necessary VRFs in most common | |
2192 | * situations. | |
2193 | */ | |
a2addae8 | 2194 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
d62a17ae | 2195 | zvrf = vrf->info; |
2196 | if (zvrf == NULL || !(zvrf->flags & ZEBRA_VRF_RIB_SCHEDULED)) | |
2197 | continue; | |
2198 | ||
2199 | zvrf->flags &= ~ZEBRA_VRF_RIB_SCHEDULED; | |
6d53d7b1 | 2200 | zebra_evaluate_rnh(zvrf, AF_INET, 0, RNH_NEXTHOP_TYPE, NULL); |
2201 | zebra_evaluate_rnh(zvrf, AF_INET, 0, RNH_IMPORT_CHECK_TYPE, | |
d62a17ae | 2202 | NULL); |
6d53d7b1 | 2203 | zebra_evaluate_rnh(zvrf, AF_INET6, 0, RNH_NEXTHOP_TYPE, NULL); |
2204 | zebra_evaluate_rnh(zvrf, AF_INET6, 0, RNH_IMPORT_CHECK_TYPE, | |
d62a17ae | 2205 | NULL); |
d62a17ae | 2206 | } |
939fba27 | 2207 | |
d62a17ae | 2208 | /* Schedule LSPs for processing, if needed. */ |
2209 | zvrf = vrf_info_lookup(VRF_DEFAULT); | |
2210 | if (mpls_should_lsps_be_processed(zvrf)) { | |
2211 | if (IS_ZEBRA_DEBUG_MPLS) | |
2212 | zlog_debug( | |
2213 | "%u: Scheduling all LSPs upon RIB completion", | |
2214 | zvrf_id(zvrf)); | |
2215 | zebra_mpls_lsp_schedule(zvrf); | |
2216 | mpls_unmark_lsps_for_processing(zvrf); | |
2217 | } | |
fb018d25 DS |
2218 | } |
2219 | ||
e96f9203 | 2220 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from |
d62a17ae | 2221 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
2222 | * data | |
e96f9203 DO |
2223 | * is pointed to the meta queue structure. |
2224 | */ | |
d62a17ae | 2225 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 2226 | { |
d62a17ae | 2227 | struct meta_queue *mq = data; |
2228 | unsigned i; | |
5110a0c6 | 2229 | |
d62a17ae | 2230 | for (i = 0; i < MQ_SIZE; i++) |
2231 | if (process_subq(mq->subq[i], i)) { | |
2232 | mq->size--; | |
2233 | break; | |
2234 | } | |
2235 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
2236 | } |
2237 | ||
9fd92e3c AS |
2238 | /* |
2239 | * Map from rib types to queue type (priority) in meta queue | |
2240 | */ | |
d7c0a89a QY |
2241 | static const uint8_t meta_queue_map[ZEBRA_ROUTE_MAX] = { |
2242 | [ZEBRA_ROUTE_SYSTEM] = 4, | |
2243 | [ZEBRA_ROUTE_KERNEL] = 0, | |
2244 | [ZEBRA_ROUTE_CONNECT] = 0, | |
2245 | [ZEBRA_ROUTE_STATIC] = 1, | |
2246 | [ZEBRA_ROUTE_RIP] = 2, | |
2247 | [ZEBRA_ROUTE_RIPNG] = 2, | |
2248 | [ZEBRA_ROUTE_OSPF] = 2, | |
2249 | [ZEBRA_ROUTE_OSPF6] = 2, | |
2250 | [ZEBRA_ROUTE_ISIS] = 2, | |
2251 | [ZEBRA_ROUTE_BGP] = 3, | |
2252 | [ZEBRA_ROUTE_PIM] = 4, // Shouldn't happen but for safety | |
2253 | [ZEBRA_ROUTE_EIGRP] = 2, | |
2254 | [ZEBRA_ROUTE_NHRP] = 2, | |
2255 | [ZEBRA_ROUTE_HSLS] = 4, | |
2256 | [ZEBRA_ROUTE_OLSR] = 4, | |
2257 | [ZEBRA_ROUTE_TABLE] = 1, | |
2258 | [ZEBRA_ROUTE_LDP] = 4, | |
2259 | [ZEBRA_ROUTE_VNC] = 3, | |
2260 | [ZEBRA_ROUTE_VNC_DIRECT] = 3, | |
2261 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = 3, | |
2262 | [ZEBRA_ROUTE_BGP_DIRECT] = 3, | |
2263 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = 3, | |
2264 | [ZEBRA_ROUTE_BABEL] = 2, | |
2265 | [ZEBRA_ROUTE_ALL] = 4, // Shouldn't happen but for safety | |
5110a0c6 SH |
2266 | }; |
2267 | ||
2268 | /* Look into the RN and queue it into one or more priority queues, | |
2269 | * increasing the size for each data push done. | |
e96f9203 | 2270 | */ |
d62a17ae | 2271 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) |
e96f9203 | 2272 | { |
d62a17ae | 2273 | struct route_entry *re; |
5110a0c6 | 2274 | |
a2addae8 | 2275 | RNODE_FOREACH_RE (rn, re) { |
d7c0a89a | 2276 | uint8_t qindex = meta_queue_map[re->type]; |
d62a17ae | 2277 | struct zebra_vrf *zvrf; |
2278 | ||
2279 | /* Invariant: at this point we always have rn->info set. */ | |
2280 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
2281 | RIB_ROUTE_QUEUED(qindex))) { | |
2282 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2283 | rnode_debug( | |
2284 | rn, re->vrf_id, | |
2285 | "rn %p is already queued in sub-queue %u", | |
2286 | (void *)rn, qindex); | |
2287 | continue; | |
2288 | } | |
5110a0c6 | 2289 | |
d62a17ae | 2290 | SET_FLAG(rib_dest_from_rnode(rn)->flags, |
2291 | RIB_ROUTE_QUEUED(qindex)); | |
2292 | listnode_add(mq->subq[qindex], rn); | |
2293 | route_lock_node(rn); | |
2294 | mq->size++; | |
5110a0c6 | 2295 | |
d62a17ae | 2296 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2297 | rnode_debug(rn, re->vrf_id, | |
2298 | "queued rn %p into sub-queue %u", | |
2299 | (void *)rn, qindex); | |
9ec6b0bb | 2300 | |
d62a17ae | 2301 | zvrf = zebra_vrf_lookup_by_id(re->vrf_id); |
2302 | if (zvrf) | |
2303 | zvrf->flags |= ZEBRA_VRF_RIB_SCHEDULED; | |
2304 | } | |
4d38fdb4 | 2305 | } |
2306 | ||
6d691129 | 2307 | /* Add route_node to work queue and schedule processing */ |
d62a17ae | 2308 | void rib_queue_add(struct route_node *rn) |
4d38fdb4 | 2309 | { |
d62a17ae | 2310 | assert(rn); |
fc328ac9 | 2311 | |
d62a17ae | 2312 | /* Pointless to queue a route_node with no RIB entries to add or remove |
2313 | */ | |
2314 | if (!rnode_to_ribs(rn)) { | |
2315 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
2316 | __func__, (void *)rn, rn->lock); | |
2317 | zlog_backtrace(LOG_DEBUG); | |
2318 | return; | |
2319 | } | |
4d38fdb4 | 2320 | |
d62a17ae | 2321 | if (zebrad.ribq == NULL) { |
e914ccbe | 2322 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2323 | "%s: work_queue does not exist!", __func__); |
d62a17ae | 2324 | return; |
2325 | } | |
2326 | ||
2327 | /* | |
2328 | * The RIB queue should normally be either empty or holding the only | |
2329 | * work_queue_item element. In the latter case this element would | |
2330 | * hold a pointer to the meta queue structure, which must be used to | |
2331 | * actually queue the route nodes to process. So create the MQ | |
2332 | * holder, if necessary, then push the work into it in any case. | |
2333 | * This semantics was introduced after 0.99.9 release. | |
2334 | */ | |
f104f6c1 | 2335 | if (work_queue_empty(zebrad.ribq)) |
d62a17ae | 2336 | work_queue_add(zebrad.ribq, zebrad.mq); |
e96f9203 | 2337 | |
d62a17ae | 2338 | rib_meta_queue_add(zebrad.mq, rn); |
fc328ac9 | 2339 | |
d62a17ae | 2340 | return; |
4d38fdb4 | 2341 | } |
2342 | ||
5110a0c6 SH |
2343 | /* Create new meta queue. |
2344 | A destructor function doesn't seem to be necessary here. | |
2345 | */ | |
d62a17ae | 2346 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 2347 | { |
d62a17ae | 2348 | struct meta_queue *new; |
2349 | unsigned i; | |
5110a0c6 | 2350 | |
d62a17ae | 2351 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
e96f9203 | 2352 | |
d62a17ae | 2353 | for (i = 0; i < MQ_SIZE; i++) { |
2354 | new->subq[i] = list_new(); | |
2355 | assert(new->subq[i]); | |
2356 | } | |
5110a0c6 | 2357 | |
d62a17ae | 2358 | return new; |
e96f9203 DO |
2359 | } |
2360 | ||
d62a17ae | 2361 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 2362 | { |
d62a17ae | 2363 | unsigned i; |
5a8dfcd8 | 2364 | |
d62a17ae | 2365 | for (i = 0; i < MQ_SIZE; i++) |
6a154c88 | 2366 | list_delete(&mq->subq[i]); |
5a8dfcd8 | 2367 | |
d62a17ae | 2368 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
2369 | } |
2370 | ||
4d38fdb4 | 2371 | /* initialise zebra rib work queue */ |
d62a17ae | 2372 | static void rib_queue_init(struct zebra_t *zebra) |
4d38fdb4 | 2373 | { |
d62a17ae | 2374 | assert(zebra); |
2375 | ||
2376 | if (!(zebra->ribq = | |
2377 | work_queue_new(zebra->master, "route_node processing"))) { | |
e914ccbe | 2378 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2379 | "%s: could not initialise work queue!", __func__); |
d62a17ae | 2380 | return; |
2381 | } | |
4d38fdb4 | 2382 | |
d62a17ae | 2383 | /* fill in the work queue spec */ |
2384 | zebra->ribq->spec.workfunc = &meta_queue_process; | |
2385 | zebra->ribq->spec.errorfunc = NULL; | |
2386 | zebra->ribq->spec.completion_func = &meta_queue_process_complete; | |
2387 | /* XXX: TODO: These should be runtime configurable via vty */ | |
2388 | zebra->ribq->spec.max_retries = 3; | |
3a30f50f | 2389 | zebra->ribq->spec.hold = ZEBRA_RIB_PROCESS_HOLD_TIME; |
d62a17ae | 2390 | |
2391 | if (!(zebra->mq = meta_queue_new())) { | |
e914ccbe | 2392 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2393 | "%s: could not initialise meta queue!", __func__); |
d62a17ae | 2394 | return; |
2395 | } | |
2396 | return; | |
718e3744 | 2397 | } |
2398 | ||
6d691129 PJ |
2399 | /* RIB updates are processed via a queue of pointers to route_nodes. |
2400 | * | |
2401 | * The queue length is bounded by the maximal size of the routing table, | |
2402 | * as a route_node will not be requeued, if already queued. | |
2403 | * | |
f0f77c9a DS |
2404 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
2405 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 2406 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 2407 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 2408 | * |
f0f77c9a | 2409 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
2410 | * |
2411 | * rib_addnode | |
f0f77c9a DS |
2412 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
2413 | * |-------->| | best RE, if required | |
3c0755dc PJ |
2414 | * | | |
2415 | * static_install->|->rib_addqueue...... -> rib_process | |
2416 | * | | | |
2417 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
2418 | * |-> set ROUTE_ENTRY_REMOVE | |
2419 | * rib_delnode (RE freed) | |
6d691129 | 2420 | * |
9fd92e3c AS |
2421 | * The 'info' pointer of a route_node points to a rib_dest_t |
2422 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
2423 | * dest is created on-demand by rib_link() and is kept around at least | |
2424 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 2425 | * |
6d691129 PJ |
2426 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
2427 | * | |
2428 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
2429 | * - dest attached to route_node: |
2430 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
2431 | * - route_node processing queue |
2432 | * - managed by: rib_addqueue, rib_process. | |
2433 | * | |
2434 | */ | |
d62a17ae | 2435 | |
f0f77c9a | 2436 | /* Add RE to head of the route node. */ |
d62a17ae | 2437 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
2438 | { | |
2439 | struct route_entry *head; | |
2440 | rib_dest_t *dest; | |
2441 | afi_t afi; | |
2442 | const char *rmap_name; | |
9fd92e3c | 2443 | |
d62a17ae | 2444 | assert(re && rn); |
9fd92e3c | 2445 | |
d62a17ae | 2446 | dest = rib_dest_from_rnode(rn); |
2447 | if (!dest) { | |
2448 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2449 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 2450 | |
d62a17ae | 2451 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); |
2452 | route_lock_node(rn); /* rn route table reference */ | |
2453 | rn->info = dest; | |
2454 | dest->rnode = rn; | |
2455 | } | |
2263a412 | 2456 | |
d62a17ae | 2457 | head = dest->routes; |
2458 | if (head) { | |
2459 | head->prev = re; | |
2460 | } | |
2461 | re->next = head; | |
2462 | dest->routes = re; | |
2463 | ||
2464 | afi = (rn->p.family == AF_INET) | |
2465 | ? AFI_IP | |
2466 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2467 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2468 | rmap_name = zebra_get_import_table_route_map(afi, re->table); | |
2469 | zebra_add_import_table_entry(rn, re, rmap_name); | |
2470 | } else if (process) | |
2471 | rib_queue_add(rn); | |
2472 | } | |
2473 | ||
7e24fdf3 DS |
2474 | static void rib_addnode(struct route_node *rn, |
2475 | struct route_entry *re, int process) | |
d62a17ae | 2476 | { |
2477 | /* RE node has been un-removed before route-node is processed. | |
2478 | * route_node must hence already be on the queue for processing.. | |
2479 | */ | |
2480 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
2481 | if (IS_ZEBRA_DEBUG_RIB) | |
2482 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
2483 | (void *)rn, (void *)re); | |
2484 | ||
2485 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2486 | return; | |
2487 | } | |
2488 | rib_link(rn, re, process); | |
6d691129 PJ |
2489 | } |
2490 | ||
9fd92e3c AS |
2491 | /* |
2492 | * rib_unlink | |
2493 | * | |
2494 | * Detach a rib structure from a route_node. | |
2495 | * | |
2496 | * Note that a call to rib_unlink() should be followed by a call to | |
2497 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
2498 | * longer required to be deleted. | |
2499 | */ | |
d62a17ae | 2500 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 2501 | { |
d62a17ae | 2502 | rib_dest_t *dest; |
9fd92e3c | 2503 | |
d62a17ae | 2504 | assert(rn && re); |
6d691129 | 2505 | |
d62a17ae | 2506 | if (IS_ZEBRA_DEBUG_RIB) |
2507 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
2508 | (void *)re); | |
6d691129 | 2509 | |
d62a17ae | 2510 | dest = rib_dest_from_rnode(rn); |
6d691129 | 2511 | |
d62a17ae | 2512 | if (re->next) |
2513 | re->next->prev = re->prev; | |
6d691129 | 2514 | |
d62a17ae | 2515 | if (re->prev) |
2516 | re->prev->next = re->next; | |
2517 | else { | |
2518 | dest->routes = re->next; | |
2519 | } | |
7a4bb9c5 | 2520 | |
2eb07de3 DS |
2521 | if (dest->selected_fib == re) |
2522 | dest->selected_fib = NULL; | |
2523 | ||
7ee30f28 | 2524 | nexthops_free(re->ng.nexthop); |
d62a17ae | 2525 | XFREE(MTYPE_RE, re); |
2526 | } | |
2527 | ||
2528 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2529 | { | |
2530 | afi_t afi; | |
2531 | ||
2532 | if (IS_ZEBRA_DEBUG_RIB) | |
2533 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2534 | (void *)rn, (void *)re); | |
2535 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2536 | ||
2537 | afi = (rn->p.family == AF_INET) | |
2538 | ? AFI_IP | |
2539 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2540 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2541 | zebra_del_import_table_entry(rn, re); | |
2542 | /* Just clean up if non main table */ | |
2543 | if (IS_ZEBRA_DEBUG_RIB) { | |
2544 | char buf[SRCDEST2STR_BUFFER]; | |
2545 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2546 | zlog_debug( | |
2547 | "%u:%s: Freeing route rn %p, re %p (type %d)", | |
2548 | re->vrf_id, buf, rn, re, re->type); | |
2549 | } | |
7a4bb9c5 | 2550 | |
d62a17ae | 2551 | rib_unlink(rn, re); |
2552 | } else { | |
2553 | rib_queue_add(rn); | |
2554 | } | |
718e3744 | 2555 | } |
2556 | ||
f0f77c9a | 2557 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2558 | * standard debug log. Calling function name and IP prefix in |
2559 | * question are passed as 1st and 2nd arguments. | |
2560 | */ | |
2561 | ||
d62a17ae | 2562 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2563 | union prefixconstptr src_pp, | |
2564 | const struct route_entry *re) | |
2565 | { | |
d62a17ae | 2566 | const struct prefix *src_p = src_pp.p; |
2567 | bool is_srcdst = src_p && src_p->prefixlen; | |
2568 | char straddr[PREFIX_STRLEN]; | |
2569 | char srcaddr[PREFIX_STRLEN]; | |
2570 | struct nexthop *nexthop; | |
2571 | ||
2572 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2573 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2574 | is_srcdst ? " from " : "", | |
2575 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2576 | : "", | |
2577 | re->vrf_id); | |
cc54cfee RW |
2578 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", |
2579 | func, (unsigned long)re->uptime, re->type, re->instance, | |
2580 | re->table); | |
d62a17ae | 2581 | zlog_debug( |
2582 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
2583 | func, re->metric, re->mtu, re->distance, re->flags, re->status); | |
2584 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func, | |
2585 | re->nexthop_num, re->nexthop_active_num); | |
2586 | ||
7ee30f28 | 2587 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
2d68a0f2 DS |
2588 | struct interface *ifp; |
2589 | struct vrf *vrf = vrf_lookup_by_id(nexthop->vrf_id); | |
2590 | ||
2591 | switch (nexthop->type) { | |
2592 | case NEXTHOP_TYPE_BLACKHOLE: | |
2593 | sprintf(straddr, "Blackhole"); | |
2594 | break; | |
2595 | case NEXTHOP_TYPE_IFINDEX: | |
2596 | ifp = if_lookup_by_index(nexthop->ifindex, | |
2597 | nexthop->vrf_id); | |
2598 | sprintf(straddr, "%s", ifp ? ifp->name : "Unknown"); | |
2599 | break; | |
2600 | case NEXTHOP_TYPE_IPV4: | |
2601 | /* fallthrough */ | |
2602 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
2603 | inet_ntop(AF_INET, &nexthop->gate, straddr, | |
2604 | INET6_ADDRSTRLEN); | |
2605 | break; | |
2606 | case NEXTHOP_TYPE_IPV6: | |
2607 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
2608 | inet_ntop(AF_INET6, &nexthop->gate, straddr, | |
2609 | INET6_ADDRSTRLEN); | |
2610 | break; | |
2611 | } | |
2612 | zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s", func, | |
d62a17ae | 2613 | (nexthop->rparent ? " NH" : "NH"), straddr, |
2d68a0f2 DS |
2614 | nexthop->ifindex, vrf ? vrf->name : "Unknown", |
2615 | nexthop->vrf_id, | |
d62a17ae | 2616 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) |
2617 | ? "ACTIVE " | |
2618 | : ""), | |
2619 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB) | |
2620 | ? "FIB " | |
2621 | : ""), | |
2622 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
2623 | ? "RECURSIVE" | |
2624 | : "")); | |
2625 | } | |
2626 | zlog_debug("%s: dump complete", func); | |
dc95824a DO |
2627 | } |
2628 | ||
2629 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2630 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2631 | */ |
2632 | ||
d62a17ae | 2633 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2634 | { | |
2635 | struct route_table *table; | |
2636 | struct route_node *rn; | |
2637 | struct route_entry *re; | |
2638 | char prefix_buf[INET_ADDRSTRLEN]; | |
2639 | ||
2640 | /* Lookup table. */ | |
2641 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2642 | if (!table) { | |
e914ccbe | 2643 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2644 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2645 | vrf_id); | |
d62a17ae | 2646 | return; |
2647 | } | |
2648 | ||
2649 | /* Scan the RIB table for exactly matching RE entry. */ | |
2650 | rn = route_node_lookup(table, (struct prefix *)p); | |
2651 | ||
2652 | /* No route for this prefix. */ | |
2653 | if (!rn) { | |
32391aff | 2654 | zlog_debug("%s:%u lookup failed for %s", __func__, vrf_id, |
d62a17ae | 2655 | prefix2str((struct prefix *)p, prefix_buf, |
2656 | sizeof(prefix_buf))); | |
2657 | return; | |
2658 | } | |
2659 | ||
2660 | /* Unlock node. */ | |
2661 | route_unlock_node(rn); | |
2662 | ||
2663 | /* let's go */ | |
a2addae8 | 2664 | RNODE_FOREACH_RE (rn, re) { |
32391aff DS |
2665 | zlog_debug("%s:%u rn %p, re %p: %s, %s", |
2666 | __func__, vrf_id, | |
2667 | (void *)rn, (void *)re, | |
d62a17ae | 2668 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) |
2669 | ? "removed" | |
2670 | : "NOT removed"), | |
2671 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2672 | ? "selected" | |
2673 | : "NOT selected")); | |
2674 | route_entry_dump(p, NULL, re); | |
2675 | } | |
dc95824a DO |
2676 | } |
2677 | ||
20e5ff0a DO |
2678 | /* Check if requested address assignment will fail due to another |
2679 | * route being installed by zebra in FIB already. Take necessary | |
2680 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2681 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2682 | */ |
d62a17ae | 2683 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2684 | { | |
2685 | struct route_table *table; | |
2686 | struct route_node *rn; | |
d62a17ae | 2687 | unsigned changed = 0; |
5f7a4718 | 2688 | rib_dest_t *dest; |
d62a17ae | 2689 | |
2690 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
e914ccbe | 2691 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2692 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2693 | vrf_id); | |
d62a17ae | 2694 | return; |
2695 | } | |
2696 | ||
2697 | /* No matches would be the simplest case. */ | |
2698 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2699 | return; | |
2700 | ||
2701 | /* Unlock node. */ | |
2702 | route_unlock_node(rn); | |
2703 | ||
5f7a4718 | 2704 | dest = rib_dest_from_rnode(rn); |
d62a17ae | 2705 | /* Check all RE entries. In case any changes have to be done, requeue |
2706 | * the RN into RIBQ head. If the routing message about the new connected | |
2707 | * route (generated by the IP address we are going to assign very soon) | |
2708 | * comes before the RIBQ is processed, the new RE entry will join | |
2709 | * RIBQ record already on head. This is necessary for proper | |
2710 | * revalidation | |
2711 | * of the rest of the RE. | |
2712 | */ | |
5f7a4718 DS |
2713 | if (dest->selected_fib && !RIB_SYSTEM_ROUTE(dest->selected_fib)) { |
2714 | changed = 1; | |
2715 | if (IS_ZEBRA_DEBUG_RIB) { | |
2716 | char buf[PREFIX_STRLEN]; | |
2717 | ||
2718 | zlog_debug("%u:%s: freeing way for connected prefix", | |
2719 | dest->selected_fib->vrf_id, | |
2720 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2721 | route_entry_dump(&rn->p, NULL, dest->selected_fib); | |
d62a17ae | 2722 | } |
5f7a4718 | 2723 | rib_uninstall(rn, dest->selected_fib); |
d62a17ae | 2724 | } |
2725 | if (changed) | |
2726 | rib_queue_add(rn); | |
20e5ff0a DO |
2727 | } |
2728 | ||
d62a17ae | 2729 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
2730 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
718e3744 | 2731 | { |
d62a17ae | 2732 | struct route_table *table; |
2733 | struct route_node *rn; | |
40ecd8e4 | 2734 | struct route_entry *same = NULL; |
d62a17ae | 2735 | struct nexthop *nexthop; |
2736 | int ret = 0; | |
b4c034b0 | 2737 | |
d62a17ae | 2738 | if (!re) |
2739 | return 0; | |
b4c034b0 | 2740 | |
1f610a1f | 2741 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
05737783 | 2742 | |
d62a17ae | 2743 | /* Lookup table. */ |
7865c65d RW |
2744 | table = zebra_vrf_table_with_table_id(afi, safi, re->vrf_id, re->table); |
2745 | if (!table) { | |
2746 | XFREE(MTYPE_RE, re); | |
d62a17ae | 2747 | return 0; |
7865c65d | 2748 | } |
cddf391b | 2749 | |
d62a17ae | 2750 | /* Make it sure prefixlen is applied to the prefix. */ |
2751 | apply_mask(p); | |
2752 | if (src_p) | |
2753 | apply_mask_ipv6(src_p); | |
718e3744 | 2754 | |
d62a17ae | 2755 | /* Set default distance by route type. */ |
2756 | if (re->distance == 0) { | |
0492eea0 | 2757 | re->distance = route_distance(re->type); |
718e3744 | 2758 | |
d62a17ae | 2759 | /* iBGP distance is 200. */ |
2760 | if (re->type == ZEBRA_ROUTE_BGP | |
2761 | && CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
2762 | re->distance = 200; | |
2763 | } | |
718e3744 | 2764 | |
d62a17ae | 2765 | /* Lookup route node.*/ |
2766 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2767 | |
40ecd8e4 DS |
2768 | /* |
2769 | * If same type of route are installed, treat it as a implicit | |
2770 | * withdraw. | |
2771 | * If the user has specified the No route replace semantics | |
2772 | * for the install don't do a route replace. | |
2773 | */ | |
a2addae8 | 2774 | RNODE_FOREACH_RE (rn, same) { |
d62a17ae | 2775 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) |
2776 | continue; | |
41ec9222 | 2777 | |
eb327fa5 RW |
2778 | if (same->type != re->type) |
2779 | continue; | |
2780 | if (same->instance != re->instance) | |
2781 | continue; | |
996c9314 LB |
2782 | if (same->type == ZEBRA_ROUTE_KERNEL |
2783 | && same->metric != re->metric) | |
eb327fa5 | 2784 | continue; |
40ecd8e4 DS |
2785 | |
2786 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && | |
2787 | same->distance != re->distance) | |
2788 | continue; | |
2789 | ||
844b3a87 | 2790 | /* |
40ecd8e4 DS |
2791 | * We should allow duplicate connected routes |
2792 | * because of IPv6 link-local routes and unnumbered | |
2793 | * interfaces on Linux. | |
844b3a87 RW |
2794 | */ |
2795 | if (same->type != ZEBRA_ROUTE_CONNECT) | |
d62a17ae | 2796 | break; |
2797 | } | |
718e3744 | 2798 | |
d62a17ae | 2799 | /* If this route is kernel route, set FIB flag to the route. */ |
8628fc61 | 2800 | if (RIB_SYSTEM_ROUTE(re)) |
7ee30f28 | 2801 | for (nexthop = re->ng.nexthop; nexthop; nexthop = nexthop->next) |
d62a17ae | 2802 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
718e3744 | 2803 | |
d62a17ae | 2804 | /* Link new re to node.*/ |
2805 | if (IS_ZEBRA_DEBUG_RIB) { | |
2806 | rnode_debug( | |
2807 | rn, re->vrf_id, | |
2808 | "Inserting route rn %p, re %p (type %d) existing %p", | |
2809 | (void *)rn, (void *)re, re->type, (void *)same); | |
718e3744 | 2810 | |
d62a17ae | 2811 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2812 | route_entry_dump(p, src_p, re); | |
718e3744 | 2813 | } |
d62a17ae | 2814 | rib_addnode(rn, re, 1); |
2815 | ret = 1; | |
6b0655a2 | 2816 | |
d62a17ae | 2817 | /* Free implicit route.*/ |
2818 | if (same) { | |
2819 | rib_delnode(rn, same); | |
2820 | ret = -1; | |
2821 | } | |
718e3744 | 2822 | |
d62a17ae | 2823 | route_unlock_node(rn); |
2824 | return ret; | |
2825 | } | |
2826 | ||
2827 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
d7c0a89a | 2828 | unsigned short instance, int flags, struct prefix *p, |
fd36be7e | 2829 | struct prefix_ipv6 *src_p, const struct nexthop *nh, |
40ecd8e4 DS |
2830 | uint32_t table_id, uint32_t metric, uint8_t distance, |
2831 | bool fromkernel) | |
d62a17ae | 2832 | { |
2833 | struct route_table *table; | |
2834 | struct route_node *rn; | |
2835 | struct route_entry *re; | |
2836 | struct route_entry *fib = NULL; | |
2837 | struct route_entry *same = NULL; | |
fd36be7e | 2838 | struct nexthop *rtnh; |
d62a17ae | 2839 | char buf2[INET6_ADDRSTRLEN]; |
5f7a4718 | 2840 | rib_dest_t *dest; |
d62a17ae | 2841 | |
1f610a1f | 2842 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
d62a17ae | 2843 | |
2844 | /* Lookup table. */ | |
2845 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2846 | if (!table) | |
2847 | return; | |
2848 | ||
2849 | /* Apply mask. */ | |
2850 | apply_mask(p); | |
2851 | if (src_p) | |
2852 | apply_mask_ipv6(src_p); | |
2853 | ||
2854 | /* Lookup route node. */ | |
2855 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2856 | if (!rn) { | |
2857 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2858 | ||
2859 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2860 | if (src_p && src_p->prefixlen) | |
2861 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2862 | else | |
2863 | src_buf[0] = '\0'; | |
2864 | ||
2865 | if (IS_ZEBRA_DEBUG_RIB) | |
2866 | zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id, | |
2867 | dst_buf, | |
2868 | (src_buf[0] != '\0') ? " from " : "", | |
2869 | src_buf); | |
2870 | return; | |
2871 | } | |
718e3744 | 2872 | |
5f7a4718 DS |
2873 | dest = rib_dest_from_rnode(rn); |
2874 | fib = dest->selected_fib; | |
2875 | ||
d62a17ae | 2876 | /* Lookup same type route. */ |
a2addae8 | 2877 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2878 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2879 | continue; | |
2880 | ||
d62a17ae | 2881 | if (re->type != type) |
2882 | continue; | |
2883 | if (re->instance != instance) | |
2884 | continue; | |
40ecd8e4 DS |
2885 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && |
2886 | distance != re->distance) | |
2887 | continue; | |
2888 | ||
996c9314 | 2889 | if (re->type == ZEBRA_ROUTE_KERNEL && re->metric != metric) |
f19435a8 | 2890 | continue; |
7ee30f28 | 2891 | if (re->type == ZEBRA_ROUTE_CONNECT && (rtnh = re->ng.nexthop) |
fd36be7e DL |
2892 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { |
2893 | if (rtnh->ifindex != nh->ifindex) | |
d62a17ae | 2894 | continue; |
d62a17ae | 2895 | same = re; |
2896 | break; | |
2897 | } | |
2898 | /* Make sure that the route found has the same gateway. */ | |
2899 | else { | |
fd36be7e | 2900 | if (nh == NULL) { |
d62a17ae | 2901 | same = re; |
2902 | break; | |
2903 | } | |
7ee30f28 | 2904 | for (ALL_NEXTHOPS(re->ng, rtnh)) |
fd36be7e | 2905 | if (nexthop_same_no_recurse(rtnh, nh)) { |
d62a17ae | 2906 | same = re; |
2907 | break; | |
2908 | } | |
2909 | if (same) | |
2910 | break; | |
2911 | } | |
2912 | } | |
2913 | /* If same type of route can't be found and this message is from | |
2914 | kernel. */ | |
2915 | if (!same) { | |
5dfeba19 DS |
2916 | /* |
2917 | * In the past(HA!) we could get here because | |
2918 | * we were receiving a route delete from the | |
2919 | * kernel and we're not marking the proto | |
2920 | * as coming from it's appropriate originator. | |
2921 | * Now that we are properly noticing the fact | |
2922 | * that the kernel has deleted our route we | |
2923 | * are not going to get called in this path | |
2924 | * I am going to leave this here because | |
2925 | * this might still work this way on non-linux | |
2926 | * platforms as well as some weird state I have | |
2927 | * not properly thought of yet. | |
2928 | * If we can show that this code path is | |
2929 | * dead then we can remove it. | |
2930 | */ | |
b8faa875 | 2931 | if (fib && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { |
d62a17ae | 2932 | if (IS_ZEBRA_DEBUG_RIB) { |
2933 | rnode_debug( | |
2934 | rn, vrf_id, | |
2935 | "rn %p, re %p (type %d) was deleted from kernel, adding", | |
2936 | rn, fib, fib->type); | |
2937 | } | |
2938 | if (allow_delete) { | |
2939 | /* Unset flags. */ | |
7ee30f28 | 2940 | for (rtnh = fib->ng.nexthop; rtnh; |
fd36be7e DL |
2941 | rtnh = rtnh->next) |
2942 | UNSET_FLAG(rtnh->flags, | |
d62a17ae | 2943 | NEXTHOP_FLAG_FIB); |
2944 | ||
ed216282 DS |
2945 | /* |
2946 | * This is a non FRR route | |
2947 | * as such we should mark | |
2948 | * it as deleted | |
2949 | */ | |
5f7a4718 | 2950 | dest->selected_fib = NULL; |
d62a17ae | 2951 | } else { |
2952 | /* This means someone else, other than Zebra, | |
2953 | * has deleted | |
2954 | * a Zebra router from the kernel. We will add | |
2955 | * it back */ | |
2956 | rib_install_kernel(rn, fib, NULL); | |
2957 | } | |
2958 | } else { | |
2959 | if (IS_ZEBRA_DEBUG_RIB) { | |
fd36be7e | 2960 | if (nh) |
d62a17ae | 2961 | rnode_debug( |
2962 | rn, vrf_id, | |
2963 | "via %s ifindex %d type %d " | |
2964 | "doesn't exist in rib", | |
36228974 | 2965 | inet_ntop(afi2family(afi), |
2966 | &nh->gate, buf2, | |
2967 | sizeof(buf2)), | |
2968 | nh->ifindex, type); | |
d62a17ae | 2969 | else |
2970 | rnode_debug( | |
2971 | rn, vrf_id, | |
fd36be7e DL |
2972 | "type %d doesn't exist in rib", |
2973 | type); | |
d62a17ae | 2974 | } |
2975 | route_unlock_node(rn); | |
2976 | return; | |
2977 | } | |
2978 | } | |
718e3744 | 2979 | |
5dfeba19 | 2980 | if (same) { |
996c9314 LB |
2981 | if (fromkernel && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE) |
2982 | && !allow_delete) { | |
5dfeba19 DS |
2983 | rib_install_kernel(rn, same, NULL); |
2984 | route_unlock_node(rn); | |
2985 | ||
2986 | return; | |
2987 | } | |
6134fd82 | 2988 | |
90264d64 | 2989 | if (CHECK_FLAG(flags, ZEBRA_FLAG_EVPN_ROUTE)) { |
6134fd82 | 2990 | struct nexthop *tmp_nh; |
2991 | ||
7ee30f28 | 2992 | for (ALL_NEXTHOPS(re->ng, tmp_nh)) { |
6134fd82 | 2993 | struct ipaddr vtep_ip; |
2994 | ||
2995 | memset(&vtep_ip, 0, sizeof(struct ipaddr)); | |
1ec31309 | 2996 | if (afi == AFI_IP) { |
2997 | vtep_ip.ipa_type = IPADDR_V4; | |
2998 | memcpy(&(vtep_ip.ipaddr_v4), | |
2999 | &(tmp_nh->gate.ipv4), | |
3000 | sizeof(struct in_addr)); | |
3001 | } else { | |
3002 | vtep_ip.ipa_type = IPADDR_V6; | |
3003 | memcpy(&(vtep_ip.ipaddr_v6), | |
3004 | &(tmp_nh->gate.ipv6), | |
3005 | sizeof(struct in6_addr)); | |
3006 | } | |
a317a9b9 | 3007 | zebra_vxlan_evpn_vrf_route_del(re->vrf_id, |
6134fd82 | 3008 | &vtep_ip, p); |
3009 | } | |
3010 | } | |
d62a17ae | 3011 | rib_delnode(rn, same); |
5dfeba19 | 3012 | } |
05737783 | 3013 | |
d62a17ae | 3014 | route_unlock_node(rn); |
3015 | return; | |
3016 | } | |
718e3744 | 3017 | |
718e3744 | 3018 | |
d7c0a89a QY |
3019 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, |
3020 | unsigned short instance, int flags, struct prefix *p, | |
3021 | struct prefix_ipv6 *src_p, const struct nexthop *nh, | |
3022 | uint32_t table_id, uint32_t metric, uint32_t mtu, uint8_t distance, | |
3023 | route_tag_t tag) | |
d62a17ae | 3024 | { |
3025 | struct route_entry *re; | |
66af6845 | 3026 | struct nexthop *nexthop; |
718e3744 | 3027 | |
66af6845 | 3028 | /* Allocate new route_entry structure. */ |
d62a17ae | 3029 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); |
d62a17ae | 3030 | re->type = type; |
3031 | re->instance = instance; | |
3032 | re->distance = distance; | |
3033 | re->flags = flags; | |
3034 | re->metric = metric; | |
3035 | re->mtu = mtu; | |
3036 | re->table = table_id; | |
3037 | re->vrf_id = vrf_id; | |
3038 | re->nexthop_num = 0; | |
3039 | re->uptime = time(NULL); | |
4e40b6d6 | 3040 | re->tag = tag; |
d62a17ae | 3041 | |
66af6845 RW |
3042 | /* Add nexthop. */ |
3043 | nexthop = nexthop_new(); | |
3044 | *nexthop = *nh; | |
3045 | route_entry_nexthop_add(re, nexthop); | |
718e3744 | 3046 | |
66af6845 | 3047 | return rib_add_multipath(afi, safi, p, src_p, re); |
718e3744 | 3048 | } |
3049 | ||
1c848137 | 3050 | /* Schedule routes of a particular table (address-family) based on event. */ |
d5b8c216 | 3051 | void rib_update_table(struct route_table *table, rib_update_event_t event) |
d62a17ae | 3052 | { |
3053 | struct route_node *rn; | |
3054 | struct route_entry *re, *next; | |
3055 | ||
3056 | /* Walk all routes and queue for processing, if appropriate for | |
3057 | * the trigger event. | |
3058 | */ | |
3059 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
1ca60f2c DS |
3060 | /* |
3061 | * If we are looking at a route node and the node | |
3062 | * has already been queued we don't | |
3063 | * need to queue it up again | |
3064 | */ | |
996c9314 LB |
3065 | if (rn->info && CHECK_FLAG(rib_dest_from_rnode(rn)->flags, |
3066 | RIB_ROUTE_ANY_QUEUED)) | |
1ca60f2c | 3067 | continue; |
d62a17ae | 3068 | switch (event) { |
3069 | case RIB_UPDATE_IF_CHANGE: | |
3070 | /* Examine all routes that won't get processed by the | |
3071 | * protocol or | |
3072 | * triggered by nexthop evaluation (NHT). This would be | |
3073 | * system, | |
3074 | * kernel and certain static routes. Note that NHT will | |
3075 | * get | |
3076 | * triggered upon an interface event as connected routes | |
3077 | * always | |
3078 | * get queued for processing. | |
3079 | */ | |
a2addae8 | 3080 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
0a16efff DS |
3081 | struct nexthop *nh; |
3082 | ||
996c9314 LB |
3083 | if (re->type != ZEBRA_ROUTE_SYSTEM |
3084 | && re->type != ZEBRA_ROUTE_KERNEL | |
3085 | && re->type != ZEBRA_ROUTE_CONNECT | |
3086 | && re->type != ZEBRA_ROUTE_STATIC) | |
0a16efff DS |
3087 | continue; |
3088 | ||
3089 | if (re->type != ZEBRA_ROUTE_STATIC) { | |
3090 | rib_queue_add(rn); | |
3091 | continue; | |
3092 | } | |
3093 | ||
7ee30f28 | 3094 | for (nh = re->ng.nexthop; nh; nh = nh->next) |
0a16efff DS |
3095 | if (!(nh->type == NEXTHOP_TYPE_IPV4 |
3096 | || nh->type == NEXTHOP_TYPE_IPV6)) | |
3097 | break; | |
3098 | ||
3099 | /* If we only have nexthops to a | |
3100 | * gateway, NHT will | |
3101 | * take care. | |
3102 | */ | |
3103 | if (nh) | |
d62a17ae | 3104 | rib_queue_add(rn); |
3105 | } | |
3106 | break; | |
3107 | ||
3108 | case RIB_UPDATE_RMAP_CHANGE: | |
3109 | case RIB_UPDATE_OTHER: | |
3110 | /* Right now, examine all routes. Can restrict to a | |
3111 | * protocol in | |
3112 | * some cases (TODO). | |
3113 | */ | |
3114 | if (rnode_to_ribs(rn)) | |
3115 | rib_queue_add(rn); | |
3116 | break; | |
3117 | ||
3118 | default: | |
3119 | break; | |
3120 | } | |
3121 | } | |
b84c7253 | 3122 | } |
3123 | ||
718e3744 | 3124 | /* RIB update function. */ |
d62a17ae | 3125 | void rib_update(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 3126 | { |
d62a17ae | 3127 | struct route_table *table; |
1c848137 | 3128 | |
d62a17ae | 3129 | /* Process routes of interested address-families. */ |
3130 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
d5b8c216 | 3131 | if (table) { |
3132 | if (IS_ZEBRA_DEBUG_EVENT) | |
3133 | zlog_debug("%s : AFI_IP event %d", __func__, event); | |
d62a17ae | 3134 | rib_update_table(table, event); |
d5b8c216 | 3135 | } |
718e3744 | 3136 | |
d62a17ae | 3137 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
d5b8c216 | 3138 | if (table) { |
3139 | if (IS_ZEBRA_DEBUG_EVENT) | |
3140 | zlog_debug("%s : AFI_IP6 event %d", __func__, event); | |
d62a17ae | 3141 | rib_update_table(table, event); |
d5b8c216 | 3142 | } |
718e3744 | 3143 | } |
3144 | ||
718e3744 | 3145 | /* Delete self installed routes after zebra is relaunched. */ |
95a29032 | 3146 | void rib_sweep_table(struct route_table *table) |
d62a17ae | 3147 | { |
3148 | struct route_node *rn; | |
3149 | struct route_entry *re; | |
3150 | struct route_entry *next; | |
915902cb | 3151 | struct nexthop *nexthop; |
d62a17ae | 3152 | |
915902cb DS |
3153 | if (!table) |
3154 | return; | |
d62a17ae | 3155 | |
915902cb | 3156 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
a2addae8 | 3157 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
915902cb DS |
3158 | if (IS_ZEBRA_DEBUG_RIB) |
3159 | route_entry_dump(&rn->p, NULL, re); | |
3160 | ||
3161 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
3162 | continue; | |
3163 | ||
3164 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
3165 | continue; | |
3166 | ||
3167 | /* | |
3168 | * So we are starting up and have received | |
3169 | * routes from the kernel that we have installed | |
3170 | * from a previous run of zebra but not cleaned | |
3171 | * up ( say a kill -9 ) | |
3172 | * But since we haven't actually installed | |
3173 | * them yet( we received them from the kernel ) | |
3174 | * we don't think they are active. | |
3175 | * So let's pretend they are active to actually | |
3176 | * remove them. | |
3177 | * In all honesty I'm not sure if we should | |
3178 | * mark them as active when we receive them | |
3179 | * This is startup only so probably ok. | |
3180 | * | |
3181 | * If we ever decide to move rib_sweep_table | |
3182 | * to a different spot (ie startup ) | |
3183 | * this decision needs to be revisited | |
3184 | */ | |
7ee30f28 | 3185 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
915902cb DS |
3186 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
3187 | ||
0c555cc6 DS |
3188 | rib_uninstall_kernel(rn, re); |
3189 | rib_delnode(rn, re); | |
915902cb DS |
3190 | } |
3191 | } | |
718e3744 | 3192 | } |
3193 | ||
3194 | /* Sweep all RIB tables. */ | |
d62a17ae | 3195 | void rib_sweep_route(void) |
718e3744 | 3196 | { |
d62a17ae | 3197 | struct vrf *vrf; |
3198 | struct zebra_vrf *zvrf; | |
78104b9b | 3199 | |
a2addae8 | 3200 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
915902cb DS |
3201 | if ((zvrf = vrf->info) == NULL) |
3202 | continue; | |
3203 | ||
d62a17ae | 3204 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); |
3205 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3206 | } | |
95a29032 DS |
3207 | |
3208 | zebra_ns_sweep_route(); | |
718e3744 | 3209 | } |
2ea1ab1c VT |
3210 | |
3211 | /* Remove specific by protocol routes from 'table'. */ | |
d7c0a89a | 3212 | unsigned long rib_score_proto_table(uint8_t proto, unsigned short instance, |
47a08aa9 | 3213 | struct route_table *table) |
d62a17ae | 3214 | { |
3215 | struct route_node *rn; | |
3216 | struct route_entry *re; | |
3217 | struct route_entry *next; | |
3218 | unsigned long n = 0; | |
3219 | ||
3220 | if (table) | |
3221 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
a2addae8 | 3222 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 3223 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
3224 | continue; | |
3225 | if (re->type == proto | |
3226 | && re->instance == instance) { | |
3227 | rib_delnode(rn, re); | |
3228 | n++; | |
3229 | } | |
3230 | } | |
3231 | return n; | |
2ea1ab1c VT |
3232 | } |
3233 | ||
3234 | /* Remove specific by protocol routes. */ | |
d7c0a89a | 3235 | unsigned long rib_score_proto(uint8_t proto, unsigned short instance) |
2ea1ab1c | 3236 | { |
d62a17ae | 3237 | struct vrf *vrf; |
3238 | struct zebra_vrf *zvrf; | |
3239 | unsigned long cnt = 0; | |
78104b9b | 3240 | |
a2addae8 RW |
3241 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) |
3242 | if ((zvrf = vrf->info) != NULL) | |
3243 | cnt += rib_score_proto_table( | |
3244 | proto, instance, | |
3245 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
3246 | + rib_score_proto_table( | |
3247 | proto, instance, | |
3248 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
78104b9b | 3249 | |
47a08aa9 DS |
3250 | cnt += zebra_ns_score_proto(proto, instance); |
3251 | ||
d62a17ae | 3252 | return cnt; |
2ea1ab1c VT |
3253 | } |
3254 | ||
718e3744 | 3255 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 3256 | void rib_close_table(struct route_table *table) |
718e3744 | 3257 | { |
d62a17ae | 3258 | struct route_node *rn; |
1e9f448f | 3259 | rib_table_info_t *info; |
5f7a4718 | 3260 | rib_dest_t *dest; |
718e3744 | 3261 | |
1e9f448f DS |
3262 | if (!table) |
3263 | return; | |
9fd92e3c | 3264 | |
6ca30e9e | 3265 | info = route_table_get_info(table); |
5adc2528 | 3266 | |
5f7a4718 DS |
3267 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
3268 | dest = rib_dest_from_rnode(rn); | |
1e9f448f | 3269 | |
5f7a4718 | 3270 | if (dest && dest->selected_fib) { |
1e9f448f DS |
3271 | if (info->safi == SAFI_UNICAST) |
3272 | hook_call(rib_update, rn, NULL); | |
3273 | ||
5f7a4718 DS |
3274 | if (!RIB_SYSTEM_ROUTE(dest->selected_fib)) |
3275 | rib_uninstall_kernel(rn, dest->selected_fib); | |
1e9f448f | 3276 | } |
5f7a4718 | 3277 | } |
718e3744 | 3278 | } |
3279 | ||
7cdb1a84 MS |
3280 | /* |
3281 | * | |
3282 | */ | |
3283 | static int rib_process_dplane_results(struct thread *thread) | |
3284 | { | |
3285 | dplane_ctx_h ctx; | |
3286 | ||
3287 | do { | |
3288 | /* Take lock controlling queue of results */ | |
3289 | pthread_mutex_lock(&dplane_mutex); | |
3290 | { | |
3291 | /* Dequeue context block */ | |
3292 | dplane_ctx_dequeue(&rib_dplane_q, &ctx); | |
3293 | } | |
3294 | pthread_mutex_unlock(&dplane_mutex); | |
3295 | ||
5709131c | 3296 | if (ctx) |
e5ac2adf | 3297 | rib_process_after(ctx); |
5709131c | 3298 | else |
7cdb1a84 | 3299 | break; |
7cdb1a84 | 3300 | |
5709131c | 3301 | } while (1); |
7cdb1a84 | 3302 | |
5709131c | 3303 | return 0; |
7cdb1a84 MS |
3304 | } |
3305 | ||
3306 | /* | |
3307 | * Results are returned from the dataplane subsystem, in the context of | |
3308 | * the dataplane thread. We enqueue the results here for processing by | |
3309 | * the main thread later. | |
3310 | */ | |
3311 | static int rib_dplane_results(dplane_ctx_h ctx) | |
3312 | { | |
3313 | /* Take lock controlling queue of results */ | |
3314 | pthread_mutex_lock(&dplane_mutex); | |
3315 | { | |
3316 | /* Enqueue context block */ | |
3317 | dplane_ctx_enqueue_tail(&rib_dplane_q, ctx); | |
3318 | } | |
3319 | pthread_mutex_unlock(&dplane_mutex); | |
3320 | ||
3321 | /* Ensure event is signalled to zebra main thread */ | |
3322 | thread_add_event(zebrad.master, rib_process_dplane_results, NULL, 0, | |
3323 | &t_dplane); | |
3324 | ||
5709131c | 3325 | return 0; |
7cdb1a84 MS |
3326 | } |
3327 | ||
718e3744 | 3328 | /* Routing information base initialize. */ |
d62a17ae | 3329 | void rib_init(void) |
718e3744 | 3330 | { |
d62a17ae | 3331 | rib_queue_init(&zebrad); |
7cdb1a84 MS |
3332 | |
3333 | /* Init dataplane, and register for results */ | |
3334 | pthread_mutex_init(&dplane_mutex, NULL); | |
3335 | TAILQ_INIT(&rib_dplane_q); | |
3336 | zebra_dplane_init(); | |
3337 | dplane_results_register(rib_dplane_results); | |
718e3744 | 3338 | } |
0915bb0c AS |
3339 | |
3340 | /* | |
3341 | * vrf_id_get_next | |
3342 | * | |
3343 | * Get the first vrf id that is greater than the given vrf id if any. | |
3344 | * | |
3345 | * Returns TRUE if a vrf id was found, FALSE otherwise. | |
3346 | */ | |
d62a17ae | 3347 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 3348 | { |
d62a17ae | 3349 | struct vrf *vrf; |
b72ede27 | 3350 | |
d62a17ae | 3351 | vrf = vrf_lookup_by_id(vrf_id); |
3352 | if (vrf) { | |
3353 | vrf = RB_NEXT(vrf_id_head, vrf); | |
3354 | if (vrf) { | |
3355 | *next_id_p = vrf->vrf_id; | |
3356 | return 1; | |
3357 | } | |
3358 | } | |
0915bb0c | 3359 | |
d62a17ae | 3360 | return 0; |
0915bb0c AS |
3361 | } |
3362 | ||
3363 | /* | |
3364 | * rib_tables_iter_next | |
3365 | * | |
3366 | * Returns the next table in the iteration. | |
3367 | */ | |
d62a17ae | 3368 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
3369 | { | |
3370 | struct route_table *table; | |
3371 | ||
3372 | /* | |
3373 | * Array that helps us go over all AFI/SAFI combinations via one | |
3374 | * index. | |
3375 | */ | |
3376 | static struct { | |
3377 | afi_t afi; | |
3378 | safi_t safi; | |
3379 | } afi_safis[] = { | |
3380 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
3381 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
3382 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
3383 | }; | |
3384 | ||
3385 | table = NULL; | |
3386 | ||
3387 | switch (iter->state) { | |
3388 | ||
3389 | case RIB_TABLES_ITER_S_INIT: | |
3390 | iter->vrf_id = VRF_DEFAULT; | |
3391 | iter->afi_safi_ix = -1; | |
3392 | ||
3393 | /* Fall through */ | |
3394 | ||
3395 | case RIB_TABLES_ITER_S_ITERATING: | |
3396 | iter->afi_safi_ix++; | |
3397 | while (1) { | |
3398 | ||
3399 | while (iter->afi_safi_ix | |
3400 | < (int)ZEBRA_NUM_OF(afi_safis)) { | |
3401 | table = zebra_vrf_table( | |
3402 | afi_safis[iter->afi_safi_ix].afi, | |
3403 | afi_safis[iter->afi_safi_ix].safi, | |
3404 | iter->vrf_id); | |
3405 | if (table) | |
3406 | break; | |
3407 | ||
3408 | iter->afi_safi_ix++; | |
3409 | } | |
3410 | ||
3411 | /* | |
3412 | * Found another table in this vrf. | |
3413 | */ | |
3414 | if (table) | |
3415 | break; | |
3416 | ||
3417 | /* | |
3418 | * Done with all tables in the current vrf, go to the | |
3419 | * next | |
3420 | * one. | |
3421 | */ | |
3422 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
3423 | break; | |
3424 | ||
3425 | iter->afi_safi_ix = 0; | |
3426 | } | |
0915bb0c | 3427 | |
0915bb0c AS |
3428 | break; |
3429 | ||
d62a17ae | 3430 | case RIB_TABLES_ITER_S_DONE: |
3431 | return NULL; | |
0915bb0c AS |
3432 | } |
3433 | ||
d62a17ae | 3434 | if (table) |
3435 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
3436 | else | |
3437 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 3438 | |
d62a17ae | 3439 | return table; |
0915bb0c | 3440 | } |