]>
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 | |
89272910 | 41 | #include "zebra/zebra_router.h" |
43e52561 QY |
42 | #include "zebra/connected.h" |
43 | #include "zebra/debug.h" | |
44 | #include "zebra/interface.h" | |
45 | #include "zebra/redistribute.h" | |
718e3744 | 46 | #include "zebra/rib.h" |
47 | #include "zebra/rt.h" | |
43e52561 QY |
48 | #include "zebra/zapi_msg.h" |
49 | #include "zebra/zebra_errors.h" | |
50 | #include "zebra/zebra_memory.h" | |
7c551956 | 51 | #include "zebra/zebra_ns.h" |
fb018d25 | 52 | #include "zebra/zebra_rnh.h" |
43e52561 QY |
53 | #include "zebra/zebra_routemap.h" |
54 | #include "zebra/zebra_vrf.h" | |
6134fd82 | 55 | #include "zebra/zebra_vxlan.h" |
7cdb1a84 MS |
56 | #include "zebra/zapi_msg.h" |
57 | #include "zebra/zebra_dplane.h" | |
58 | ||
59 | /* | |
60 | * Event, list, and mutex for delivery of dataplane results | |
61 | */ | |
62 | static pthread_mutex_t dplane_mutex; | |
63 | static struct thread *t_dplane; | |
25779064 | 64 | static struct dplane_ctx_q rib_dplane_q; |
718e3744 | 65 | |
d62a17ae | 66 | DEFINE_HOOK(rib_update, (struct route_node * rn, const char *reason), |
67 | (rn, reason)) | |
4f8ea50c | 68 | |
6baf7bb8 DS |
69 | /* Should we allow non Quagga processes to delete our routes */ |
70 | extern int allow_delete; | |
71 | ||
718e3744 | 72 | /* Each route type's string and default distance value. */ |
d62a17ae | 73 | static const struct { |
74 | int key; | |
75 | int distance; | |
76 | } route_info[ZEBRA_ROUTE_MAX] = { | |
9d303b37 DL |
77 | [ZEBRA_ROUTE_SYSTEM] = {ZEBRA_ROUTE_SYSTEM, 0}, |
78 | [ZEBRA_ROUTE_KERNEL] = {ZEBRA_ROUTE_KERNEL, 0}, | |
79 | [ZEBRA_ROUTE_CONNECT] = {ZEBRA_ROUTE_CONNECT, 0}, | |
80 | [ZEBRA_ROUTE_STATIC] = {ZEBRA_ROUTE_STATIC, 1}, | |
81 | [ZEBRA_ROUTE_RIP] = {ZEBRA_ROUTE_RIP, 120}, | |
82 | [ZEBRA_ROUTE_RIPNG] = {ZEBRA_ROUTE_RIPNG, 120}, | |
83 | [ZEBRA_ROUTE_OSPF] = {ZEBRA_ROUTE_OSPF, 110}, | |
84 | [ZEBRA_ROUTE_OSPF6] = {ZEBRA_ROUTE_OSPF6, 110}, | |
85 | [ZEBRA_ROUTE_ISIS] = {ZEBRA_ROUTE_ISIS, 115}, | |
86 | [ZEBRA_ROUTE_BGP] = {ZEBRA_ROUTE_BGP, 20 /* IBGP is 200. */}, | |
c710b277 DS |
87 | [ZEBRA_ROUTE_PIM] = {ZEBRA_ROUTE_PIM, 255}, |
88 | [ZEBRA_ROUTE_EIGRP] = {ZEBRA_ROUTE_EIGRP, 90}, | |
9d303b37 | 89 | [ZEBRA_ROUTE_NHRP] = {ZEBRA_ROUTE_NHRP, 10}, |
c710b277 DS |
90 | [ZEBRA_ROUTE_HSLS] = {ZEBRA_ROUTE_HSLS, 255}, |
91 | [ZEBRA_ROUTE_OLSR] = {ZEBRA_ROUTE_OLSR, 255}, | |
92 | [ZEBRA_ROUTE_TABLE] = {ZEBRA_ROUTE_TABLE, 150}, | |
93 | [ZEBRA_ROUTE_LDP] = {ZEBRA_ROUTE_LDP, 150}, | |
94 | [ZEBRA_ROUTE_VNC] = {ZEBRA_ROUTE_VNC, 20}, | |
95 | [ZEBRA_ROUTE_VNC_DIRECT] = {ZEBRA_ROUTE_VNC_DIRECT, 20}, | |
96 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = {ZEBRA_ROUTE_VNC_DIRECT_RH, 20}, | |
97 | [ZEBRA_ROUTE_BGP_DIRECT] = {ZEBRA_ROUTE_BGP_DIRECT, 20}, | |
98 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = {ZEBRA_ROUTE_BGP_DIRECT_EXT, 20}, | |
99 | [ZEBRA_ROUTE_BABEL] = {ZEBRA_ROUTE_BABEL, 100}, | |
400a663b | 100 | [ZEBRA_ROUTE_SHARP] = {ZEBRA_ROUTE_SHARP, 150}, |
c710b277 | 101 | |
d62a17ae | 102 | /* no entry/default: 150 */ |
718e3744 | 103 | }; |
6b0655a2 | 104 | |
4623d897 DL |
105 | /* RPF lookup behaviour */ |
106 | static enum multicast_mode ipv4_multicast_mode = MCAST_NO_CONFIG; | |
107 | ||
6c4f4e6e | 108 | |
d62a17ae | 109 | static void __attribute__((format(printf, 5, 6))) |
110 | _rnode_zlog(const char *_func, vrf_id_t vrf_id, struct route_node *rn, | |
111 | int priority, const char *msgfmt, ...) | |
2263a412 | 112 | { |
d62a17ae | 113 | char buf[SRCDEST2STR_BUFFER + sizeof(" (MRIB)")]; |
114 | char msgbuf[512]; | |
115 | va_list ap; | |
2263a412 | 116 | |
d62a17ae | 117 | va_start(ap, msgfmt); |
118 | vsnprintf(msgbuf, sizeof(msgbuf), msgfmt, ap); | |
119 | va_end(ap); | |
2263a412 | 120 | |
d62a17ae | 121 | if (rn) { |
122 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
123 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
cb653491 | 124 | |
d62a17ae | 125 | if (info->safi == SAFI_MULTICAST) |
126 | strcat(buf, " (MRIB)"); | |
127 | } else { | |
128 | snprintf(buf, sizeof(buf), "{(route_node *) NULL}"); | |
129 | } | |
2263a412 | 130 | |
d62a17ae | 131 | zlog(priority, "%s: %d:%s: %s", _func, vrf_id, buf, msgbuf); |
2263a412 DL |
132 | } |
133 | ||
d62a17ae | 134 | #define rnode_debug(node, vrf_id, ...) \ |
2263a412 | 135 | _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__) |
d62a17ae | 136 | #define rnode_info(node, ...) \ |
2263a412 DL |
137 | _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__) |
138 | ||
fd289fc8 | 139 | uint8_t route_distance(int type) |
40c7bdb0 | 140 | { |
fd289fc8 | 141 | uint8_t distance; |
40c7bdb0 | 142 | |
d62a17ae | 143 | if ((unsigned)type >= array_size(route_info)) |
144 | distance = 150; | |
145 | else | |
146 | distance = route_info[type].distance; | |
40c7bdb0 | 147 | |
d62a17ae | 148 | return distance; |
40c7bdb0 | 149 | } |
150 | ||
d7c0a89a | 151 | int is_zebra_valid_kernel_table(uint32_t table_id) |
7a4bb9c5 | 152 | { |
8f500a1c | 153 | #ifdef linux |
d62a17ae | 154 | if ((table_id == RT_TABLE_UNSPEC) || (table_id == RT_TABLE_LOCAL) |
155 | || (table_id == RT_TABLE_COMPAT)) | |
156 | return 0; | |
8f500a1c RW |
157 | #endif |
158 | ||
d62a17ae | 159 | return 1; |
7a4bb9c5 DS |
160 | } |
161 | ||
d7c0a89a | 162 | int is_zebra_main_routing_table(uint32_t table_id) |
7a4bb9c5 | 163 | { |
d62a17ae | 164 | if ((table_id == RT_TABLE_MAIN) |
165 | || (table_id == zebrad.rtm_table_default)) | |
166 | return 1; | |
167 | return 0; | |
7a4bb9c5 DS |
168 | } |
169 | ||
86391e56 | 170 | int zebra_check_addr(const struct prefix *p) |
0aabccc0 | 171 | { |
d62a17ae | 172 | if (p->family == AF_INET) { |
d7c0a89a | 173 | uint32_t addr; |
0aabccc0 | 174 | |
d62a17ae | 175 | addr = p->u.prefix4.s_addr; |
176 | addr = ntohl(addr); | |
0aabccc0 | 177 | |
d62a17ae | 178 | if (IPV4_NET127(addr) || IN_CLASSD(addr) |
179 | || IPV4_LINKLOCAL(addr)) | |
180 | return 0; | |
181 | } | |
182 | if (p->family == AF_INET6) { | |
183 | if (IN6_IS_ADDR_LOOPBACK(&p->u.prefix6)) | |
184 | return 0; | |
185 | if (IN6_IS_ADDR_LINKLOCAL(&p->u.prefix6)) | |
186 | return 0; | |
187 | } | |
188 | return 1; | |
0aabccc0 DD |
189 | } |
190 | ||
fa713d9e | 191 | /* Add nexthop to the end of a rib node's nexthop list */ |
d62a17ae | 192 | void route_entry_nexthop_add(struct route_entry *re, struct nexthop *nexthop) |
fa713d9e | 193 | { |
7ee30f28 | 194 | nexthop_add(&re->ng.nexthop, nexthop); |
d62a17ae | 195 | re->nexthop_num++; |
718e3744 | 196 | } |
197 | ||
6e26278c | 198 | |
6e26278c DS |
199 | /** |
200 | * copy_nexthop - copy a nexthop to the rib structure. | |
201 | */ | |
d62a17ae | 202 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) |
6e26278c | 203 | { |
7ee30f28 DS |
204 | assert(!re->ng.nexthop); |
205 | copy_nexthops(&re->ng.nexthop, nh, NULL); | |
d62a17ae | 206 | for (struct nexthop *nexthop = nh; nexthop; nexthop = nexthop->next) |
207 | re->nexthop_num++; | |
6e26278c DS |
208 | } |
209 | ||
718e3744 | 210 | /* Delete specified nexthop from the list. */ |
d62a17ae | 211 | void route_entry_nexthop_delete(struct route_entry *re, struct nexthop *nexthop) |
718e3744 | 212 | { |
d62a17ae | 213 | if (nexthop->next) |
214 | nexthop->next->prev = nexthop->prev; | |
215 | if (nexthop->prev) | |
216 | nexthop->prev->next = nexthop->next; | |
217 | else | |
7ee30f28 | 218 | re->ng.nexthop = nexthop->next; |
d62a17ae | 219 | re->nexthop_num--; |
718e3744 | 220 | } |
221 | ||
fa713d9e | 222 | |
d62a17ae | 223 | struct nexthop *route_entry_nexthop_ifindex_add(struct route_entry *re, |
4a7371e9 DS |
224 | ifindex_t ifindex, |
225 | vrf_id_t nh_vrf_id) | |
718e3744 | 226 | { |
d62a17ae | 227 | struct nexthop *nexthop; |
718e3744 | 228 | |
d62a17ae | 229 | nexthop = nexthop_new(); |
230 | nexthop->type = NEXTHOP_TYPE_IFINDEX; | |
231 | nexthop->ifindex = ifindex; | |
4a7371e9 | 232 | nexthop->vrf_id = nh_vrf_id; |
718e3744 | 233 | |
d62a17ae | 234 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 235 | |
d62a17ae | 236 | return nexthop; |
718e3744 | 237 | } |
238 | ||
d62a17ae | 239 | struct nexthop *route_entry_nexthop_ipv4_add(struct route_entry *re, |
240 | struct in_addr *ipv4, | |
4a7371e9 DS |
241 | struct in_addr *src, |
242 | vrf_id_t nh_vrf_id) | |
718e3744 | 243 | { |
d62a17ae | 244 | struct nexthop *nexthop; |
718e3744 | 245 | |
d62a17ae | 246 | nexthop = nexthop_new(); |
247 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
4a7371e9 | 248 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 249 | nexthop->gate.ipv4 = *ipv4; |
250 | if (src) | |
251 | nexthop->src.ipv4 = *src; | |
718e3744 | 252 | |
d62a17ae | 253 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 254 | |
d62a17ae | 255 | return nexthop; |
718e3744 | 256 | } |
257 | ||
d62a17ae | 258 | struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re, |
259 | struct in_addr *ipv4, | |
260 | struct in_addr *src, | |
4a7371e9 DS |
261 | ifindex_t ifindex, |
262 | vrf_id_t nh_vrf_id) | |
718e3744 | 263 | { |
d62a17ae | 264 | struct nexthop *nexthop; |
265 | struct interface *ifp; | |
718e3744 | 266 | |
d62a17ae | 267 | nexthop = nexthop_new(); |
4a7371e9 | 268 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 269 | nexthop->type = NEXTHOP_TYPE_IPV4_IFINDEX; |
270 | nexthop->gate.ipv4 = *ipv4; | |
271 | if (src) | |
272 | nexthop->src.ipv4 = *src; | |
273 | nexthop->ifindex = ifindex; | |
4a7371e9 | 274 | ifp = if_lookup_by_index(nexthop->ifindex, nh_vrf_id); |
d62a17ae | 275 | /*Pending: need to think if null ifp here is ok during bootup? |
276 | There was a crash because ifp here was coming to be NULL */ | |
277 | if (ifp) | |
996c9314 | 278 | if (connected_is_unnumbered(ifp) |
103e4a71 CF |
279 | || CHECK_FLAG(re->flags, ZEBRA_FLAG_EVPN_ROUTE) |
280 | || CHECK_FLAG(re->flags, ZEBRA_FLAG_ONLINK)) { | |
d62a17ae | 281 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); |
282 | } | |
718e3744 | 283 | |
d62a17ae | 284 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 285 | |
d62a17ae | 286 | return nexthop; |
718e3744 | 287 | } |
288 | ||
d62a17ae | 289 | struct nexthop *route_entry_nexthop_ipv6_add(struct route_entry *re, |
4a7371e9 DS |
290 | struct in6_addr *ipv6, |
291 | vrf_id_t nh_vrf_id) | |
718e3744 | 292 | { |
d62a17ae | 293 | struct nexthop *nexthop; |
718e3744 | 294 | |
d62a17ae | 295 | nexthop = nexthop_new(); |
4a7371e9 | 296 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 297 | nexthop->type = NEXTHOP_TYPE_IPV6; |
298 | nexthop->gate.ipv6 = *ipv6; | |
718e3744 | 299 | |
d62a17ae | 300 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 301 | |
d62a17ae | 302 | return nexthop; |
718e3744 | 303 | } |
304 | ||
d62a17ae | 305 | struct nexthop *route_entry_nexthop_ipv6_ifindex_add(struct route_entry *re, |
306 | struct in6_addr *ipv6, | |
4a7371e9 DS |
307 | ifindex_t ifindex, |
308 | vrf_id_t nh_vrf_id) | |
718e3744 | 309 | { |
d62a17ae | 310 | struct nexthop *nexthop; |
718e3744 | 311 | |
d62a17ae | 312 | nexthop = nexthop_new(); |
4a7371e9 | 313 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 314 | nexthop->type = NEXTHOP_TYPE_IPV6_IFINDEX; |
315 | nexthop->gate.ipv6 = *ipv6; | |
316 | nexthop->ifindex = ifindex; | |
103e4a71 CF |
317 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_EVPN_ROUTE) |
318 | || CHECK_FLAG(re->flags, ZEBRA_FLAG_ONLINK)) { | |
1ec31309 | 319 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); |
103e4a71 | 320 | } |
718e3744 | 321 | |
d62a17ae | 322 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 323 | |
d62a17ae | 324 | return nexthop; |
718e3744 | 325 | } |
718e3744 | 326 | |
a8309422 | 327 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re, |
60466a63 | 328 | enum blackhole_type bh_type) |
595db7f1 | 329 | { |
d62a17ae | 330 | struct nexthop *nexthop; |
595db7f1 | 331 | |
d62a17ae | 332 | nexthop = nexthop_new(); |
4a7371e9 | 333 | nexthop->vrf_id = VRF_DEFAULT; |
d62a17ae | 334 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; |
a8309422 | 335 | nexthop->bh_type = bh_type; |
595db7f1 | 336 | |
d62a17ae | 337 | route_entry_nexthop_add(re, nexthop); |
595db7f1 | 338 | |
d62a17ae | 339 | return nexthop; |
595db7f1 | 340 | } |
341 | ||
86391e56 | 342 | static void nexthop_set_resolved(afi_t afi, const struct nexthop *newhop, |
d62a17ae | 343 | struct nexthop *nexthop) |
4491a88f | 344 | { |
d62a17ae | 345 | struct nexthop *resolved_hop; |
4491a88f | 346 | |
d62a17ae | 347 | resolved_hop = nexthop_new(); |
348 | SET_FLAG(resolved_hop->flags, NEXTHOP_FLAG_ACTIVE); | |
d855d11f | 349 | |
4a7371e9 | 350 | resolved_hop->vrf_id = nexthop->vrf_id; |
d855d11f RW |
351 | switch (newhop->type) { |
352 | case NEXTHOP_TYPE_IPV4: | |
353 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
354 | /* If the resolving route specifies a gateway, use it */ | |
d62a17ae | 355 | resolved_hop->type = newhop->type; |
356 | resolved_hop->gate.ipv4 = newhop->gate.ipv4; | |
4491a88f | 357 | |
d62a17ae | 358 | if (newhop->ifindex) { |
359 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
360 | resolved_hop->ifindex = newhop->ifindex; | |
d62a17ae | 361 | } |
d855d11f RW |
362 | break; |
363 | case NEXTHOP_TYPE_IPV6: | |
364 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
d62a17ae | 365 | resolved_hop->type = newhop->type; |
366 | resolved_hop->gate.ipv6 = newhop->gate.ipv6; | |
367 | ||
368 | if (newhop->ifindex) { | |
369 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
370 | resolved_hop->ifindex = newhop->ifindex; | |
371 | } | |
d855d11f RW |
372 | break; |
373 | case NEXTHOP_TYPE_IFINDEX: | |
374 | /* If the resolving route is an interface route, | |
375 | * it means the gateway we are looking up is connected | |
376 | * to that interface. (The actual network is _not_ onlink). | |
377 | * Therefore, the resolved route should have the original | |
378 | * gateway as nexthop as it is directly connected. | |
379 | * | |
380 | * On Linux, we have to set the onlink netlink flag because | |
381 | * otherwise, the kernel won't accept the route. | |
382 | */ | |
d62a17ae | 383 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; |
384 | if (afi == AFI_IP) { | |
385 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
386 | resolved_hop->gate.ipv4 = nexthop->gate.ipv4; | |
387 | } else if (afi == AFI_IP6) { | |
388 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
389 | resolved_hop->gate.ipv6 = nexthop->gate.ipv6; | |
390 | } | |
391 | resolved_hop->ifindex = newhop->ifindex; | |
d855d11f RW |
392 | break; |
393 | case NEXTHOP_TYPE_BLACKHOLE: | |
59693377 DS |
394 | resolved_hop->type = NEXTHOP_TYPE_BLACKHOLE; |
395 | resolved_hop->bh_type = nexthop->bh_type; | |
d855d11f | 396 | break; |
59693377 | 397 | } |
d855d11f | 398 | |
effcfaeb DS |
399 | if (newhop->flags & NEXTHOP_FLAG_ONLINK) |
400 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
401 | ||
f674dfe2 RW |
402 | /* Copy labels of the resolved route */ |
403 | if (newhop->nh_label) | |
404 | nexthop_add_labels(resolved_hop, newhop->nh_label_type, | |
405 | newhop->nh_label->num_labels, | |
406 | &newhop->nh_label->label[0]); | |
407 | ||
d62a17ae | 408 | resolved_hop->rparent = nexthop; |
409 | nexthop_add(&nexthop->resolved, resolved_hop); | |
4491a88f DS |
410 | } |
411 | ||
718e3744 | 412 | /* If force flag is not set, do not modify falgs at all for uninstall |
413 | the route from FIB. */ | |
d62a17ae | 414 | static int nexthop_active(afi_t afi, struct route_entry *re, |
0dddbf72 | 415 | struct nexthop *nexthop, bool set, |
d62a17ae | 416 | struct route_node *top) |
417 | { | |
418 | struct prefix p; | |
419 | struct route_table *table; | |
420 | struct route_node *rn; | |
5f7a4718 | 421 | struct route_entry *match = NULL; |
d62a17ae | 422 | int resolved; |
423 | struct nexthop *newhop; | |
424 | struct interface *ifp; | |
5f7a4718 | 425 | rib_dest_t *dest; |
d62a17ae | 426 | |
427 | if ((nexthop->type == NEXTHOP_TYPE_IPV4) | |
428 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
429 | nexthop->ifindex = 0; | |
430 | ||
431 | if (set) { | |
432 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE); | |
d62a17ae | 433 | nexthops_free(nexthop->resolved); |
434 | nexthop->resolved = NULL; | |
435 | re->nexthop_mtu = 0; | |
d44ca835 | 436 | } |
18ff3edd | 437 | |
1ec31309 | 438 | /* Next hops (remote VTEPs) for EVPN routes are fully resolved. */ |
439 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_EVPN_RVTEP)) | |
440 | return 1; | |
441 | ||
d62a17ae | 442 | /* Skip nexthops that have been filtered out due to route-map */ |
443 | /* The nexthops are specific to this route and so the same */ | |
444 | /* nexthop for a different route may not have this flag set */ | |
34815ea3 DS |
445 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FILTERED)) { |
446 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
447 | zlog_debug("\t%s: Nexthop Filtered", | |
448 | __PRETTY_FUNCTION__); | |
d62a17ae | 449 | return 0; |
34815ea3 | 450 | } |
d62a17ae | 451 | |
452 | /* | |
453 | * Check to see if we should trust the passed in information | |
454 | * for UNNUMBERED interfaces as that we won't find the GW | |
455 | * address in the routing table. | |
456 | */ | |
457 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK)) { | |
4a7371e9 | 458 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); |
103e4a71 CF |
459 | if ((ifp && connected_is_unnumbered(ifp)) |
460 | || CHECK_FLAG(re->flags, ZEBRA_FLAG_ONLINK)) { | |
d62a17ae | 461 | if (if_is_operative(ifp)) |
462 | return 1; | |
34815ea3 DS |
463 | else { |
464 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
465 | zlog_debug( | |
466 | "\t%s: Onlink and interface %s is not operative", | |
467 | __PRETTY_FUNCTION__, ifp->name); | |
d62a17ae | 468 | return 0; |
34815ea3 DS |
469 | } |
470 | } else { | |
471 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
472 | zlog_debug( | |
473 | "\t%s: Interface %s is not unnumbered", | |
474 | __PRETTY_FUNCTION__, | |
475 | ifp ? ifp->name : "Unknown"); | |
d62a17ae | 476 | return 0; |
34815ea3 | 477 | } |
16814f96 | 478 | } |
718e3744 | 479 | |
d62a17ae | 480 | /* Make lookup prefix. */ |
481 | memset(&p, 0, sizeof(struct prefix)); | |
482 | switch (afi) { | |
483 | case AFI_IP: | |
484 | p.family = AF_INET; | |
485 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
486 | p.u.prefix4 = nexthop->gate.ipv4; | |
487 | break; | |
488 | case AFI_IP6: | |
489 | p.family = AF_INET6; | |
490 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
491 | p.u.prefix6 = nexthop->gate.ipv6; | |
492 | break; | |
493 | default: | |
494 | assert(afi != AFI_IP && afi != AFI_IP6); | |
495 | break; | |
718e3744 | 496 | } |
d62a17ae | 497 | /* Lookup table. */ |
4a7371e9 | 498 | table = zebra_vrf_table(afi, SAFI_UNICAST, nexthop->vrf_id); |
34815ea3 DS |
499 | if (!table) { |
500 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
501 | zlog_debug("\t%s: Table not found", | |
502 | __PRETTY_FUNCTION__); | |
d62a17ae | 503 | return 0; |
34815ea3 | 504 | } |
d62a17ae | 505 | |
506 | rn = route_node_match(table, (struct prefix *)&p); | |
507 | while (rn) { | |
508 | route_unlock_node(rn); | |
509 | ||
fd7fd9e5 DS |
510 | /* Lookup should halt if we've matched against ourselves ('top', |
511 | * if specified) - i.e., we cannot have a nexthop NH1 is | |
512 | * resolved by a route NH1. The exception is if the route is a | |
513 | * host route. | |
514 | */ | |
515 | if (top && rn == top) | |
996c9314 | 516 | if (((afi == AFI_IP) && (rn->p.prefixlen != 32)) |
34815ea3 DS |
517 | || ((afi == AFI_IP6) && (rn->p.prefixlen != 128))) { |
518 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
519 | zlog_debug( | |
520 | "\t%s: Matched against ourself and prefix length is not max bit length", | |
521 | __PRETTY_FUNCTION__); | |
fd7fd9e5 | 522 | return 0; |
34815ea3 | 523 | } |
d62a17ae | 524 | |
525 | /* Pick up selected route. */ | |
526 | /* However, do not resolve over default route unless explicitly | |
527 | * allowed. */ | |
528 | if (is_default_prefix(&rn->p) | |
34815ea3 DS |
529 | && !rnh_resolve_via_default(p.family)) { |
530 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
531 | zlog_debug( | |
532 | "\t:%s: Resolved against default route", | |
533 | __PRETTY_FUNCTION__); | |
d62a17ae | 534 | return 0; |
34815ea3 | 535 | } |
d62a17ae | 536 | |
5f7a4718 | 537 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
538 | if (dest && dest->selected_fib |
539 | && !CHECK_FLAG(dest->selected_fib->status, | |
540 | ROUTE_ENTRY_REMOVED) | |
541 | && dest->selected_fib->type != ZEBRA_ROUTE_TABLE) | |
5f7a4718 | 542 | match = dest->selected_fib; |
48a53dc7 | 543 | |
d62a17ae | 544 | /* If there is no selected route or matched route is EGP, go up |
545 | tree. */ | |
546 | if (!match) { | |
547 | do { | |
548 | rn = rn->parent; | |
549 | } while (rn && rn->info == NULL); | |
550 | if (rn) | |
551 | route_lock_node(rn); | |
c87bdd2b | 552 | |
d62a17ae | 553 | continue; |
554 | } | |
555 | ||
d62a17ae | 556 | if (match->type == ZEBRA_ROUTE_CONNECT) { |
557 | /* Directly point connected route. */ | |
7ee30f28 | 558 | newhop = match->ng.nexthop; |
d62a17ae | 559 | if (newhop) { |
560 | if (nexthop->type == NEXTHOP_TYPE_IPV4 | |
561 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
562 | nexthop->ifindex = newhop->ifindex; | |
563 | } | |
564 | return 1; | |
4e8b02f4 | 565 | } else if (CHECK_FLAG(re->flags, ZEBRA_FLAG_ALLOW_RECURSION)) { |
d62a17ae | 566 | resolved = 0; |
7ee30f28 | 567 | for (ALL_NEXTHOPS(match->ng, newhop)) { |
a8309422 DL |
568 | if (!CHECK_FLAG(newhop->flags, |
569 | NEXTHOP_FLAG_FIB)) | |
570 | continue; | |
571 | if (CHECK_FLAG(newhop->flags, | |
572 | NEXTHOP_FLAG_RECURSIVE)) | |
573 | continue; | |
574 | ||
575 | if (set) { | |
576 | SET_FLAG(nexthop->flags, | |
577 | NEXTHOP_FLAG_RECURSIVE); | |
578 | SET_FLAG(re->status, | |
579 | ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
580 | nexthop_set_resolved(afi, newhop, | |
581 | nexthop); | |
d62a17ae | 582 | } |
a8309422 DL |
583 | resolved = 1; |
584 | } | |
43e31305 JB |
585 | if (resolved && set) |
586 | re->nexthop_mtu = match->mtu; | |
34815ea3 DS |
587 | if (!resolved && IS_ZEBRA_DEBUG_RIB_DETAILED) |
588 | zlog_debug("\t%s: Recursion failed to find", | |
589 | __PRETTY_FUNCTION__); | |
d62a17ae | 590 | return resolved; |
591 | } else if (re->type == ZEBRA_ROUTE_STATIC) { | |
592 | resolved = 0; | |
7ee30f28 | 593 | for (ALL_NEXTHOPS(match->ng, newhop)) { |
a8309422 DL |
594 | if (!CHECK_FLAG(newhop->flags, |
595 | NEXTHOP_FLAG_FIB)) | |
596 | continue; | |
597 | ||
598 | if (set) { | |
599 | SET_FLAG(nexthop->flags, | |
600 | NEXTHOP_FLAG_RECURSIVE); | |
601 | nexthop_set_resolved(afi, newhop, | |
602 | nexthop); | |
d62a17ae | 603 | } |
a8309422 DL |
604 | resolved = 1; |
605 | } | |
d62a17ae | 606 | if (resolved && set) |
607 | re->nexthop_mtu = match->mtu; | |
34815ea3 DS |
608 | |
609 | if (!resolved && IS_ZEBRA_DEBUG_RIB_DETAILED) | |
610 | zlog_debug( | |
611 | "\t%s: Static route unable to resolve", | |
612 | __PRETTY_FUNCTION__); | |
d62a17ae | 613 | return resolved; |
614 | } else { | |
7939ff76 DS |
615 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
616 | zlog_debug("\t%s: Route Type %s has not turned on recursion", | |
617 | __PRETTY_FUNCTION__, | |
618 | zebra_route_string(re->type)); | |
619 | if (re->type == ZEBRA_ROUTE_BGP && | |
620 | !CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
621 | zlog_debug("\tEBGP: see \"disable-ebgp-connected-route-check\" or \"disable-connected-check\""); | |
622 | } | |
d62a17ae | 623 | return 0; |
624 | } | |
718e3744 | 625 | } |
7939ff76 DS |
626 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
627 | zlog_debug("\t%s: Nexthop did not lookup in table", | |
628 | __PRETTY_FUNCTION__); | |
d62a17ae | 629 | return 0; |
718e3744 | 630 | } |
718e3744 | 631 | |
d62a17ae | 632 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
633 | union g_addr *addr, struct route_node **rn_out) | |
634 | { | |
635 | struct prefix p; | |
636 | struct route_table *table; | |
637 | struct route_node *rn; | |
5f7a4718 | 638 | struct route_entry *match = NULL; |
d62a17ae | 639 | struct nexthop *newhop; |
640 | ||
641 | /* Lookup table. */ | |
642 | table = zebra_vrf_table(afi, safi, vrf_id); | |
643 | if (!table) | |
644 | return 0; | |
645 | ||
646 | memset(&p, 0, sizeof(struct prefix)); | |
647 | p.family = afi; | |
648 | if (afi == AFI_IP) { | |
649 | p.u.prefix4 = addr->ipv4; | |
650 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
651 | } else { | |
652 | p.u.prefix6 = addr->ipv6; | |
653 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
654 | } | |
718e3744 | 655 | |
d62a17ae | 656 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 657 | |
d62a17ae | 658 | while (rn) { |
5f7a4718 DS |
659 | rib_dest_t *dest; |
660 | ||
d62a17ae | 661 | route_unlock_node(rn); |
718e3744 | 662 | |
5f7a4718 | 663 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
664 | if (dest && dest->selected_fib |
665 | && !CHECK_FLAG(dest->selected_fib->status, | |
666 | ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 667 | match = dest->selected_fib; |
718e3744 | 668 | |
d62a17ae | 669 | /* If there is no selected route or matched route is EGP, go up |
670 | tree. */ | |
671 | if (!match) { | |
672 | do { | |
673 | rn = rn->parent; | |
674 | } while (rn && rn->info == NULL); | |
675 | if (rn) | |
676 | route_lock_node(rn); | |
677 | } else { | |
678 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
679 | int found = 0; | |
7ee30f28 | 680 | for (ALL_NEXTHOPS(match->ng, newhop)) |
d62a17ae | 681 | if (CHECK_FLAG(newhop->flags, |
682 | NEXTHOP_FLAG_FIB)) { | |
683 | found = 1; | |
684 | break; | |
685 | } | |
686 | if (!found) | |
687 | return NULL; | |
688 | } | |
689 | ||
690 | if (rn_out) | |
691 | *rn_out = rn; | |
692 | return match; | |
693 | } | |
16814f96 | 694 | } |
d62a17ae | 695 | return NULL; |
696 | } | |
697 | ||
698 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
699 | struct in_addr addr, | |
700 | struct route_node **rn_out) | |
701 | { | |
702 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
703 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
704 | union g_addr gaddr = {.ipv4 = addr}; | |
705 | ||
706 | switch (ipv4_multicast_mode) { | |
707 | case MCAST_MRIB_ONLY: | |
708 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
709 | rn_out); | |
710 | case MCAST_URIB_ONLY: | |
711 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
712 | case MCAST_NO_CONFIG: | |
713 | case MCAST_MIX_MRIB_FIRST: | |
714 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
715 | &m_rn); | |
716 | if (!mre) | |
717 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
718 | &gaddr, &u_rn); | |
719 | break; | |
720 | case MCAST_MIX_DISTANCE: | |
721 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
722 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
723 | if (mre && ure) | |
724 | re = ure->distance < mre->distance ? ure : mre; | |
725 | else if (mre) | |
726 | re = mre; | |
727 | else if (ure) | |
728 | re = ure; | |
729 | break; | |
730 | case MCAST_MIX_PFXLEN: | |
731 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
732 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
733 | if (mre && ure) | |
734 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
735 | else if (mre) | |
736 | re = mre; | |
737 | else if (ure) | |
738 | re = ure; | |
739 | break; | |
718e3744 | 740 | } |
33550aa8 | 741 | |
d62a17ae | 742 | if (rn_out) |
743 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 744 | |
d62a17ae | 745 | if (IS_ZEBRA_DEBUG_RIB) { |
746 | char buf[BUFSIZ]; | |
747 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 748 | |
32391aff DS |
749 | zlog_debug("%s: %s: vrf: %u found %s, using %s", |
750 | __func__, buf, vrf_id, | |
d62a17ae | 751 | mre ? (ure ? "MRIB+URIB" : "MRIB") |
752 | : ure ? "URIB" : "nothing", | |
753 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
754 | } | |
755 | return re; | |
4623d897 DL |
756 | } |
757 | ||
d62a17ae | 758 | void multicast_mode_ipv4_set(enum multicast_mode mode) |
4623d897 | 759 | { |
d62a17ae | 760 | if (IS_ZEBRA_DEBUG_RIB) |
761 | zlog_debug("%s: multicast lookup mode set (%d)", __func__, | |
762 | mode); | |
763 | ipv4_multicast_mode = mode; | |
4623d897 DL |
764 | } |
765 | ||
d62a17ae | 766 | enum multicast_mode multicast_mode_ipv4_get(void) |
4623d897 | 767 | { |
d62a17ae | 768 | return ipv4_multicast_mode; |
4623d897 DL |
769 | } |
770 | ||
d62a17ae | 771 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 772 | { |
d62a17ae | 773 | struct route_table *table; |
774 | struct route_node *rn; | |
5f7a4718 | 775 | struct route_entry *match = NULL; |
d62a17ae | 776 | struct nexthop *nexthop; |
5f7a4718 | 777 | rib_dest_t *dest; |
718e3744 | 778 | |
d62a17ae | 779 | /* Lookup table. */ |
780 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
781 | if (!table) | |
782 | return 0; | |
718e3744 | 783 | |
d62a17ae | 784 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 785 | |
d62a17ae | 786 | /* No route for this prefix. */ |
787 | if (!rn) | |
788 | return NULL; | |
718e3744 | 789 | |
d62a17ae | 790 | /* Unlock node. */ |
791 | route_unlock_node(rn); | |
5f7a4718 | 792 | dest = rib_dest_from_rnode(rn); |
718e3744 | 793 | |
996c9314 LB |
794 | if (dest && dest->selected_fib |
795 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 796 | match = dest->selected_fib; |
718e3744 | 797 | |
d62a17ae | 798 | if (!match) |
799 | return NULL; | |
718e3744 | 800 | |
d62a17ae | 801 | if (match->type == ZEBRA_ROUTE_CONNECT) |
802 | return match; | |
f9e1b38e | 803 | |
7ee30f28 | 804 | for (ALL_NEXTHOPS(match->ng, nexthop)) |
d62a17ae | 805 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) |
806 | return match; | |
718e3744 | 807 | |
d62a17ae | 808 | return NULL; |
718e3744 | 809 | } |
810 | ||
dc95824a DO |
811 | /* |
812 | * This clone function, unlike its original rib_lookup_ipv4(), checks | |
813 | * if specified IPv4 route record (prefix/mask -> gate) exists in | |
f0f77c9a | 814 | * the whole RIB and has ROUTE_ENTRY_SELECTED_FIB set. |
dc95824a DO |
815 | * |
816 | * Return values: | |
817 | * -1: error | |
818 | * 0: exact match found | |
819 | * 1: a match was found with a different gate | |
820 | * 2: connected route found | |
821 | * 3: no matches found | |
822 | */ | |
d62a17ae | 823 | int rib_lookup_ipv4_route(struct prefix_ipv4 *p, union sockunion *qgate, |
824 | vrf_id_t vrf_id) | |
dc95824a | 825 | { |
d62a17ae | 826 | struct route_table *table; |
827 | struct route_node *rn; | |
5f7a4718 | 828 | struct route_entry *match = NULL; |
d62a17ae | 829 | struct nexthop *nexthop; |
830 | int nexthops_active; | |
5f7a4718 | 831 | rib_dest_t *dest; |
dc95824a | 832 | |
d62a17ae | 833 | /* Lookup table. */ |
834 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
835 | if (!table) | |
836 | return ZEBRA_RIB_LOOKUP_ERROR; | |
dc95824a | 837 | |
d62a17ae | 838 | /* Scan the RIB table for exactly matching RIB entry. */ |
839 | rn = route_node_lookup(table, (struct prefix *)p); | |
dc95824a | 840 | |
d62a17ae | 841 | /* No route for this prefix. */ |
842 | if (!rn) | |
843 | return ZEBRA_RIB_NOTFOUND; | |
dc95824a | 844 | |
d62a17ae | 845 | /* Unlock node. */ |
846 | route_unlock_node(rn); | |
5f7a4718 | 847 | dest = rib_dest_from_rnode(rn); |
dc95824a | 848 | |
d62a17ae | 849 | /* Find out if a "selected" RR for the discovered RIB entry exists ever. |
850 | */ | |
996c9314 LB |
851 | if (dest && dest->selected_fib |
852 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 853 | match = dest->selected_fib; |
dc95824a | 854 | |
d62a17ae | 855 | /* None such found :( */ |
856 | if (!match) | |
857 | return ZEBRA_RIB_NOTFOUND; | |
858 | ||
859 | if (match->type == ZEBRA_ROUTE_CONNECT) | |
860 | return ZEBRA_RIB_FOUND_CONNECTED; | |
861 | ||
862 | /* Ok, we have a cood candidate, let's check it's nexthop list... */ | |
863 | nexthops_active = 0; | |
7ee30f28 | 864 | for (ALL_NEXTHOPS(match->ng, nexthop)) |
d62a17ae | 865 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)) { |
866 | nexthops_active = 1; | |
867 | if (nexthop->gate.ipv4.s_addr == sockunion2ip(qgate)) | |
868 | return ZEBRA_RIB_FOUND_EXACT; | |
869 | if (IS_ZEBRA_DEBUG_RIB) { | |
870 | char gate_buf[INET_ADDRSTRLEN], | |
871 | qgate_buf[INET_ADDRSTRLEN]; | |
872 | inet_ntop(AF_INET, &nexthop->gate.ipv4.s_addr, | |
873 | gate_buf, INET_ADDRSTRLEN); | |
874 | inet_ntop(AF_INET, &sockunion2ip(qgate), | |
875 | qgate_buf, INET_ADDRSTRLEN); | |
876 | zlog_debug("%s: qgate == %s, %s == %s", | |
877 | __func__, qgate_buf, | |
878 | nexthop->rparent ? "rgate" : "gate", | |
879 | gate_buf); | |
880 | } | |
881 | } | |
882 | ||
883 | if (nexthops_active) | |
884 | return ZEBRA_RIB_FOUND_NOGATE; | |
885 | ||
886 | return ZEBRA_RIB_NOTFOUND; | |
887 | } | |
888 | ||
889 | #define RIB_SYSTEM_ROUTE(R) \ | |
890 | ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT) | |
7514fb77 | 891 | |
212df1de PG |
892 | #define RIB_KERNEL_ROUTE(R) \ |
893 | ((R)->type == ZEBRA_ROUTE_KERNEL) | |
894 | ||
dc95824a DO |
895 | /* This function verifies reachability of one given nexthop, which can be |
896 | * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored | |
897 | * in nexthop->flags field. If the 4th parameter, 'set', is non-zero, | |
898 | * nexthop->ifindex will be updated appropriately as well. | |
899 | * An existing route map can turn (otherwise active) nexthop into inactive, but | |
900 | * not vice versa. | |
901 | * | |
902 | * The return value is the final value of 'ACTIVE' flag. | |
903 | */ | |
904 | ||
d62a17ae | 905 | static unsigned nexthop_active_check(struct route_node *rn, |
906 | struct route_entry *re, | |
0dddbf72 | 907 | struct nexthop *nexthop, bool set) |
d62a17ae | 908 | { |
909 | struct interface *ifp; | |
910 | route_map_result_t ret = RMAP_MATCH; | |
911 | int family; | |
912 | char buf[SRCDEST2STR_BUFFER]; | |
86391e56 | 913 | const struct prefix *p, *src_p; |
ac6eebce | 914 | struct zebra_vrf *zvrf; |
915 | ||
d62a17ae | 916 | srcdest_rnode_prefixes(rn, &p, &src_p); |
917 | ||
918 | if (rn->p.family == AF_INET) | |
919 | family = AFI_IP; | |
920 | else if (rn->p.family == AF_INET6) | |
921 | family = AFI_IP6; | |
922 | else | |
923 | family = 0; | |
924 | switch (nexthop->type) { | |
925 | case NEXTHOP_TYPE_IFINDEX: | |
4a7371e9 | 926 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); |
d62a17ae | 927 | if (ifp && if_is_operative(ifp)) |
928 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
929 | else | |
930 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
931 | break; | |
932 | case NEXTHOP_TYPE_IPV4: | |
933 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
934 | family = AFI_IP; | |
1ec31309 | 935 | if (nexthop_active(AFI_IP, re, nexthop, set, rn)) |
d62a17ae | 936 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
937 | else | |
938 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
939 | break; | |
940 | case NEXTHOP_TYPE_IPV6: | |
941 | family = AFI_IP6; | |
942 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
943 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
944 | else | |
945 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
946 | break; | |
947 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
948 | /* RFC 5549, v4 prefix with v6 NH */ | |
949 | if (rn->p.family != AF_INET) | |
950 | family = AFI_IP6; | |
951 | if (IN6_IS_ADDR_LINKLOCAL(&nexthop->gate.ipv6)) { | |
007dbee6 | 952 | ifp = if_lookup_by_index(nexthop->ifindex, |
4a7371e9 | 953 | nexthop->vrf_id); |
d62a17ae | 954 | if (ifp && if_is_operative(ifp)) |
955 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
956 | else | |
957 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
958 | } else { | |
959 | if (nexthop_active(AFI_IP6, re, nexthop, set, rn)) | |
960 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
961 | else | |
962 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
963 | } | |
964 | break; | |
965 | case NEXTHOP_TYPE_BLACKHOLE: | |
966 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
967 | break; | |
968 | default: | |
969 | break; | |
970 | } | |
34815ea3 DS |
971 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) { |
972 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
973 | zlog_debug("\t%s: Unable to find a active nexthop", | |
974 | __PRETTY_FUNCTION__); | |
d62a17ae | 975 | return 0; |
34815ea3 | 976 | } |
d62a17ae | 977 | |
978 | /* XXX: What exactly do those checks do? Do we support | |
ac6eebce | 979 | * e.g. IPv4 routes with IPv6 nexthops or vice versa? |
980 | */ | |
d62a17ae | 981 | if (RIB_SYSTEM_ROUTE(re) || (family == AFI_IP && p->family != AF_INET) |
982 | || (family == AFI_IP6 && p->family != AF_INET6)) | |
983 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
984 | ||
985 | /* The original code didn't determine the family correctly | |
986 | * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi | |
987 | * from the rib_table_info in those cases. | |
988 | * Possibly it may be better to use only the rib_table_info | |
989 | * in every case. | |
990 | */ | |
991 | if (!family) { | |
992 | rib_table_info_t *info; | |
993 | ||
994 | info = srcdest_rnode_table_info(rn); | |
995 | family = info->afi; | |
718e3744 | 996 | } |
c52ef59f | 997 | |
d62a17ae | 998 | memset(&nexthop->rmap_src.ipv6, 0, sizeof(union g_addr)); |
999 | ||
ac6eebce | 1000 | zvrf = zebra_vrf_lookup_by_id(nexthop->vrf_id); |
1001 | if (!zvrf) { | |
1002 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
1003 | zlog_debug("\t%s: zvrf is NULL", __PRETTY_FUNCTION__); | |
1004 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
1005 | } | |
1006 | ||
d62a17ae | 1007 | /* It'll get set if required inside */ |
ac6eebce | 1008 | ret = zebra_route_map_check(family, re->type, re->instance, p, |
1009 | nexthop, zvrf, re->tag); | |
d62a17ae | 1010 | if (ret == RMAP_DENYMATCH) { |
1011 | if (IS_ZEBRA_DEBUG_RIB) { | |
1012 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1013 | zlog_debug( | |
1014 | "%u:%s: Filtering out with NH out %s due to route map", | |
1015 | re->vrf_id, buf, | |
99b9d960 | 1016 | ifindex2ifname(nexthop->ifindex, |
4a7371e9 | 1017 | nexthop->vrf_id)); |
d62a17ae | 1018 | } |
1019 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
518f0eb1 | 1020 | } |
d62a17ae | 1021 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
718e3744 | 1022 | } |
1023 | ||
03e232a4 | 1024 | /* Iterate over all nexthops of the given RIB entry and refresh their |
f0f77c9a DS |
1025 | * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any |
1026 | * nexthop is found to toggle the ACTIVE flag, the whole re structure | |
1027 | * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is | |
03e232a4 DO |
1028 | * transparently passed to nexthop_active_check(). |
1029 | * | |
1030 | * Return value is the new number of active nexthops. | |
1031 | */ | |
1032 | ||
d62a17ae | 1033 | static int nexthop_active_update(struct route_node *rn, struct route_entry *re, |
0dddbf72 | 1034 | bool set) |
d62a17ae | 1035 | { |
1036 | struct nexthop *nexthop; | |
1037 | union g_addr prev_src; | |
1038 | unsigned int prev_active, new_active, old_num_nh; | |
1039 | ifindex_t prev_index; | |
ac6eebce | 1040 | |
d62a17ae | 1041 | old_num_nh = re->nexthop_active_num; |
1042 | ||
1043 | re->nexthop_active_num = 0; | |
1044 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1045 | ||
7ee30f28 | 1046 | for (nexthop = re->ng.nexthop; nexthop; nexthop = nexthop->next) { |
d62a17ae | 1047 | /* No protocol daemon provides src and so we're skipping |
1048 | * tracking it */ | |
1049 | prev_src = nexthop->rmap_src; | |
1050 | prev_active = CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
1051 | prev_index = nexthop->ifindex; | |
1052 | if ((new_active = nexthop_active_check(rn, re, nexthop, set))) | |
1053 | re->nexthop_active_num++; | |
1054 | /* Don't allow src setting on IPv6 addr for now */ | |
1055 | if (prev_active != new_active || prev_index != nexthop->ifindex | |
1056 | || ((nexthop->type >= NEXTHOP_TYPE_IFINDEX | |
1057 | && nexthop->type < NEXTHOP_TYPE_IPV6) | |
1058 | && prev_src.ipv4.s_addr | |
1059 | != nexthop->rmap_src.ipv4.s_addr) | |
1060 | || ((nexthop->type >= NEXTHOP_TYPE_IPV6 | |
1061 | && nexthop->type < NEXTHOP_TYPE_BLACKHOLE) | |
1062 | && !(IPV6_ADDR_SAME(&prev_src.ipv6, | |
1063 | &nexthop->rmap_src.ipv6)))) { | |
1064 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1065 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1066 | } | |
1067 | } | |
1068 | ||
1069 | if (old_num_nh != re->nexthop_active_num) | |
1070 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
6e26278c | 1071 | |
d62a17ae | 1072 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED)) { |
1073 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1074 | } | |
1075 | ||
1076 | return re->nexthop_active_num; | |
718e3744 | 1077 | } |
6baeb988 | 1078 | |
a64448ba DS |
1079 | /* |
1080 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
1081 | * (nexthops) must have a label. | |
1082 | */ | |
d62a17ae | 1083 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 1084 | { |
d62a17ae | 1085 | struct nexthop *nexthop = NULL; |
a64448ba | 1086 | |
d62a17ae | 1087 | if (re->type != ZEBRA_ROUTE_BGP) |
1088 | return 0; | |
a64448ba | 1089 | |
7ee30f28 | 1090 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d62a17ae | 1091 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) |
1092 | return 0; | |
6b0655a2 | 1093 | |
d62a17ae | 1094 | return 1; |
a64448ba | 1095 | } |
718e3744 | 1096 | |
6ae24471 DS |
1097 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
1098 | * is only used for IPv4. | |
1099 | */ | |
0c555cc6 DS |
1100 | void rib_install_kernel(struct route_node *rn, struct route_entry *re, |
1101 | struct route_entry *old) | |
718e3744 | 1102 | { |
d62a17ae | 1103 | struct nexthop *nexthop; |
1104 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 1105 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
97f5b441 MS |
1106 | const struct prefix *p, *src_p; |
1107 | enum zebra_dplane_result ret; | |
1108 | ||
1109 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
718e3744 | 1110 | |
d62a17ae | 1111 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 1112 | |
d62a17ae | 1113 | if (info->safi != SAFI_UNICAST) { |
7ee30f28 | 1114 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d62a17ae | 1115 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 1116 | return; |
25b9cb0c DL |
1117 | } else { |
1118 | struct nexthop *prev; | |
1119 | ||
7ee30f28 | 1120 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
25b9cb0c | 1121 | UNSET_FLAG (nexthop->flags, NEXTHOP_FLAG_DUPLICATE); |
7ee30f28 | 1122 | for (ALL_NEXTHOPS(re->ng, prev)) { |
25b9cb0c DL |
1123 | if (prev == nexthop) |
1124 | break; | |
996c9314 LB |
1125 | if (nexthop_same_firsthop(nexthop, prev)) { |
1126 | SET_FLAG(nexthop->flags, | |
1127 | NEXTHOP_FLAG_DUPLICATE); | |
25b9cb0c DL |
1128 | break; |
1129 | } | |
1130 | } | |
1131 | } | |
d62a17ae | 1132 | } |
718e3744 | 1133 | |
2063a814 DS |
1134 | /* |
1135 | * If this is a replace to a new RE let the originator of the RE | |
1136 | * know that they've lost | |
1137 | */ | |
9a9f8926 | 1138 | if (old && (old != re) && (old->type != re->type)) |
28610f7e | 1139 | zsend_route_notify_owner(old, p, ZAPI_ROUTE_BETTER_ADMIN_WON); |
25b9cb0c | 1140 | |
97f5b441 MS |
1141 | /* Update fib selection */ |
1142 | dest->selected_fib = re; | |
1143 | ||
d62a17ae | 1144 | /* |
1145 | * Make sure we update the FPM any time we send new information to | |
1146 | * the kernel. | |
1147 | */ | |
1148 | hook_call(rib_update, rn, "installing in kernel"); | |
97f5b441 MS |
1149 | |
1150 | /* Send add or update */ | |
5709131c | 1151 | if (old && (old != re)) |
97f5b441 | 1152 | ret = dplane_route_update(rn, re, old); |
5709131c | 1153 | else |
97f5b441 | 1154 | ret = dplane_route_add(rn, re); |
97f5b441 MS |
1155 | |
1156 | switch (ret) { | |
ea1c14f6 | 1157 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
97f5b441 MS |
1158 | if (zvrf) |
1159 | zvrf->installs_queued++; | |
1e885672 | 1160 | break; |
ea1c14f6 | 1161 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
1162 | { |
1163 | char str[SRCDEST2STR_BUFFER]; | |
1164 | ||
1165 | srcdest_rnode2str(rn, str, sizeof(str)); | |
1166 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
1167 | "%u:%s: Failed to enqueue dataplane install", | |
1168 | re->vrf_id, str); | |
1e885672 | 1169 | break; |
97f5b441 | 1170 | } |
ea1c14f6 | 1171 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
97f5b441 MS |
1172 | if (zvrf) |
1173 | zvrf->installs++; | |
1e885672 DS |
1174 | break; |
1175 | } | |
d62a17ae | 1176 | |
0c555cc6 | 1177 | return; |
718e3744 | 1178 | } |
1179 | ||
1180 | /* Uninstall the route from kernel. */ | |
0c555cc6 | 1181 | void rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 1182 | { |
d62a17ae | 1183 | struct nexthop *nexthop; |
1184 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 1185 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
05737783 | 1186 | |
d62a17ae | 1187 | if (info->safi != SAFI_UNICAST) { |
7ee30f28 | 1188 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d6792f9d | 1189 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 1190 | return; |
d62a17ae | 1191 | } |
416ec78d | 1192 | |
d62a17ae | 1193 | /* |
1194 | * Make sure we update the FPM any time we send new information to | |
97f5b441 | 1195 | * the dataplane. |
d62a17ae | 1196 | */ |
1197 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
97f5b441 MS |
1198 | |
1199 | switch (dplane_route_delete(rn, re)) { | |
ea1c14f6 | 1200 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
97f5b441 MS |
1201 | if (zvrf) |
1202 | zvrf->removals_queued++; | |
1e885672 | 1203 | break; |
ea1c14f6 | 1204 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
1205 | { |
1206 | char str[SRCDEST2STR_BUFFER]; | |
1207 | ||
1208 | srcdest_rnode2str(rn, str, sizeof(str)); | |
1209 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
1210 | "%u:%s: Failed to enqueue dataplane uninstall", | |
1211 | re->vrf_id, str); | |
1e885672 | 1212 | break; |
97f5b441 | 1213 | } |
ea1c14f6 | 1214 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
1e885672 DS |
1215 | if (zvrf) |
1216 | zvrf->removals++; | |
1217 | break; | |
1218 | } | |
718e3744 | 1219 | |
0c555cc6 | 1220 | return; |
718e3744 | 1221 | } |
1222 | ||
1223 | /* Uninstall the route from kernel. */ | |
d62a17ae | 1224 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 1225 | { |
d62a17ae | 1226 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
5f7a4718 | 1227 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
97f5b441 | 1228 | struct nexthop *nexthop; |
416ec78d | 1229 | |
5f7a4718 | 1230 | if (dest && dest->selected_fib == re) { |
d62a17ae | 1231 | if (info->safi == SAFI_UNICAST) |
1232 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 1233 | |
d62a17ae | 1234 | /* If labeled-unicast route, uninstall transit LSP. */ |
1235 | if (zebra_rib_labeled_unicast(re)) | |
1236 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
7cdb1a84 MS |
1237 | |
1238 | if (!RIB_SYSTEM_ROUTE(re)) | |
1239 | rib_uninstall_kernel(rn, re); | |
97f5b441 MS |
1240 | |
1241 | dest->selected_fib = NULL; | |
1242 | ||
1243 | for (ALL_NEXTHOPS(re->ng, nexthop)) | |
1244 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
d62a17ae | 1245 | } |
446bb95e | 1246 | |
d62a17ae | 1247 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
86391e56 MS |
1248 | const struct prefix *p, *src_p; |
1249 | ||
d62a17ae | 1250 | srcdest_rnode_prefixes(rn, &p, &src_p); |
05737783 | 1251 | |
d62a17ae | 1252 | redistribute_delete(p, src_p, re); |
1253 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); | |
1254 | } | |
718e3744 | 1255 | } |
1256 | ||
9fd92e3c AS |
1257 | /* |
1258 | * rib_can_delete_dest | |
1259 | * | |
1260 | * Returns TRUE if the given dest can be deleted from the table. | |
1261 | */ | |
d62a17ae | 1262 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 1263 | { |
d62a17ae | 1264 | if (dest->routes) { |
1265 | return 0; | |
1266 | } | |
9fd92e3c | 1267 | |
d62a17ae | 1268 | /* |
1269 | * Don't delete the dest if we have to update the FPM about this | |
1270 | * prefix. | |
1271 | */ | |
1272 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
1273 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
1274 | return 0; | |
5adc2528 | 1275 | |
d62a17ae | 1276 | return 1; |
9fd92e3c AS |
1277 | } |
1278 | ||
1279 | /* | |
1280 | * rib_gc_dest | |
1281 | * | |
1282 | * Garbage collect the rib dest corresponding to the given route node | |
1283 | * if appropriate. | |
1284 | * | |
1285 | * Returns TRUE if the dest was deleted, FALSE otherwise. | |
1286 | */ | |
d62a17ae | 1287 | int rib_gc_dest(struct route_node *rn) |
9fd92e3c | 1288 | { |
d62a17ae | 1289 | rib_dest_t *dest; |
9fd92e3c | 1290 | |
d62a17ae | 1291 | dest = rib_dest_from_rnode(rn); |
1292 | if (!dest) | |
1293 | return 0; | |
9fd92e3c | 1294 | |
d62a17ae | 1295 | if (!rib_can_delete_dest(dest)) |
1296 | return 0; | |
9fd92e3c | 1297 | |
c9abf558 DS |
1298 | if (IS_ZEBRA_DEBUG_RIB) { |
1299 | struct zebra_vrf *zvrf; | |
1300 | ||
1301 | zvrf = rib_dest_vrf(dest); | |
d62a17ae | 1302 | rnode_debug(rn, zvrf_id(zvrf), "removing dest from table"); |
c9abf558 | 1303 | } |
9fd92e3c | 1304 | |
d62a17ae | 1305 | dest->rnode = NULL; |
1306 | XFREE(MTYPE_RIB_DEST, dest); | |
1307 | rn->info = NULL; | |
9fd92e3c | 1308 | |
d62a17ae | 1309 | /* |
1310 | * Release the one reference that we keep on the route node. | |
1311 | */ | |
1312 | route_unlock_node(rn); | |
1313 | return 1; | |
9fd92e3c AS |
1314 | } |
1315 | ||
d62a17ae | 1316 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1317 | struct route_entry *new) | |
3e5c6e00 | 1318 | { |
5f7a4718 DS |
1319 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
1320 | ||
d62a17ae | 1321 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 1322 | |
d62a17ae | 1323 | /* Update real nexthop. This may actually determine if nexthop is active |
1324 | * or not. */ | |
0dddbf72 | 1325 | if (!nexthop_active_update(rn, new, true)) { |
d62a17ae | 1326 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
1327 | return; | |
1328 | } | |
3e5c6e00 | 1329 | |
d62a17ae | 1330 | if (IS_ZEBRA_DEBUG_RIB) { |
1331 | char buf[SRCDEST2STR_BUFFER]; | |
1332 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1333 | zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)", | |
1334 | zvrf_id(zvrf), buf, rn, new, new->type); | |
1335 | } | |
3e5c6e00 | 1336 | |
d62a17ae | 1337 | /* If labeled-unicast route, install transit LSP. */ |
1338 | if (zebra_rib_labeled_unicast(new)) | |
1339 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 1340 | |
0c555cc6 DS |
1341 | if (!RIB_SYSTEM_ROUTE(new)) |
1342 | rib_install_kernel(rn, new, NULL); | |
ed216282 DS |
1343 | else |
1344 | dest->selected_fib = new; | |
3e5c6e00 | 1345 | |
d62a17ae | 1346 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1347 | } |
1348 | ||
d62a17ae | 1349 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1350 | struct route_entry *old) | |
3e5c6e00 | 1351 | { |
5f7a4718 | 1352 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
d62a17ae | 1353 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 1354 | |
d62a17ae | 1355 | /* Uninstall from kernel. */ |
1356 | if (IS_ZEBRA_DEBUG_RIB) { | |
1357 | char buf[SRCDEST2STR_BUFFER]; | |
1358 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1359 | zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)", | |
1360 | zvrf_id(zvrf), buf, rn, old, old->type); | |
1361 | } | |
3e5c6e00 | 1362 | |
d62a17ae | 1363 | /* If labeled-unicast route, uninstall transit LSP. */ |
1364 | if (zebra_rib_labeled_unicast(old)) | |
1365 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1366 | ||
1367 | if (!RIB_SYSTEM_ROUTE(old)) | |
1368 | rib_uninstall_kernel(rn, old); | |
ed216282 DS |
1369 | else { |
1370 | /* | |
1371 | * We are setting this to NULL here | |
1372 | * because that is what we traditionally | |
1373 | * have been doing. I am not positive | |
1374 | * that this is the right thing to do | |
1375 | * but let's leave the code alone | |
1376 | * for the RIB_SYSTEM_ROUTE case | |
1377 | */ | |
1378 | dest->selected_fib = NULL; | |
1379 | } | |
d62a17ae | 1380 | |
1381 | /* Update nexthop for route, reset changed flag. */ | |
ef57f35f DL |
1382 | /* Note: this code also handles the Linux case when an interface goes |
1383 | * down, causing the kernel to delete routes without sending DELROUTE | |
1384 | * notifications | |
1385 | */ | |
0dddbf72 | 1386 | if (!nexthop_active_update(rn, old, true) && |
212df1de PG |
1387 | (RIB_KERNEL_ROUTE(old))) |
1388 | SET_FLAG(old->status, ROUTE_ENTRY_REMOVED); | |
1389 | else | |
1390 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1391 | } |
1392 | ||
1393 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
1394 | struct route_node *rn, | |
1395 | struct route_entry *old, | |
1396 | struct route_entry *new) | |
1397 | { | |
1398 | struct nexthop *nexthop = NULL; | |
1399 | int nh_active = 0; | |
5f7a4718 | 1400 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
d62a17ae | 1401 | |
1402 | /* | |
1403 | * We have to install or update if a new route has been selected or | |
1404 | * something has changed. | |
1405 | */ | |
1406 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
1407 | hook_call(rib_update, rn, "updating existing route"); | |
1408 | ||
1409 | /* Update the nexthop; we could determine here that nexthop is | |
1410 | * inactive. */ | |
0dddbf72 | 1411 | if (nexthop_active_update(rn, new, true)) |
d62a17ae | 1412 | nh_active = 1; |
1413 | ||
1414 | /* If nexthop is active, install the selected route, if | |
1415 | * appropriate. If | |
1416 | * the install succeeds, cleanup flags for prior route, if | |
1417 | * different from | |
1418 | * newly selected. | |
1419 | */ | |
1420 | if (nh_active) { | |
1421 | if (IS_ZEBRA_DEBUG_RIB) { | |
1422 | char buf[SRCDEST2STR_BUFFER]; | |
1423 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1424 | if (new != old) | |
1425 | zlog_debug( | |
1426 | "%u:%s: Updating route rn %p, re %p (type %d) " | |
1427 | "old %p (type %d)", | |
1428 | zvrf_id(zvrf), buf, rn, new, | |
1429 | new->type, old, old->type); | |
1430 | else | |
1431 | zlog_debug( | |
1432 | "%u:%s: Updating route rn %p, re %p (type %d)", | |
1433 | zvrf_id(zvrf), buf, rn, new, | |
1434 | new->type); | |
1435 | } | |
1436 | ||
1437 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1438 | if (zebra_rib_labeled_unicast(old)) | |
1439 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1440 | ||
1441 | /* Non-system route should be installed. */ | |
1442 | if (!RIB_SYSTEM_ROUTE(new)) { | |
1443 | /* If labeled-unicast route, install transit | |
1444 | * LSP. */ | |
1445 | if (zebra_rib_labeled_unicast(new)) | |
1446 | zebra_mpls_lsp_install(zvrf, rn, new); | |
1447 | ||
0c555cc6 | 1448 | rib_install_kernel(rn, new, old); |
ed216282 DS |
1449 | } else { |
1450 | /* | |
1451 | * We do not need to install the | |
1452 | * selected route because it | |
1453 | * is already isntalled by | |
1454 | * the system( ie not us ) | |
1455 | * so just mark it as winning | |
1456 | * we do need to ensure that | |
1457 | * if we uninstall a route | |
1458 | * from ourselves we don't | |
1459 | * over write this pointer | |
1460 | */ | |
1461 | dest->selected_fib = NULL; | |
d62a17ae | 1462 | } |
d62a17ae | 1463 | /* If install succeeded or system route, cleanup flags |
1464 | * for prior route. */ | |
ed216282 | 1465 | if (new != old) { |
d62a17ae | 1466 | if (RIB_SYSTEM_ROUTE(new)) { |
1467 | if (!RIB_SYSTEM_ROUTE(old)) | |
1468 | rib_uninstall_kernel(rn, old); | |
1469 | } else { | |
7ee30f28 | 1470 | for (nexthop = old->ng.nexthop; nexthop; |
d62a17ae | 1471 | nexthop = nexthop->next) |
1472 | UNSET_FLAG(nexthop->flags, | |
1473 | NEXTHOP_FLAG_FIB); | |
1474 | } | |
1475 | } | |
d62a17ae | 1476 | } |
a64448ba | 1477 | |
d62a17ae | 1478 | /* |
1479 | * If nexthop for selected route is not active or install | |
1480 | * failed, we | |
1481 | * may need to uninstall and delete for redistribution. | |
1482 | */ | |
ed216282 | 1483 | if (!nh_active) { |
d62a17ae | 1484 | if (IS_ZEBRA_DEBUG_RIB) { |
1485 | char buf[SRCDEST2STR_BUFFER]; | |
1486 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1487 | if (new != old) | |
1488 | zlog_debug( | |
1489 | "%u:%s: Deleting route rn %p, re %p (type %d) " | |
0cfbff74 | 1490 | "old %p (type %d) - nexthop inactive", |
d62a17ae | 1491 | zvrf_id(zvrf), buf, rn, new, |
0cfbff74 | 1492 | new->type, old, old->type); |
d62a17ae | 1493 | else |
1494 | zlog_debug( | |
0cfbff74 | 1495 | "%u:%s: Deleting route rn %p, re %p (type %d) - nexthop inactive", |
d62a17ae | 1496 | zvrf_id(zvrf), buf, rn, new, |
0cfbff74 | 1497 | new->type); |
d62a17ae | 1498 | } |
1499 | ||
1500 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1501 | if (zebra_rib_labeled_unicast(old)) | |
1502 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1503 | ||
1504 | if (!RIB_SYSTEM_ROUTE(old)) | |
1505 | rib_uninstall_kernel(rn, old); | |
ed216282 DS |
1506 | else |
1507 | dest->selected_fib = NULL; | |
d62a17ae | 1508 | } |
1509 | } else { | |
1510 | /* | |
1511 | * Same route selected; check if in the FIB and if not, | |
1512 | * re-install. This | |
1513 | * is housekeeping code to deal with race conditions in kernel | |
1514 | * with linux | |
1515 | * netlink reporting interface up before IPv4 or IPv6 protocol | |
1516 | * is ready | |
1517 | * to add routes. | |
1518 | */ | |
1519 | if (!RIB_SYSTEM_ROUTE(new)) { | |
ed216282 | 1520 | bool in_fib = false; |
d62a17ae | 1521 | |
7ee30f28 | 1522 | for (ALL_NEXTHOPS(new->ng, nexthop)) |
d62a17ae | 1523 | if (CHECK_FLAG(nexthop->flags, |
1524 | NEXTHOP_FLAG_FIB)) { | |
ed216282 | 1525 | in_fib = true; |
d62a17ae | 1526 | break; |
1527 | } | |
1528 | if (!in_fib) | |
1529 | rib_install_kernel(rn, new, NULL); | |
1530 | } | |
1531 | } | |
1532 | ||
1533 | /* Update prior route. */ | |
1534 | if (new != old) { | |
d62a17ae | 1535 | /* Set real nexthop. */ |
0dddbf72 | 1536 | nexthop_active_update(rn, old, true); |
d62a17ae | 1537 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); |
1538 | } | |
3e5c6e00 | 1539 | |
d62a17ae | 1540 | /* Clear changed flag. */ |
1541 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 1542 | } |
1543 | ||
d62a17ae | 1544 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
1545 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1546 | struct route_entry *alternate) | |
1547 | { | |
1548 | if (current == NULL) | |
1549 | return alternate; | |
1550 | ||
1551 | /* filter route selection in following order: | |
1552 | * - connected beats other types | |
fec4ca19 | 1553 | * - if both connected, loopback or vrf wins |
d62a17ae | 1554 | * - lower distance beats higher |
1555 | * - lower metric beats higher for equal distance | |
1556 | * - last, hence oldest, route wins tie break. | |
1557 | */ | |
1558 | ||
fec4ca19 DS |
1559 | /* Connected routes. Check to see if either are a vrf |
1560 | * or loopback interface. If not, pick the last connected | |
d62a17ae | 1561 | * route of the set of lowest metric connected routes. |
1562 | */ | |
1563 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
fec4ca19 DS |
1564 | if (current->type != ZEBRA_ROUTE_CONNECT) |
1565 | return alternate; | |
1566 | ||
1567 | /* both are connected. are either loop or vrf? */ | |
1568 | struct nexthop *nexthop = NULL; | |
1569 | ||
1570 | for (ALL_NEXTHOPS(alternate->ng, nexthop)) { | |
1571 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1572 | nexthop->ifindex, alternate->vrf_id))) | |
1573 | return alternate; | |
1574 | } | |
1575 | ||
1576 | for (ALL_NEXTHOPS(current->ng, nexthop)) { | |
1577 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1578 | nexthop->ifindex, current->vrf_id))) | |
1579 | return current; | |
1580 | } | |
1581 | ||
1582 | /* Neither are loop or vrf so pick best metric */ | |
1583 | if (alternate->metric <= current->metric) | |
d62a17ae | 1584 | return alternate; |
1585 | ||
1586 | return current; | |
1587 | } | |
3e5c6e00 | 1588 | |
d62a17ae | 1589 | if (current->type == ZEBRA_ROUTE_CONNECT) |
1590 | return current; | |
3e5c6e00 | 1591 | |
d62a17ae | 1592 | /* higher distance loses */ |
1593 | if (alternate->distance < current->distance) | |
1594 | return alternate; | |
1595 | if (current->distance < alternate->distance) | |
1596 | return current; | |
3e5c6e00 | 1597 | |
d62a17ae | 1598 | /* metric tie-breaks equal distance */ |
1599 | if (alternate->metric <= current->metric) | |
1600 | return alternate; | |
3e5c6e00 | 1601 | |
d62a17ae | 1602 | return current; |
3e5c6e00 | 1603 | } |
1604 | ||
d62a17ae | 1605 | /* Core function for processing routing information base. */ |
1606 | static void rib_process(struct route_node *rn) | |
1607 | { | |
1608 | struct route_entry *re; | |
1609 | struct route_entry *next; | |
1610 | struct route_entry *old_selected = NULL; | |
1611 | struct route_entry *new_selected = NULL; | |
1612 | struct route_entry *old_fib = NULL; | |
1613 | struct route_entry *new_fib = NULL; | |
1614 | struct route_entry *best = NULL; | |
1615 | char buf[SRCDEST2STR_BUFFER]; | |
1616 | rib_dest_t *dest; | |
1617 | struct zebra_vrf *zvrf = NULL; | |
86391e56 MS |
1618 | const struct prefix *p, *src_p; |
1619 | ||
d62a17ae | 1620 | srcdest_rnode_prefixes(rn, &p, &src_p); |
1621 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1622 | ||
1623 | assert(rn); | |
1624 | ||
1625 | dest = rib_dest_from_rnode(rn); | |
1626 | if (dest) { | |
1627 | zvrf = rib_dest_vrf(dest); | |
1628 | vrf_id = zvrf_id(zvrf); | |
1629 | } | |
bab85d4f | 1630 | |
d62a17ae | 1631 | if (IS_ZEBRA_DEBUG_RIB) |
1632 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1633 | |
d62a17ae | 1634 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1635 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1636 | |
607425e5 DS |
1637 | /* |
1638 | * we can have rn's that have a NULL info pointer | |
1639 | * (dest). As such let's not let the deref happen | |
1640 | * additionally we know RNODE_FOREACH_RE_SAFE | |
1641 | * will not iterate so we are ok. | |
1642 | */ | |
1643 | if (dest) | |
1644 | old_fib = dest->selected_fib; | |
5f7a4718 | 1645 | |
a2addae8 | 1646 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1647 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1648 | zlog_debug( | |
1649 | "%u:%s: Examine re %p (type %d) status %x flags %x " | |
1650 | "dist %d metric %d", | |
1651 | vrf_id, buf, re, re->type, re->status, | |
1652 | re->flags, re->distance, re->metric); | |
1653 | ||
1654 | UNSET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1655 | ||
1656 | /* Currently selected re. */ | |
1657 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1658 | assert(old_selected == NULL); | |
1659 | old_selected = re; | |
1660 | } | |
bab85d4f | 1661 | |
d62a17ae | 1662 | /* Skip deleted entries from selection */ |
1663 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1664 | continue; | |
1665 | ||
1666 | /* Skip unreachable nexthop. */ | |
1667 | /* This first call to nexthop_active_update is merely to | |
1668 | * determine if | |
1669 | * there's any change to nexthops associated with this RIB | |
1670 | * entry. Now, | |
1671 | * rib_process() can be invoked due to an external event such as | |
1672 | * link | |
1673 | * down or due to next-hop-tracking evaluation. In the latter | |
1674 | * case, | |
1675 | * a decision has already been made that the NHs have changed. | |
1676 | * So, no | |
1677 | * need to invoke a potentially expensive call again. Further, | |
1678 | * since | |
1679 | * the change might be in a recursive NH which is not caught in | |
1680 | * the nexthop_active_update() code. Thus, we might miss changes | |
1681 | * to | |
1682 | * recursive NHs. | |
1683 | */ | |
1684 | if (!CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) | |
0dddbf72 | 1685 | && !nexthop_active_update(rn, re, false)) { |
d62a17ae | 1686 | if (re->type == ZEBRA_ROUTE_TABLE) { |
1687 | /* XXX: HERE BE DRAGONS!!!!! | |
1688 | * In all honesty, I have not yet figured out | |
1689 | * what this part | |
1690 | * does or why the ROUTE_ENTRY_CHANGED test | |
1691 | * above is correct | |
1692 | * or why we need to delete a route here, and | |
1693 | * also not whether | |
1694 | * this concerns both selected and fib route, or | |
1695 | * only selected | |
1696 | * or only fib */ | |
1697 | /* This entry was denied by the 'ip protocol | |
1698 | * table' route-map, we | |
1699 | * need to delete it */ | |
1700 | if (re != old_selected) { | |
1701 | if (IS_ZEBRA_DEBUG_RIB) | |
1702 | zlog_debug( | |
32391aff | 1703 | "%s: %u:%s: imported via import-table but denied " |
d62a17ae | 1704 | "by the ip protocol table route-map", |
32391aff | 1705 | __func__, vrf_id, buf); |
d62a17ae | 1706 | rib_unlink(rn, re); |
1707 | } else | |
1708 | SET_FLAG(re->status, | |
1709 | ROUTE_ENTRY_REMOVED); | |
1710 | } | |
1711 | ||
1712 | continue; | |
1713 | } | |
bab85d4f | 1714 | |
d62a17ae | 1715 | /* Infinite distance. */ |
1716 | if (re->distance == DISTANCE_INFINITY) { | |
1717 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1718 | continue; | |
1719 | } | |
bab85d4f | 1720 | |
d62a17ae | 1721 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1722 | best = rib_choose_best(new_fib, re); | |
1723 | if (new_fib && best != new_fib) | |
1724 | UNSET_FLAG(new_fib->status, | |
1725 | ROUTE_ENTRY_CHANGED); | |
1726 | new_fib = best; | |
1727 | } else { | |
1728 | best = rib_choose_best(new_selected, re); | |
1729 | if (new_selected && best != new_selected) | |
1730 | UNSET_FLAG(new_selected->status, | |
1731 | ROUTE_ENTRY_CHANGED); | |
1732 | new_selected = best; | |
1733 | } | |
1734 | if (best != re) | |
1735 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1736 | } /* RNODE_FOREACH_RE */ | |
1737 | ||
1738 | /* If no FIB override route, use the selected route also for FIB */ | |
1739 | if (new_fib == NULL) | |
1740 | new_fib = new_selected; | |
1741 | ||
1742 | /* After the cycle is finished, the following pointers will be set: | |
1743 | * old_selected --- RE entry currently having SELECTED | |
1744 | * new_selected --- RE entry that is newly SELECTED | |
1745 | * old_fib --- RE entry currently in kernel FIB | |
1746 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1747 | * | |
1748 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1749 | * the zclients. new_fib (which can be new_selected) will be installed | |
1750 | * in kernel. | |
1751 | */ | |
1752 | ||
1753 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1754 | zlog_debug( | |
1755 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1756 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1757 | (void *)old_fib, (void *)new_fib); | |
1758 | } | |
446bb95e | 1759 | |
d62a17ae | 1760 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1761 | * fib == selected */ | |
9d303b37 DL |
1762 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, |
1763 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1764 | |
1765 | /* Update fib according to selection results */ | |
1766 | if (new_fib && old_fib) | |
1767 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1768 | else if (new_fib) | |
1769 | rib_process_add_fib(zvrf, rn, new_fib); | |
1770 | else if (old_fib) | |
1771 | rib_process_del_fib(zvrf, rn, old_fib); | |
1772 | ||
8cb41cd6 | 1773 | /* Update SELECTED entry */ |
d62a17ae | 1774 | if (old_selected != new_selected || selected_changed) { |
93bdadae | 1775 | |
d62a17ae | 1776 | if (new_selected && new_selected != new_fib) { |
0dddbf72 | 1777 | nexthop_active_update(rn, new_selected, true); |
d62a17ae | 1778 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); |
1779 | } | |
41ec9222 | 1780 | |
5af4b346 MS |
1781 | if (new_selected) { |
1782 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); | |
1783 | ||
1784 | /* Special case: new route is system route, so | |
1785 | * dataplane update will not be done - ensure we | |
1786 | * redistribute the route. | |
1787 | */ | |
1788 | if (RIB_SYSTEM_ROUTE(new_selected)) | |
1789 | redistribute_update(p, src_p, new_selected, | |
1790 | old_selected); | |
1791 | } | |
1792 | ||
d62a17ae | 1793 | if (old_selected) { |
1794 | if (!new_selected) | |
1795 | redistribute_delete(p, src_p, old_selected); | |
1796 | if (old_selected != new_selected) | |
1797 | UNSET_FLAG(old_selected->flags, | |
1798 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1799 | } |
d62a17ae | 1800 | } |
3e5c6e00 | 1801 | |
d62a17ae | 1802 | /* Remove all RE entries queued for removal */ |
a2addae8 | 1803 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1804 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { |
1805 | if (IS_ZEBRA_DEBUG_RIB) { | |
1806 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1807 | (void *)rn, (void *)re); | |
1808 | } | |
1809 | rib_unlink(rn, re); | |
1810 | } | |
1811 | } | |
4d38fdb4 | 1812 | |
d62a17ae | 1813 | /* |
1814 | * Check if the dest can be deleted now. | |
1815 | */ | |
1816 | rib_gc_dest(rn); | |
e96f9203 DO |
1817 | } |
1818 | ||
e5ac2adf MS |
1819 | /* |
1820 | * Utility to match route with dplane context data | |
1821 | */ | |
1822 | static bool rib_route_match_ctx(const struct route_entry *re, | |
25779064 MS |
1823 | const struct zebra_dplane_ctx *ctx, |
1824 | bool is_update) | |
e5ac2adf MS |
1825 | { |
1826 | bool result = false; | |
1827 | ||
1828 | if (is_update) { | |
1829 | /* | |
1830 | * In 'update' case, we test info about the 'previous' or | |
1831 | * 'old' route | |
1832 | */ | |
1833 | if ((re->type == dplane_ctx_get_old_type(ctx)) && | |
1834 | (re->instance == dplane_ctx_get_old_instance(ctx))) { | |
1835 | result = true; | |
1836 | ||
1837 | /* TODO -- we're using this extra test, but it's not | |
1838 | * exactly clear why. | |
1839 | */ | |
1840 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1841 | (re->distance != dplane_ctx_get_old_distance(ctx) || | |
1842 | re->tag != dplane_ctx_get_old_tag(ctx))) { | |
1843 | result = false; | |
1844 | } | |
1845 | } | |
1846 | ||
1847 | } else { | |
1848 | /* | |
1849 | * Ordinary, single-route case using primary context info | |
1850 | */ | |
1851 | if ((dplane_ctx_get_op(ctx) != DPLANE_OP_ROUTE_DELETE) && | |
1852 | CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1853 | /* Skip route that's been deleted */ | |
1854 | goto done; | |
1855 | } | |
1856 | ||
1857 | if ((re->type == dplane_ctx_get_type(ctx)) && | |
1858 | (re->instance == dplane_ctx_get_instance(ctx))) { | |
1859 | result = true; | |
1860 | ||
1861 | /* TODO -- we're using this extra test, but it's not | |
1862 | * exactly clear why. | |
1863 | */ | |
1864 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1865 | (re->distance != dplane_ctx_get_distance(ctx) || | |
1866 | re->tag != dplane_ctx_get_tag(ctx))) { | |
1867 | result = false; | |
1868 | } | |
1869 | } | |
1870 | } | |
1871 | ||
1872 | done: | |
1873 | ||
1874 | return (result); | |
1875 | } | |
1876 | ||
1877 | /* | |
91f16812 | 1878 | * Route-update results processing after async dataplane update. |
e5ac2adf | 1879 | */ |
25779064 | 1880 | static void rib_process_after(struct zebra_dplane_ctx *ctx) |
e5ac2adf MS |
1881 | { |
1882 | struct route_table *table = NULL; | |
97f5b441 | 1883 | struct zebra_vrf *zvrf = NULL; |
e5ac2adf MS |
1884 | struct route_node *rn = NULL; |
1885 | struct route_entry *re = NULL, *old_re = NULL, *rib; | |
1886 | bool is_update = false; | |
f183e380 | 1887 | struct nexthop *nexthop, *ctx_nexthop; |
97f5b441 | 1888 | char dest_str[PREFIX_STRLEN] = ""; |
5709131c | 1889 | enum dplane_op_e op; |
e5ac2adf MS |
1890 | enum zebra_dplane_result status; |
1891 | const struct prefix *dest_pfx, *src_pfx; | |
1892 | ||
1893 | /* Locate rn and re(s) from ctx */ | |
1894 | ||
1895 | table = zebra_vrf_table_with_table_id(dplane_ctx_get_afi(ctx), | |
1896 | dplane_ctx_get_safi(ctx), | |
1897 | dplane_ctx_get_vrf(ctx), | |
1898 | dplane_ctx_get_table(ctx)); | |
1899 | if (table == NULL) { | |
1900 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fe2c53d4 | 1901 | zlog_debug("Failed to process dplane results: no table for afi %d, safi %d, vrf %u", |
e5ac2adf MS |
1902 | dplane_ctx_get_afi(ctx), |
1903 | dplane_ctx_get_safi(ctx), | |
1904 | dplane_ctx_get_vrf(ctx)); | |
1905 | } | |
1906 | goto done; | |
1907 | } | |
1908 | ||
97f5b441 MS |
1909 | zvrf = vrf_info_lookup(dplane_ctx_get_vrf(ctx)); |
1910 | ||
e5ac2adf MS |
1911 | dest_pfx = dplane_ctx_get_dest(ctx); |
1912 | ||
1913 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1914 | * unconditional log messages will have to generate the string. | |
1915 | */ | |
5709131c | 1916 | if (IS_ZEBRA_DEBUG_DPLANE) |
e5ac2adf | 1917 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
e5ac2adf MS |
1918 | |
1919 | src_pfx = dplane_ctx_get_src(ctx); | |
1920 | rn = srcdest_rnode_get(table, dplane_ctx_get_dest(ctx), | |
5709131c | 1921 | src_pfx ? (struct prefix_ipv6 *)src_pfx : NULL); |
e5ac2adf MS |
1922 | if (rn == NULL) { |
1923 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fe2c53d4 | 1924 | zlog_debug("Failed to process dplane results: no route for %u:%s", |
e5ac2adf MS |
1925 | dplane_ctx_get_vrf(ctx), dest_str); |
1926 | } | |
1927 | goto done; | |
1928 | } | |
1929 | ||
1930 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); | |
1931 | ||
1932 | op = dplane_ctx_get_op(ctx); | |
1933 | status = dplane_ctx_get_status(ctx); | |
1934 | ||
c831033f | 1935 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
f183e380 | 1936 | zlog_debug("%u:%s Processing dplane ctx %p, op %s result %s", |
e5ac2adf | 1937 | dplane_ctx_get_vrf(ctx), dest_str, ctx, |
f183e380 | 1938 | dplane_op2str(op), dplane_res2str(status)); |
e5ac2adf MS |
1939 | |
1940 | if (op == DPLANE_OP_ROUTE_DELETE) { | |
1941 | /* | |
1942 | * In the delete case, the zebra core datastructs were | |
1943 | * updated (or removed) at the time the delete was issued, | |
1944 | * so we're just notifying the route owner. | |
1945 | */ | |
1946 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
1947 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_REMOVED); | |
97f5b441 | 1948 | |
5709131c | 1949 | if (zvrf) |
97f5b441 | 1950 | zvrf->removals++; |
e5ac2adf MS |
1951 | } else { |
1952 | zsend_route_notify_owner_ctx(ctx, | |
1953 | ZAPI_ROUTE_FAIL_INSTALL); | |
1954 | ||
1955 | zlog_warn("%u:%s: Route Deletion failure", | |
1956 | dplane_ctx_get_vrf(ctx), | |
1957 | prefix2str(dest_pfx, | |
1958 | dest_str, sizeof(dest_str))); | |
1959 | } | |
1960 | ||
1961 | /* Nothing more to do in delete case */ | |
1962 | goto done; | |
1963 | } | |
1964 | ||
1965 | /* | |
1966 | * Update is a bit of a special case, where we may have both old and new | |
1967 | * routes to post-process. | |
1968 | */ | |
1969 | is_update = dplane_ctx_is_update(ctx); | |
1970 | ||
1971 | /* | |
1972 | * Take a pass through the routes, look for matches with the context | |
1973 | * info. | |
1974 | */ | |
1975 | RNODE_FOREACH_RE(rn, rib) { | |
1976 | ||
1977 | if (re == NULL) { | |
5709131c | 1978 | if (rib_route_match_ctx(rib, ctx, false)) |
e5ac2adf | 1979 | re = rib; |
e5ac2adf MS |
1980 | } |
1981 | ||
1982 | /* Check for old route match */ | |
1983 | if (is_update && (old_re == NULL)) { | |
5709131c | 1984 | if (rib_route_match_ctx(rib, ctx, true /*is_update*/)) |
e5ac2adf | 1985 | old_re = rib; |
e5ac2adf MS |
1986 | } |
1987 | ||
1988 | /* Have we found the routes we need to work on? */ | |
5709131c | 1989 | if (re && ((!is_update || old_re))) |
e5ac2adf | 1990 | break; |
e5ac2adf MS |
1991 | } |
1992 | ||
1993 | /* | |
1994 | * Check sequence number(s) to detect stale results before continuing | |
1995 | */ | |
1996 | if (re && (re->dplane_sequence != dplane_ctx_get_seq(ctx))) { | |
1997 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) { | |
1998 | zlog_debug("%u:%s Stale dplane result for re %p", | |
1999 | dplane_ctx_get_vrf(ctx), dest_str, re); | |
2000 | } | |
2001 | re = NULL; | |
2002 | } | |
2003 | ||
2004 | if (old_re && | |
2005 | (old_re->dplane_sequence != dplane_ctx_get_old_seq(ctx))) { | |
2006 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) { | |
2007 | zlog_debug("%u:%s Stale dplane result for old_re %p", | |
2008 | dplane_ctx_get_vrf(ctx), dest_str, old_re); | |
2009 | } | |
2010 | old_re = NULL; | |
2011 | } | |
2012 | ||
2013 | /* | |
2014 | * Here's sort of a tough one: the route update result is stale. | |
2015 | * Is it better to use the context block info to generate | |
2016 | * redist and owner notification, or is it better to wait | |
2017 | * for the up-to-date result to arrive? | |
2018 | */ | |
2019 | if (re == NULL) { | |
2020 | /* TODO -- for now, only expose up-to-date results */ | |
2021 | goto done; | |
2022 | } | |
2023 | ||
2024 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
f183e380 MS |
2025 | /* Update zebra nexthop FIB flag for each |
2026 | * nexthop that was installed. | |
2027 | */ | |
2028 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), ctx_nexthop)) { | |
2029 | ||
2030 | for (ALL_NEXTHOPS(re->ng, nexthop)) { | |
2031 | if (nexthop_same(ctx_nexthop, nexthop)) | |
2032 | break; | |
2033 | } | |
2034 | ||
2035 | if (nexthop == NULL) | |
2036 | continue; | |
2037 | ||
e5ac2adf MS |
2038 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) |
2039 | continue; | |
2040 | ||
f183e380 MS |
2041 | if (CHECK_FLAG(ctx_nexthop->flags, |
2042 | NEXTHOP_FLAG_FIB)) | |
e5ac2adf MS |
2043 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
2044 | else | |
2045 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
2046 | } | |
2047 | ||
f183e380 | 2048 | if (zvrf) { |
97f5b441 | 2049 | zvrf->installs++; |
f183e380 MS |
2050 | /* Set flag for nexthop tracking processing */ |
2051 | zvrf->flags |= ZEBRA_VRF_RIB_SCHEDULED; | |
2052 | } | |
97f5b441 | 2053 | |
e5ac2adf MS |
2054 | /* Redistribute */ |
2055 | /* TODO -- still calling the redist api using the route_entries, | |
2056 | * and there's a corner-case here: if there's no client | |
2057 | * for the 'new' route, a redist deleting the 'old' route | |
2058 | * will be sent. But if the 'old' context info was stale, | |
2059 | * 'old_re' will be NULL here and that delete will not be sent. | |
2060 | */ | |
2061 | redistribute_update(dest_pfx, src_pfx, re, old_re); | |
2062 | ||
2063 | /* Notify route owner */ | |
2064 | zsend_route_notify_owner(re, | |
2065 | dest_pfx, ZAPI_ROUTE_INSTALLED); | |
2066 | ||
2067 | } else { | |
2068 | zsend_route_notify_owner(re, dest_pfx, | |
2069 | ZAPI_ROUTE_FAIL_INSTALL); | |
2070 | ||
2071 | zlog_warn("%u:%s: Route install failed", | |
2072 | dplane_ctx_get_vrf(ctx), | |
2073 | prefix2str(dest_pfx, | |
2074 | dest_str, sizeof(dest_str))); | |
2075 | } | |
2076 | ||
2077 | done: | |
2078 | ||
2079 | /* Return context to dataplane module */ | |
2080 | dplane_ctx_fini(&ctx); | |
2081 | } | |
2082 | ||
5110a0c6 | 2083 | /* Take a list of route_node structs and return 1, if there was a record |
d62a17ae | 2084 | * picked from it and processed by rib_process(). Don't process more, |
5110a0c6 | 2085 | * than one RN record; operate only in the specified sub-queue. |
e96f9203 | 2086 | */ |
d7c0a89a | 2087 | static unsigned int process_subq(struct list *subq, uint8_t qindex) |
e96f9203 | 2088 | { |
d62a17ae | 2089 | struct listnode *lnode = listhead(subq); |
2090 | struct route_node *rnode; | |
2091 | rib_dest_t *dest; | |
2092 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 2093 | |
d62a17ae | 2094 | if (!lnode) |
2095 | return 0; | |
5110a0c6 | 2096 | |
d62a17ae | 2097 | rnode = listgetdata(lnode); |
2098 | dest = rib_dest_from_rnode(rnode); | |
2099 | if (dest) | |
2100 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 2101 | |
d62a17ae | 2102 | rib_process(rnode); |
5110a0c6 | 2103 | |
d62a17ae | 2104 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
2105 | char buf[SRCDEST2STR_BUFFER]; | |
2106 | srcdest_rnode2str(rnode, buf, sizeof(buf)); | |
2107 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
2108 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
2109 | } | |
41ec9222 | 2110 | |
d62a17ae | 2111 | if (rnode->info) |
2112 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
2113 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 2114 | |
67b9467f | 2115 | #if 0 |
5110a0c6 SH |
2116 | else |
2117 | { | |
2118 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
2119 | __func__, rnode, rnode->lock); | |
2120 | zlog_backtrace(LOG_DEBUG); | |
2121 | } | |
67b9467f | 2122 | #endif |
d62a17ae | 2123 | route_unlock_node(rnode); |
2124 | list_delete_node(subq, lnode); | |
2125 | return 1; | |
e96f9203 DO |
2126 | } |
2127 | ||
fb018d25 | 2128 | /* |
f183e380 | 2129 | * Perform next-hop tracking processing after RIB updates. |
fb018d25 | 2130 | */ |
f183e380 | 2131 | static void do_nht_processing(void) |
fb018d25 | 2132 | { |
d62a17ae | 2133 | struct vrf *vrf; |
2134 | struct zebra_vrf *zvrf; | |
9ec6b0bb | 2135 | |
d62a17ae | 2136 | /* Evaluate nexthops for those VRFs which underwent route processing. |
2137 | * This | |
2138 | * should limit the evaluation to the necessary VRFs in most common | |
2139 | * situations. | |
2140 | */ | |
a2addae8 | 2141 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
d62a17ae | 2142 | zvrf = vrf->info; |
2143 | if (zvrf == NULL || !(zvrf->flags & ZEBRA_VRF_RIB_SCHEDULED)) | |
2144 | continue; | |
2145 | ||
f183e380 MS |
2146 | if (IS_ZEBRA_DEBUG_RIB_DETAILED || IS_ZEBRA_DEBUG_NHT) |
2147 | zlog_debug("NHT processing check for zvrf %s", | |
2148 | zvrf_name(zvrf)); | |
2149 | ||
d62a17ae | 2150 | zvrf->flags &= ~ZEBRA_VRF_RIB_SCHEDULED; |
6d53d7b1 | 2151 | zebra_evaluate_rnh(zvrf, AF_INET, 0, RNH_NEXTHOP_TYPE, NULL); |
2152 | zebra_evaluate_rnh(zvrf, AF_INET, 0, RNH_IMPORT_CHECK_TYPE, | |
d62a17ae | 2153 | NULL); |
6d53d7b1 | 2154 | zebra_evaluate_rnh(zvrf, AF_INET6, 0, RNH_NEXTHOP_TYPE, NULL); |
2155 | zebra_evaluate_rnh(zvrf, AF_INET6, 0, RNH_IMPORT_CHECK_TYPE, | |
d62a17ae | 2156 | NULL); |
d62a17ae | 2157 | } |
939fba27 | 2158 | |
d62a17ae | 2159 | /* Schedule LSPs for processing, if needed. */ |
2160 | zvrf = vrf_info_lookup(VRF_DEFAULT); | |
2161 | if (mpls_should_lsps_be_processed(zvrf)) { | |
2162 | if (IS_ZEBRA_DEBUG_MPLS) | |
2163 | zlog_debug( | |
2164 | "%u: Scheduling all LSPs upon RIB completion", | |
2165 | zvrf_id(zvrf)); | |
2166 | zebra_mpls_lsp_schedule(zvrf); | |
2167 | mpls_unmark_lsps_for_processing(zvrf); | |
2168 | } | |
fb018d25 DS |
2169 | } |
2170 | ||
f183e380 MS |
2171 | /* |
2172 | * All meta queues have been processed. Trigger next-hop evaluation. | |
2173 | */ | |
2174 | static void meta_queue_process_complete(struct work_queue *dummy) | |
2175 | { | |
2176 | do_nht_processing(); | |
2177 | } | |
2178 | ||
e96f9203 | 2179 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from |
d62a17ae | 2180 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
2181 | * data | |
e96f9203 DO |
2182 | * is pointed to the meta queue structure. |
2183 | */ | |
d62a17ae | 2184 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 2185 | { |
d62a17ae | 2186 | struct meta_queue *mq = data; |
2187 | unsigned i; | |
91f16812 MS |
2188 | uint32_t queue_len, queue_limit; |
2189 | ||
2190 | /* Ensure there's room for more dataplane updates */ | |
2191 | queue_limit = dplane_get_in_queue_limit(); | |
2192 | queue_len = dplane_get_in_queue_len(); | |
2193 | if (queue_len > queue_limit) { | |
2194 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
8b962e77 MS |
2195 | zlog_debug("rib queue: dplane queue len %u, limit %u, retrying", |
2196 | queue_len, queue_limit); | |
91f16812 MS |
2197 | |
2198 | /* Ensure that the meta-queue is actually enqueued */ | |
2199 | if (work_queue_empty(zebrad.ribq)) | |
2200 | work_queue_add(zebrad.ribq, zebrad.mq); | |
2201 | ||
2202 | return WQ_QUEUE_BLOCKED; | |
2203 | } | |
5110a0c6 | 2204 | |
d62a17ae | 2205 | for (i = 0; i < MQ_SIZE; i++) |
2206 | if (process_subq(mq->subq[i], i)) { | |
2207 | mq->size--; | |
2208 | break; | |
2209 | } | |
2210 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
2211 | } |
2212 | ||
9fd92e3c AS |
2213 | /* |
2214 | * Map from rib types to queue type (priority) in meta queue | |
2215 | */ | |
d7c0a89a QY |
2216 | static const uint8_t meta_queue_map[ZEBRA_ROUTE_MAX] = { |
2217 | [ZEBRA_ROUTE_SYSTEM] = 4, | |
2218 | [ZEBRA_ROUTE_KERNEL] = 0, | |
2219 | [ZEBRA_ROUTE_CONNECT] = 0, | |
2220 | [ZEBRA_ROUTE_STATIC] = 1, | |
2221 | [ZEBRA_ROUTE_RIP] = 2, | |
2222 | [ZEBRA_ROUTE_RIPNG] = 2, | |
2223 | [ZEBRA_ROUTE_OSPF] = 2, | |
2224 | [ZEBRA_ROUTE_OSPF6] = 2, | |
2225 | [ZEBRA_ROUTE_ISIS] = 2, | |
2226 | [ZEBRA_ROUTE_BGP] = 3, | |
2227 | [ZEBRA_ROUTE_PIM] = 4, // Shouldn't happen but for safety | |
2228 | [ZEBRA_ROUTE_EIGRP] = 2, | |
2229 | [ZEBRA_ROUTE_NHRP] = 2, | |
2230 | [ZEBRA_ROUTE_HSLS] = 4, | |
2231 | [ZEBRA_ROUTE_OLSR] = 4, | |
2232 | [ZEBRA_ROUTE_TABLE] = 1, | |
2233 | [ZEBRA_ROUTE_LDP] = 4, | |
2234 | [ZEBRA_ROUTE_VNC] = 3, | |
2235 | [ZEBRA_ROUTE_VNC_DIRECT] = 3, | |
2236 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = 3, | |
2237 | [ZEBRA_ROUTE_BGP_DIRECT] = 3, | |
2238 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = 3, | |
2239 | [ZEBRA_ROUTE_BABEL] = 2, | |
2240 | [ZEBRA_ROUTE_ALL] = 4, // Shouldn't happen but for safety | |
5110a0c6 SH |
2241 | }; |
2242 | ||
2243 | /* Look into the RN and queue it into one or more priority queues, | |
2244 | * increasing the size for each data push done. | |
e96f9203 | 2245 | */ |
d62a17ae | 2246 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) |
e96f9203 | 2247 | { |
d62a17ae | 2248 | struct route_entry *re; |
5110a0c6 | 2249 | |
a2addae8 | 2250 | RNODE_FOREACH_RE (rn, re) { |
d7c0a89a | 2251 | uint8_t qindex = meta_queue_map[re->type]; |
d62a17ae | 2252 | struct zebra_vrf *zvrf; |
2253 | ||
2254 | /* Invariant: at this point we always have rn->info set. */ | |
2255 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
2256 | RIB_ROUTE_QUEUED(qindex))) { | |
2257 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2258 | rnode_debug( | |
2259 | rn, re->vrf_id, | |
2260 | "rn %p is already queued in sub-queue %u", | |
2261 | (void *)rn, qindex); | |
2262 | continue; | |
2263 | } | |
5110a0c6 | 2264 | |
d62a17ae | 2265 | SET_FLAG(rib_dest_from_rnode(rn)->flags, |
2266 | RIB_ROUTE_QUEUED(qindex)); | |
2267 | listnode_add(mq->subq[qindex], rn); | |
2268 | route_lock_node(rn); | |
2269 | mq->size++; | |
5110a0c6 | 2270 | |
d62a17ae | 2271 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2272 | rnode_debug(rn, re->vrf_id, | |
2273 | "queued rn %p into sub-queue %u", | |
2274 | (void *)rn, qindex); | |
9ec6b0bb | 2275 | |
d62a17ae | 2276 | zvrf = zebra_vrf_lookup_by_id(re->vrf_id); |
2277 | if (zvrf) | |
2278 | zvrf->flags |= ZEBRA_VRF_RIB_SCHEDULED; | |
2279 | } | |
4d38fdb4 | 2280 | } |
2281 | ||
6d691129 | 2282 | /* Add route_node to work queue and schedule processing */ |
d62a17ae | 2283 | void rib_queue_add(struct route_node *rn) |
4d38fdb4 | 2284 | { |
d62a17ae | 2285 | assert(rn); |
fc328ac9 | 2286 | |
d62a17ae | 2287 | /* Pointless to queue a route_node with no RIB entries to add or remove |
2288 | */ | |
2289 | if (!rnode_to_ribs(rn)) { | |
2290 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
2291 | __func__, (void *)rn, rn->lock); | |
2292 | zlog_backtrace(LOG_DEBUG); | |
2293 | return; | |
2294 | } | |
4d38fdb4 | 2295 | |
d62a17ae | 2296 | if (zebrad.ribq == NULL) { |
e914ccbe | 2297 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2298 | "%s: work_queue does not exist!", __func__); |
d62a17ae | 2299 | return; |
2300 | } | |
2301 | ||
2302 | /* | |
2303 | * The RIB queue should normally be either empty or holding the only | |
2304 | * work_queue_item element. In the latter case this element would | |
2305 | * hold a pointer to the meta queue structure, which must be used to | |
2306 | * actually queue the route nodes to process. So create the MQ | |
2307 | * holder, if necessary, then push the work into it in any case. | |
2308 | * This semantics was introduced after 0.99.9 release. | |
2309 | */ | |
f104f6c1 | 2310 | if (work_queue_empty(zebrad.ribq)) |
d62a17ae | 2311 | work_queue_add(zebrad.ribq, zebrad.mq); |
e96f9203 | 2312 | |
d62a17ae | 2313 | rib_meta_queue_add(zebrad.mq, rn); |
fc328ac9 | 2314 | |
d62a17ae | 2315 | return; |
4d38fdb4 | 2316 | } |
2317 | ||
5110a0c6 SH |
2318 | /* Create new meta queue. |
2319 | A destructor function doesn't seem to be necessary here. | |
2320 | */ | |
d62a17ae | 2321 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 2322 | { |
d62a17ae | 2323 | struct meta_queue *new; |
2324 | unsigned i; | |
5110a0c6 | 2325 | |
d62a17ae | 2326 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
e96f9203 | 2327 | |
d62a17ae | 2328 | for (i = 0; i < MQ_SIZE; i++) { |
2329 | new->subq[i] = list_new(); | |
2330 | assert(new->subq[i]); | |
2331 | } | |
5110a0c6 | 2332 | |
d62a17ae | 2333 | return new; |
e96f9203 DO |
2334 | } |
2335 | ||
d62a17ae | 2336 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 2337 | { |
d62a17ae | 2338 | unsigned i; |
5a8dfcd8 | 2339 | |
d62a17ae | 2340 | for (i = 0; i < MQ_SIZE; i++) |
6a154c88 | 2341 | list_delete(&mq->subq[i]); |
5a8dfcd8 | 2342 | |
d62a17ae | 2343 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
2344 | } |
2345 | ||
4d38fdb4 | 2346 | /* initialise zebra rib work queue */ |
d62a17ae | 2347 | static void rib_queue_init(struct zebra_t *zebra) |
4d38fdb4 | 2348 | { |
d62a17ae | 2349 | assert(zebra); |
2350 | ||
2351 | if (!(zebra->ribq = | |
2352 | work_queue_new(zebra->master, "route_node processing"))) { | |
e914ccbe | 2353 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2354 | "%s: could not initialise work queue!", __func__); |
d62a17ae | 2355 | return; |
2356 | } | |
4d38fdb4 | 2357 | |
d62a17ae | 2358 | /* fill in the work queue spec */ |
2359 | zebra->ribq->spec.workfunc = &meta_queue_process; | |
2360 | zebra->ribq->spec.errorfunc = NULL; | |
2361 | zebra->ribq->spec.completion_func = &meta_queue_process_complete; | |
2362 | /* XXX: TODO: These should be runtime configurable via vty */ | |
2363 | zebra->ribq->spec.max_retries = 3; | |
3a30f50f | 2364 | zebra->ribq->spec.hold = ZEBRA_RIB_PROCESS_HOLD_TIME; |
d62a17ae | 2365 | |
2366 | if (!(zebra->mq = meta_queue_new())) { | |
e914ccbe | 2367 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2368 | "%s: could not initialise meta queue!", __func__); |
d62a17ae | 2369 | return; |
2370 | } | |
2371 | return; | |
718e3744 | 2372 | } |
2373 | ||
6d691129 PJ |
2374 | /* RIB updates are processed via a queue of pointers to route_nodes. |
2375 | * | |
2376 | * The queue length is bounded by the maximal size of the routing table, | |
2377 | * as a route_node will not be requeued, if already queued. | |
2378 | * | |
f0f77c9a DS |
2379 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
2380 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 2381 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 2382 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 2383 | * |
f0f77c9a | 2384 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
2385 | * |
2386 | * rib_addnode | |
f0f77c9a DS |
2387 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
2388 | * |-------->| | best RE, if required | |
3c0755dc PJ |
2389 | * | | |
2390 | * static_install->|->rib_addqueue...... -> rib_process | |
2391 | * | | | |
2392 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
2393 | * |-> set ROUTE_ENTRY_REMOVE | |
2394 | * rib_delnode (RE freed) | |
6d691129 | 2395 | * |
9fd92e3c AS |
2396 | * The 'info' pointer of a route_node points to a rib_dest_t |
2397 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
2398 | * dest is created on-demand by rib_link() and is kept around at least | |
2399 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 2400 | * |
6d691129 PJ |
2401 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
2402 | * | |
2403 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
2404 | * - dest attached to route_node: |
2405 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
2406 | * - route_node processing queue |
2407 | * - managed by: rib_addqueue, rib_process. | |
2408 | * | |
2409 | */ | |
d62a17ae | 2410 | |
f0f77c9a | 2411 | /* Add RE to head of the route node. */ |
d62a17ae | 2412 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
2413 | { | |
2414 | struct route_entry *head; | |
2415 | rib_dest_t *dest; | |
2416 | afi_t afi; | |
2417 | const char *rmap_name; | |
9fd92e3c | 2418 | |
d62a17ae | 2419 | assert(re && rn); |
9fd92e3c | 2420 | |
d62a17ae | 2421 | dest = rib_dest_from_rnode(rn); |
2422 | if (!dest) { | |
2423 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2424 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 2425 | |
d62a17ae | 2426 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); |
2427 | route_lock_node(rn); /* rn route table reference */ | |
2428 | rn->info = dest; | |
2429 | dest->rnode = rn; | |
2430 | } | |
2263a412 | 2431 | |
d62a17ae | 2432 | head = dest->routes; |
2433 | if (head) { | |
2434 | head->prev = re; | |
2435 | } | |
2436 | re->next = head; | |
2437 | dest->routes = re; | |
2438 | ||
2439 | afi = (rn->p.family == AF_INET) | |
2440 | ? AFI_IP | |
2441 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2442 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2443 | rmap_name = zebra_get_import_table_route_map(afi, re->table); | |
2444 | zebra_add_import_table_entry(rn, re, rmap_name); | |
2445 | } else if (process) | |
2446 | rib_queue_add(rn); | |
2447 | } | |
2448 | ||
7e24fdf3 DS |
2449 | static void rib_addnode(struct route_node *rn, |
2450 | struct route_entry *re, int process) | |
d62a17ae | 2451 | { |
2452 | /* RE node has been un-removed before route-node is processed. | |
2453 | * route_node must hence already be on the queue for processing.. | |
2454 | */ | |
2455 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
2456 | if (IS_ZEBRA_DEBUG_RIB) | |
2457 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
2458 | (void *)rn, (void *)re); | |
2459 | ||
2460 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2461 | return; | |
2462 | } | |
2463 | rib_link(rn, re, process); | |
6d691129 PJ |
2464 | } |
2465 | ||
9fd92e3c AS |
2466 | /* |
2467 | * rib_unlink | |
2468 | * | |
2469 | * Detach a rib structure from a route_node. | |
2470 | * | |
2471 | * Note that a call to rib_unlink() should be followed by a call to | |
2472 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
2473 | * longer required to be deleted. | |
2474 | */ | |
d62a17ae | 2475 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 2476 | { |
d62a17ae | 2477 | rib_dest_t *dest; |
9fd92e3c | 2478 | |
d62a17ae | 2479 | assert(rn && re); |
6d691129 | 2480 | |
d62a17ae | 2481 | if (IS_ZEBRA_DEBUG_RIB) |
2482 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
2483 | (void *)re); | |
6d691129 | 2484 | |
d62a17ae | 2485 | dest = rib_dest_from_rnode(rn); |
6d691129 | 2486 | |
d62a17ae | 2487 | if (re->next) |
2488 | re->next->prev = re->prev; | |
6d691129 | 2489 | |
d62a17ae | 2490 | if (re->prev) |
2491 | re->prev->next = re->next; | |
2492 | else { | |
2493 | dest->routes = re->next; | |
2494 | } | |
7a4bb9c5 | 2495 | |
2eb07de3 DS |
2496 | if (dest->selected_fib == re) |
2497 | dest->selected_fib = NULL; | |
2498 | ||
7ee30f28 | 2499 | nexthops_free(re->ng.nexthop); |
d62a17ae | 2500 | XFREE(MTYPE_RE, re); |
2501 | } | |
2502 | ||
2503 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2504 | { | |
2505 | afi_t afi; | |
2506 | ||
2507 | if (IS_ZEBRA_DEBUG_RIB) | |
2508 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2509 | (void *)rn, (void *)re); | |
2510 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2511 | ||
2512 | afi = (rn->p.family == AF_INET) | |
2513 | ? AFI_IP | |
2514 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2515 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2516 | zebra_del_import_table_entry(rn, re); | |
2517 | /* Just clean up if non main table */ | |
2518 | if (IS_ZEBRA_DEBUG_RIB) { | |
2519 | char buf[SRCDEST2STR_BUFFER]; | |
2520 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2521 | zlog_debug( | |
2522 | "%u:%s: Freeing route rn %p, re %p (type %d)", | |
2523 | re->vrf_id, buf, rn, re, re->type); | |
2524 | } | |
7a4bb9c5 | 2525 | |
d62a17ae | 2526 | rib_unlink(rn, re); |
2527 | } else { | |
2528 | rib_queue_add(rn); | |
2529 | } | |
718e3744 | 2530 | } |
2531 | ||
f0f77c9a | 2532 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2533 | * standard debug log. Calling function name and IP prefix in |
2534 | * question are passed as 1st and 2nd arguments. | |
2535 | */ | |
2536 | ||
d62a17ae | 2537 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2538 | union prefixconstptr src_pp, | |
2539 | const struct route_entry *re) | |
2540 | { | |
d62a17ae | 2541 | const struct prefix *src_p = src_pp.p; |
2542 | bool is_srcdst = src_p && src_p->prefixlen; | |
2543 | char straddr[PREFIX_STRLEN]; | |
2544 | char srcaddr[PREFIX_STRLEN]; | |
2545 | struct nexthop *nexthop; | |
2546 | ||
2547 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2548 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2549 | is_srcdst ? " from " : "", | |
2550 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2551 | : "", | |
2552 | re->vrf_id); | |
cc54cfee RW |
2553 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", |
2554 | func, (unsigned long)re->uptime, re->type, re->instance, | |
2555 | re->table); | |
d62a17ae | 2556 | zlog_debug( |
2557 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
2558 | func, re->metric, re->mtu, re->distance, re->flags, re->status); | |
2559 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func, | |
2560 | re->nexthop_num, re->nexthop_active_num); | |
2561 | ||
7ee30f28 | 2562 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
2d68a0f2 DS |
2563 | struct interface *ifp; |
2564 | struct vrf *vrf = vrf_lookup_by_id(nexthop->vrf_id); | |
2565 | ||
2566 | switch (nexthop->type) { | |
2567 | case NEXTHOP_TYPE_BLACKHOLE: | |
2568 | sprintf(straddr, "Blackhole"); | |
2569 | break; | |
2570 | case NEXTHOP_TYPE_IFINDEX: | |
2571 | ifp = if_lookup_by_index(nexthop->ifindex, | |
2572 | nexthop->vrf_id); | |
2573 | sprintf(straddr, "%s", ifp ? ifp->name : "Unknown"); | |
2574 | break; | |
2575 | case NEXTHOP_TYPE_IPV4: | |
2576 | /* fallthrough */ | |
2577 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
2578 | inet_ntop(AF_INET, &nexthop->gate, straddr, | |
2579 | INET6_ADDRSTRLEN); | |
2580 | break; | |
2581 | case NEXTHOP_TYPE_IPV6: | |
2582 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
2583 | inet_ntop(AF_INET6, &nexthop->gate, straddr, | |
2584 | INET6_ADDRSTRLEN); | |
2585 | break; | |
2586 | } | |
2587 | zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s", func, | |
d62a17ae | 2588 | (nexthop->rparent ? " NH" : "NH"), straddr, |
2d68a0f2 DS |
2589 | nexthop->ifindex, vrf ? vrf->name : "Unknown", |
2590 | nexthop->vrf_id, | |
d62a17ae | 2591 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) |
2592 | ? "ACTIVE " | |
2593 | : ""), | |
2594 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB) | |
2595 | ? "FIB " | |
2596 | : ""), | |
2597 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
2598 | ? "RECURSIVE" | |
2599 | : "")); | |
2600 | } | |
2601 | zlog_debug("%s: dump complete", func); | |
dc95824a DO |
2602 | } |
2603 | ||
2604 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2605 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2606 | */ |
2607 | ||
d62a17ae | 2608 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2609 | { | |
2610 | struct route_table *table; | |
2611 | struct route_node *rn; | |
2612 | struct route_entry *re; | |
2613 | char prefix_buf[INET_ADDRSTRLEN]; | |
2614 | ||
2615 | /* Lookup table. */ | |
2616 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2617 | if (!table) { | |
e914ccbe | 2618 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2619 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2620 | vrf_id); | |
d62a17ae | 2621 | return; |
2622 | } | |
2623 | ||
2624 | /* Scan the RIB table for exactly matching RE entry. */ | |
2625 | rn = route_node_lookup(table, (struct prefix *)p); | |
2626 | ||
2627 | /* No route for this prefix. */ | |
2628 | if (!rn) { | |
32391aff | 2629 | zlog_debug("%s:%u lookup failed for %s", __func__, vrf_id, |
d62a17ae | 2630 | prefix2str((struct prefix *)p, prefix_buf, |
2631 | sizeof(prefix_buf))); | |
2632 | return; | |
2633 | } | |
2634 | ||
2635 | /* Unlock node. */ | |
2636 | route_unlock_node(rn); | |
2637 | ||
2638 | /* let's go */ | |
a2addae8 | 2639 | RNODE_FOREACH_RE (rn, re) { |
32391aff DS |
2640 | zlog_debug("%s:%u rn %p, re %p: %s, %s", |
2641 | __func__, vrf_id, | |
2642 | (void *)rn, (void *)re, | |
d62a17ae | 2643 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) |
2644 | ? "removed" | |
2645 | : "NOT removed"), | |
2646 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2647 | ? "selected" | |
2648 | : "NOT selected")); | |
2649 | route_entry_dump(p, NULL, re); | |
2650 | } | |
dc95824a DO |
2651 | } |
2652 | ||
20e5ff0a DO |
2653 | /* Check if requested address assignment will fail due to another |
2654 | * route being installed by zebra in FIB already. Take necessary | |
2655 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2656 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2657 | */ |
d62a17ae | 2658 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2659 | { | |
2660 | struct route_table *table; | |
2661 | struct route_node *rn; | |
d62a17ae | 2662 | unsigned changed = 0; |
5f7a4718 | 2663 | rib_dest_t *dest; |
d62a17ae | 2664 | |
2665 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
e914ccbe | 2666 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2667 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2668 | vrf_id); | |
d62a17ae | 2669 | return; |
2670 | } | |
2671 | ||
2672 | /* No matches would be the simplest case. */ | |
2673 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2674 | return; | |
2675 | ||
2676 | /* Unlock node. */ | |
2677 | route_unlock_node(rn); | |
2678 | ||
5f7a4718 | 2679 | dest = rib_dest_from_rnode(rn); |
d62a17ae | 2680 | /* Check all RE entries. In case any changes have to be done, requeue |
2681 | * the RN into RIBQ head. If the routing message about the new connected | |
2682 | * route (generated by the IP address we are going to assign very soon) | |
2683 | * comes before the RIBQ is processed, the new RE entry will join | |
2684 | * RIBQ record already on head. This is necessary for proper | |
2685 | * revalidation | |
2686 | * of the rest of the RE. | |
2687 | */ | |
5f7a4718 DS |
2688 | if (dest->selected_fib && !RIB_SYSTEM_ROUTE(dest->selected_fib)) { |
2689 | changed = 1; | |
2690 | if (IS_ZEBRA_DEBUG_RIB) { | |
2691 | char buf[PREFIX_STRLEN]; | |
2692 | ||
2693 | zlog_debug("%u:%s: freeing way for connected prefix", | |
2694 | dest->selected_fib->vrf_id, | |
2695 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2696 | route_entry_dump(&rn->p, NULL, dest->selected_fib); | |
d62a17ae | 2697 | } |
5f7a4718 | 2698 | rib_uninstall(rn, dest->selected_fib); |
d62a17ae | 2699 | } |
2700 | if (changed) | |
2701 | rib_queue_add(rn); | |
20e5ff0a DO |
2702 | } |
2703 | ||
d62a17ae | 2704 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
2705 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
718e3744 | 2706 | { |
d62a17ae | 2707 | struct route_table *table; |
2708 | struct route_node *rn; | |
40ecd8e4 | 2709 | struct route_entry *same = NULL; |
d62a17ae | 2710 | struct nexthop *nexthop; |
2711 | int ret = 0; | |
b4c034b0 | 2712 | |
d62a17ae | 2713 | if (!re) |
2714 | return 0; | |
b4c034b0 | 2715 | |
1f610a1f | 2716 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
05737783 | 2717 | |
d62a17ae | 2718 | /* Lookup table. */ |
7865c65d RW |
2719 | table = zebra_vrf_table_with_table_id(afi, safi, re->vrf_id, re->table); |
2720 | if (!table) { | |
2721 | XFREE(MTYPE_RE, re); | |
d62a17ae | 2722 | return 0; |
7865c65d | 2723 | } |
cddf391b | 2724 | |
d62a17ae | 2725 | /* Make it sure prefixlen is applied to the prefix. */ |
2726 | apply_mask(p); | |
2727 | if (src_p) | |
2728 | apply_mask_ipv6(src_p); | |
718e3744 | 2729 | |
d62a17ae | 2730 | /* Set default distance by route type. */ |
2731 | if (re->distance == 0) { | |
0492eea0 | 2732 | re->distance = route_distance(re->type); |
718e3744 | 2733 | |
d62a17ae | 2734 | /* iBGP distance is 200. */ |
2735 | if (re->type == ZEBRA_ROUTE_BGP | |
2736 | && CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
2737 | re->distance = 200; | |
2738 | } | |
718e3744 | 2739 | |
d62a17ae | 2740 | /* Lookup route node.*/ |
2741 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2742 | |
40ecd8e4 DS |
2743 | /* |
2744 | * If same type of route are installed, treat it as a implicit | |
2745 | * withdraw. | |
2746 | * If the user has specified the No route replace semantics | |
2747 | * for the install don't do a route replace. | |
2748 | */ | |
a2addae8 | 2749 | RNODE_FOREACH_RE (rn, same) { |
d62a17ae | 2750 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) |
2751 | continue; | |
41ec9222 | 2752 | |
eb327fa5 RW |
2753 | if (same->type != re->type) |
2754 | continue; | |
2755 | if (same->instance != re->instance) | |
2756 | continue; | |
996c9314 LB |
2757 | if (same->type == ZEBRA_ROUTE_KERNEL |
2758 | && same->metric != re->metric) | |
eb327fa5 | 2759 | continue; |
40ecd8e4 DS |
2760 | |
2761 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && | |
2762 | same->distance != re->distance) | |
2763 | continue; | |
2764 | ||
844b3a87 | 2765 | /* |
40ecd8e4 DS |
2766 | * We should allow duplicate connected routes |
2767 | * because of IPv6 link-local routes and unnumbered | |
2768 | * interfaces on Linux. | |
844b3a87 RW |
2769 | */ |
2770 | if (same->type != ZEBRA_ROUTE_CONNECT) | |
d62a17ae | 2771 | break; |
2772 | } | |
718e3744 | 2773 | |
d62a17ae | 2774 | /* If this route is kernel route, set FIB flag to the route. */ |
8628fc61 | 2775 | if (RIB_SYSTEM_ROUTE(re)) |
7ee30f28 | 2776 | for (nexthop = re->ng.nexthop; nexthop; nexthop = nexthop->next) |
d62a17ae | 2777 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
718e3744 | 2778 | |
d62a17ae | 2779 | /* Link new re to node.*/ |
2780 | if (IS_ZEBRA_DEBUG_RIB) { | |
2781 | rnode_debug( | |
2782 | rn, re->vrf_id, | |
2783 | "Inserting route rn %p, re %p (type %d) existing %p", | |
2784 | (void *)rn, (void *)re, re->type, (void *)same); | |
718e3744 | 2785 | |
d62a17ae | 2786 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2787 | route_entry_dump(p, src_p, re); | |
718e3744 | 2788 | } |
d62a17ae | 2789 | rib_addnode(rn, re, 1); |
2790 | ret = 1; | |
6b0655a2 | 2791 | |
d62a17ae | 2792 | /* Free implicit route.*/ |
2793 | if (same) { | |
2794 | rib_delnode(rn, same); | |
2795 | ret = -1; | |
2796 | } | |
718e3744 | 2797 | |
d62a17ae | 2798 | route_unlock_node(rn); |
2799 | return ret; | |
2800 | } | |
2801 | ||
2802 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
d7c0a89a | 2803 | unsigned short instance, int flags, struct prefix *p, |
fd36be7e | 2804 | struct prefix_ipv6 *src_p, const struct nexthop *nh, |
40ecd8e4 DS |
2805 | uint32_t table_id, uint32_t metric, uint8_t distance, |
2806 | bool fromkernel) | |
d62a17ae | 2807 | { |
2808 | struct route_table *table; | |
2809 | struct route_node *rn; | |
2810 | struct route_entry *re; | |
2811 | struct route_entry *fib = NULL; | |
2812 | struct route_entry *same = NULL; | |
fd36be7e | 2813 | struct nexthop *rtnh; |
d62a17ae | 2814 | char buf2[INET6_ADDRSTRLEN]; |
5f7a4718 | 2815 | rib_dest_t *dest; |
d62a17ae | 2816 | |
1f610a1f | 2817 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
d62a17ae | 2818 | |
2819 | /* Lookup table. */ | |
2820 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2821 | if (!table) | |
2822 | return; | |
2823 | ||
2824 | /* Apply mask. */ | |
2825 | apply_mask(p); | |
2826 | if (src_p) | |
2827 | apply_mask_ipv6(src_p); | |
2828 | ||
2829 | /* Lookup route node. */ | |
2830 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2831 | if (!rn) { | |
2832 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2833 | ||
2834 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2835 | if (src_p && src_p->prefixlen) | |
2836 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2837 | else | |
2838 | src_buf[0] = '\0'; | |
2839 | ||
2840 | if (IS_ZEBRA_DEBUG_RIB) | |
2841 | zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id, | |
2842 | dst_buf, | |
2843 | (src_buf[0] != '\0') ? " from " : "", | |
2844 | src_buf); | |
2845 | return; | |
2846 | } | |
718e3744 | 2847 | |
5f7a4718 DS |
2848 | dest = rib_dest_from_rnode(rn); |
2849 | fib = dest->selected_fib; | |
2850 | ||
d62a17ae | 2851 | /* Lookup same type route. */ |
a2addae8 | 2852 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2853 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2854 | continue; | |
2855 | ||
d62a17ae | 2856 | if (re->type != type) |
2857 | continue; | |
2858 | if (re->instance != instance) | |
2859 | continue; | |
40ecd8e4 DS |
2860 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && |
2861 | distance != re->distance) | |
2862 | continue; | |
2863 | ||
996c9314 | 2864 | if (re->type == ZEBRA_ROUTE_KERNEL && re->metric != metric) |
f19435a8 | 2865 | continue; |
7ee30f28 | 2866 | if (re->type == ZEBRA_ROUTE_CONNECT && (rtnh = re->ng.nexthop) |
fd36be7e DL |
2867 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { |
2868 | if (rtnh->ifindex != nh->ifindex) | |
d62a17ae | 2869 | continue; |
d62a17ae | 2870 | same = re; |
2871 | break; | |
2872 | } | |
2873 | /* Make sure that the route found has the same gateway. */ | |
2874 | else { | |
fd36be7e | 2875 | if (nh == NULL) { |
d62a17ae | 2876 | same = re; |
2877 | break; | |
2878 | } | |
7ee30f28 | 2879 | for (ALL_NEXTHOPS(re->ng, rtnh)) |
fd36be7e | 2880 | if (nexthop_same_no_recurse(rtnh, nh)) { |
d62a17ae | 2881 | same = re; |
2882 | break; | |
2883 | } | |
2884 | if (same) | |
2885 | break; | |
2886 | } | |
2887 | } | |
2888 | /* If same type of route can't be found and this message is from | |
2889 | kernel. */ | |
2890 | if (!same) { | |
5dfeba19 DS |
2891 | /* |
2892 | * In the past(HA!) we could get here because | |
2893 | * we were receiving a route delete from the | |
2894 | * kernel and we're not marking the proto | |
2895 | * as coming from it's appropriate originator. | |
2896 | * Now that we are properly noticing the fact | |
2897 | * that the kernel has deleted our route we | |
2898 | * are not going to get called in this path | |
2899 | * I am going to leave this here because | |
2900 | * this might still work this way on non-linux | |
2901 | * platforms as well as some weird state I have | |
2902 | * not properly thought of yet. | |
2903 | * If we can show that this code path is | |
2904 | * dead then we can remove it. | |
2905 | */ | |
b8faa875 | 2906 | if (fib && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { |
d62a17ae | 2907 | if (IS_ZEBRA_DEBUG_RIB) { |
2908 | rnode_debug( | |
2909 | rn, vrf_id, | |
2910 | "rn %p, re %p (type %d) was deleted from kernel, adding", | |
2911 | rn, fib, fib->type); | |
2912 | } | |
2913 | if (allow_delete) { | |
2914 | /* Unset flags. */ | |
7ee30f28 | 2915 | for (rtnh = fib->ng.nexthop; rtnh; |
fd36be7e DL |
2916 | rtnh = rtnh->next) |
2917 | UNSET_FLAG(rtnh->flags, | |
d62a17ae | 2918 | NEXTHOP_FLAG_FIB); |
2919 | ||
ed216282 DS |
2920 | /* |
2921 | * This is a non FRR route | |
2922 | * as such we should mark | |
2923 | * it as deleted | |
2924 | */ | |
5f7a4718 | 2925 | dest->selected_fib = NULL; |
d62a17ae | 2926 | } else { |
2927 | /* This means someone else, other than Zebra, | |
2928 | * has deleted | |
2929 | * a Zebra router from the kernel. We will add | |
2930 | * it back */ | |
2931 | rib_install_kernel(rn, fib, NULL); | |
2932 | } | |
2933 | } else { | |
2934 | if (IS_ZEBRA_DEBUG_RIB) { | |
fd36be7e | 2935 | if (nh) |
d62a17ae | 2936 | rnode_debug( |
2937 | rn, vrf_id, | |
2938 | "via %s ifindex %d type %d " | |
2939 | "doesn't exist in rib", | |
36228974 | 2940 | inet_ntop(afi2family(afi), |
2941 | &nh->gate, buf2, | |
2942 | sizeof(buf2)), | |
2943 | nh->ifindex, type); | |
d62a17ae | 2944 | else |
2945 | rnode_debug( | |
2946 | rn, vrf_id, | |
fd36be7e DL |
2947 | "type %d doesn't exist in rib", |
2948 | type); | |
d62a17ae | 2949 | } |
2950 | route_unlock_node(rn); | |
2951 | return; | |
2952 | } | |
2953 | } | |
718e3744 | 2954 | |
5dfeba19 | 2955 | if (same) { |
996c9314 LB |
2956 | if (fromkernel && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE) |
2957 | && !allow_delete) { | |
5dfeba19 DS |
2958 | rib_install_kernel(rn, same, NULL); |
2959 | route_unlock_node(rn); | |
2960 | ||
2961 | return; | |
2962 | } | |
6134fd82 | 2963 | |
90264d64 | 2964 | if (CHECK_FLAG(flags, ZEBRA_FLAG_EVPN_ROUTE)) { |
6134fd82 | 2965 | struct nexthop *tmp_nh; |
2966 | ||
7ee30f28 | 2967 | for (ALL_NEXTHOPS(re->ng, tmp_nh)) { |
6134fd82 | 2968 | struct ipaddr vtep_ip; |
2969 | ||
2970 | memset(&vtep_ip, 0, sizeof(struct ipaddr)); | |
1ec31309 | 2971 | if (afi == AFI_IP) { |
2972 | vtep_ip.ipa_type = IPADDR_V4; | |
2973 | memcpy(&(vtep_ip.ipaddr_v4), | |
2974 | &(tmp_nh->gate.ipv4), | |
2975 | sizeof(struct in_addr)); | |
2976 | } else { | |
2977 | vtep_ip.ipa_type = IPADDR_V6; | |
2978 | memcpy(&(vtep_ip.ipaddr_v6), | |
2979 | &(tmp_nh->gate.ipv6), | |
2980 | sizeof(struct in6_addr)); | |
2981 | } | |
a317a9b9 | 2982 | zebra_vxlan_evpn_vrf_route_del(re->vrf_id, |
6134fd82 | 2983 | &vtep_ip, p); |
2984 | } | |
2985 | } | |
d62a17ae | 2986 | rib_delnode(rn, same); |
5dfeba19 | 2987 | } |
05737783 | 2988 | |
d62a17ae | 2989 | route_unlock_node(rn); |
2990 | return; | |
2991 | } | |
718e3744 | 2992 | |
718e3744 | 2993 | |
d7c0a89a QY |
2994 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, |
2995 | unsigned short instance, int flags, struct prefix *p, | |
2996 | struct prefix_ipv6 *src_p, const struct nexthop *nh, | |
2997 | uint32_t table_id, uint32_t metric, uint32_t mtu, uint8_t distance, | |
2998 | route_tag_t tag) | |
d62a17ae | 2999 | { |
3000 | struct route_entry *re; | |
66af6845 | 3001 | struct nexthop *nexthop; |
718e3744 | 3002 | |
66af6845 | 3003 | /* Allocate new route_entry structure. */ |
d62a17ae | 3004 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); |
d62a17ae | 3005 | re->type = type; |
3006 | re->instance = instance; | |
3007 | re->distance = distance; | |
3008 | re->flags = flags; | |
3009 | re->metric = metric; | |
3010 | re->mtu = mtu; | |
3011 | re->table = table_id; | |
3012 | re->vrf_id = vrf_id; | |
3013 | re->nexthop_num = 0; | |
3014 | re->uptime = time(NULL); | |
4e40b6d6 | 3015 | re->tag = tag; |
d62a17ae | 3016 | |
66af6845 RW |
3017 | /* Add nexthop. */ |
3018 | nexthop = nexthop_new(); | |
3019 | *nexthop = *nh; | |
3020 | route_entry_nexthop_add(re, nexthop); | |
718e3744 | 3021 | |
66af6845 | 3022 | return rib_add_multipath(afi, safi, p, src_p, re); |
718e3744 | 3023 | } |
3024 | ||
1c848137 | 3025 | /* Schedule routes of a particular table (address-family) based on event. */ |
d5b8c216 | 3026 | void rib_update_table(struct route_table *table, rib_update_event_t event) |
d62a17ae | 3027 | { |
3028 | struct route_node *rn; | |
3029 | struct route_entry *re, *next; | |
3030 | ||
3031 | /* Walk all routes and queue for processing, if appropriate for | |
3032 | * the trigger event. | |
3033 | */ | |
3034 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
1ca60f2c DS |
3035 | /* |
3036 | * If we are looking at a route node and the node | |
3037 | * has already been queued we don't | |
3038 | * need to queue it up again | |
3039 | */ | |
996c9314 LB |
3040 | if (rn->info && CHECK_FLAG(rib_dest_from_rnode(rn)->flags, |
3041 | RIB_ROUTE_ANY_QUEUED)) | |
1ca60f2c | 3042 | continue; |
d62a17ae | 3043 | switch (event) { |
3044 | case RIB_UPDATE_IF_CHANGE: | |
3045 | /* Examine all routes that won't get processed by the | |
3046 | * protocol or | |
3047 | * triggered by nexthop evaluation (NHT). This would be | |
3048 | * system, | |
3049 | * kernel and certain static routes. Note that NHT will | |
3050 | * get | |
3051 | * triggered upon an interface event as connected routes | |
3052 | * always | |
3053 | * get queued for processing. | |
3054 | */ | |
a2addae8 | 3055 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
0a16efff DS |
3056 | struct nexthop *nh; |
3057 | ||
996c9314 LB |
3058 | if (re->type != ZEBRA_ROUTE_SYSTEM |
3059 | && re->type != ZEBRA_ROUTE_KERNEL | |
3060 | && re->type != ZEBRA_ROUTE_CONNECT | |
3061 | && re->type != ZEBRA_ROUTE_STATIC) | |
0a16efff DS |
3062 | continue; |
3063 | ||
3064 | if (re->type != ZEBRA_ROUTE_STATIC) { | |
3065 | rib_queue_add(rn); | |
3066 | continue; | |
3067 | } | |
3068 | ||
7ee30f28 | 3069 | for (nh = re->ng.nexthop; nh; nh = nh->next) |
0a16efff DS |
3070 | if (!(nh->type == NEXTHOP_TYPE_IPV4 |
3071 | || nh->type == NEXTHOP_TYPE_IPV6)) | |
3072 | break; | |
3073 | ||
3074 | /* If we only have nexthops to a | |
3075 | * gateway, NHT will | |
3076 | * take care. | |
3077 | */ | |
3078 | if (nh) | |
d62a17ae | 3079 | rib_queue_add(rn); |
3080 | } | |
3081 | break; | |
3082 | ||
3083 | case RIB_UPDATE_RMAP_CHANGE: | |
3084 | case RIB_UPDATE_OTHER: | |
3085 | /* Right now, examine all routes. Can restrict to a | |
3086 | * protocol in | |
3087 | * some cases (TODO). | |
3088 | */ | |
3089 | if (rnode_to_ribs(rn)) | |
3090 | rib_queue_add(rn); | |
3091 | break; | |
3092 | ||
3093 | default: | |
3094 | break; | |
3095 | } | |
3096 | } | |
b84c7253 | 3097 | } |
3098 | ||
718e3744 | 3099 | /* RIB update function. */ |
d62a17ae | 3100 | void rib_update(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 3101 | { |
d62a17ae | 3102 | struct route_table *table; |
1c848137 | 3103 | |
d62a17ae | 3104 | /* Process routes of interested address-families. */ |
3105 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
d5b8c216 | 3106 | if (table) { |
3107 | if (IS_ZEBRA_DEBUG_EVENT) | |
3108 | zlog_debug("%s : AFI_IP event %d", __func__, event); | |
d62a17ae | 3109 | rib_update_table(table, event); |
d5b8c216 | 3110 | } |
718e3744 | 3111 | |
d62a17ae | 3112 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
d5b8c216 | 3113 | if (table) { |
3114 | if (IS_ZEBRA_DEBUG_EVENT) | |
3115 | zlog_debug("%s : AFI_IP6 event %d", __func__, event); | |
d62a17ae | 3116 | rib_update_table(table, event); |
d5b8c216 | 3117 | } |
718e3744 | 3118 | } |
3119 | ||
718e3744 | 3120 | /* Delete self installed routes after zebra is relaunched. */ |
95a29032 | 3121 | void rib_sweep_table(struct route_table *table) |
d62a17ae | 3122 | { |
3123 | struct route_node *rn; | |
3124 | struct route_entry *re; | |
3125 | struct route_entry *next; | |
915902cb | 3126 | struct nexthop *nexthop; |
d62a17ae | 3127 | |
915902cb DS |
3128 | if (!table) |
3129 | return; | |
d62a17ae | 3130 | |
915902cb | 3131 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
a2addae8 | 3132 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
915902cb DS |
3133 | if (IS_ZEBRA_DEBUG_RIB) |
3134 | route_entry_dump(&rn->p, NULL, re); | |
3135 | ||
3136 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
3137 | continue; | |
3138 | ||
3139 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
3140 | continue; | |
3141 | ||
3142 | /* | |
3143 | * So we are starting up and have received | |
3144 | * routes from the kernel that we have installed | |
3145 | * from a previous run of zebra but not cleaned | |
3146 | * up ( say a kill -9 ) | |
3147 | * But since we haven't actually installed | |
3148 | * them yet( we received them from the kernel ) | |
3149 | * we don't think they are active. | |
3150 | * So let's pretend they are active to actually | |
3151 | * remove them. | |
3152 | * In all honesty I'm not sure if we should | |
3153 | * mark them as active when we receive them | |
3154 | * This is startup only so probably ok. | |
3155 | * | |
3156 | * If we ever decide to move rib_sweep_table | |
3157 | * to a different spot (ie startup ) | |
3158 | * this decision needs to be revisited | |
3159 | */ | |
7ee30f28 | 3160 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
915902cb DS |
3161 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
3162 | ||
0c555cc6 DS |
3163 | rib_uninstall_kernel(rn, re); |
3164 | rib_delnode(rn, re); | |
915902cb DS |
3165 | } |
3166 | } | |
718e3744 | 3167 | } |
3168 | ||
3169 | /* Sweep all RIB tables. */ | |
d62a17ae | 3170 | void rib_sweep_route(void) |
718e3744 | 3171 | { |
d62a17ae | 3172 | struct vrf *vrf; |
3173 | struct zebra_vrf *zvrf; | |
78104b9b | 3174 | |
a2addae8 | 3175 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
915902cb DS |
3176 | if ((zvrf = vrf->info) == NULL) |
3177 | continue; | |
3178 | ||
d62a17ae | 3179 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); |
3180 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3181 | } | |
95a29032 | 3182 | |
89272910 | 3183 | zebra_router_sweep_route(); |
718e3744 | 3184 | } |
2ea1ab1c VT |
3185 | |
3186 | /* Remove specific by protocol routes from 'table'. */ | |
d7c0a89a | 3187 | unsigned long rib_score_proto_table(uint8_t proto, unsigned short instance, |
47a08aa9 | 3188 | struct route_table *table) |
d62a17ae | 3189 | { |
3190 | struct route_node *rn; | |
3191 | struct route_entry *re; | |
3192 | struct route_entry *next; | |
3193 | unsigned long n = 0; | |
3194 | ||
3195 | if (table) | |
3196 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
a2addae8 | 3197 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 3198 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
3199 | continue; | |
3200 | if (re->type == proto | |
3201 | && re->instance == instance) { | |
3202 | rib_delnode(rn, re); | |
3203 | n++; | |
3204 | } | |
3205 | } | |
3206 | return n; | |
2ea1ab1c VT |
3207 | } |
3208 | ||
3209 | /* Remove specific by protocol routes. */ | |
d7c0a89a | 3210 | unsigned long rib_score_proto(uint8_t proto, unsigned short instance) |
2ea1ab1c | 3211 | { |
d62a17ae | 3212 | struct vrf *vrf; |
3213 | struct zebra_vrf *zvrf; | |
3214 | unsigned long cnt = 0; | |
78104b9b | 3215 | |
a2addae8 RW |
3216 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) |
3217 | if ((zvrf = vrf->info) != NULL) | |
3218 | cnt += rib_score_proto_table( | |
3219 | proto, instance, | |
3220 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
3221 | + rib_score_proto_table( | |
3222 | proto, instance, | |
3223 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
78104b9b | 3224 | |
89272910 | 3225 | cnt += zebra_router_score_proto(proto, instance); |
47a08aa9 | 3226 | |
d62a17ae | 3227 | return cnt; |
2ea1ab1c VT |
3228 | } |
3229 | ||
718e3744 | 3230 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 3231 | void rib_close_table(struct route_table *table) |
718e3744 | 3232 | { |
d62a17ae | 3233 | struct route_node *rn; |
1e9f448f | 3234 | rib_table_info_t *info; |
5f7a4718 | 3235 | rib_dest_t *dest; |
718e3744 | 3236 | |
1e9f448f DS |
3237 | if (!table) |
3238 | return; | |
9fd92e3c | 3239 | |
6ca30e9e | 3240 | info = route_table_get_info(table); |
5adc2528 | 3241 | |
5f7a4718 DS |
3242 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
3243 | dest = rib_dest_from_rnode(rn); | |
1e9f448f | 3244 | |
5f7a4718 | 3245 | if (dest && dest->selected_fib) { |
1e9f448f DS |
3246 | if (info->safi == SAFI_UNICAST) |
3247 | hook_call(rib_update, rn, NULL); | |
3248 | ||
258c07e4 | 3249 | if (!RIB_SYSTEM_ROUTE(dest->selected_fib)) { |
5f7a4718 | 3250 | rib_uninstall_kernel(rn, dest->selected_fib); |
258c07e4 MS |
3251 | dest->selected_fib = NULL; |
3252 | } | |
1e9f448f | 3253 | } |
5f7a4718 | 3254 | } |
718e3744 | 3255 | } |
3256 | ||
7cdb1a84 MS |
3257 | /* |
3258 | * | |
3259 | */ | |
3260 | static int rib_process_dplane_results(struct thread *thread) | |
3261 | { | |
25779064 | 3262 | struct zebra_dplane_ctx *ctx; |
7cdb1a84 MS |
3263 | |
3264 | do { | |
3265 | /* Take lock controlling queue of results */ | |
3266 | pthread_mutex_lock(&dplane_mutex); | |
3267 | { | |
3268 | /* Dequeue context block */ | |
68b375e0 | 3269 | ctx = dplane_ctx_dequeue(&rib_dplane_q); |
7cdb1a84 MS |
3270 | } |
3271 | pthread_mutex_unlock(&dplane_mutex); | |
3272 | ||
5709131c | 3273 | if (ctx) |
e5ac2adf | 3274 | rib_process_after(ctx); |
5709131c | 3275 | else |
7cdb1a84 | 3276 | break; |
7cdb1a84 | 3277 | |
5709131c | 3278 | } while (1); |
7cdb1a84 | 3279 | |
f183e380 MS |
3280 | /* Check for nexthop tracking processing after finishing with results */ |
3281 | do_nht_processing(); | |
3282 | ||
5709131c | 3283 | return 0; |
7cdb1a84 MS |
3284 | } |
3285 | ||
3286 | /* | |
3287 | * Results are returned from the dataplane subsystem, in the context of | |
1bcea841 | 3288 | * the dataplane pthread. We enqueue the results here for processing by |
7cdb1a84 MS |
3289 | * the main thread later. |
3290 | */ | |
c831033f | 3291 | static int rib_dplane_results(struct zebra_dplane_ctx *ctx) |
7cdb1a84 MS |
3292 | { |
3293 | /* Take lock controlling queue of results */ | |
3294 | pthread_mutex_lock(&dplane_mutex); | |
3295 | { | |
3296 | /* Enqueue context block */ | |
3297 | dplane_ctx_enqueue_tail(&rib_dplane_q, ctx); | |
3298 | } | |
3299 | pthread_mutex_unlock(&dplane_mutex); | |
3300 | ||
3301 | /* Ensure event is signalled to zebra main thread */ | |
3302 | thread_add_event(zebrad.master, rib_process_dplane_results, NULL, 0, | |
3303 | &t_dplane); | |
3304 | ||
5709131c | 3305 | return 0; |
7cdb1a84 MS |
3306 | } |
3307 | ||
718e3744 | 3308 | /* Routing information base initialize. */ |
d62a17ae | 3309 | void rib_init(void) |
718e3744 | 3310 | { |
d62a17ae | 3311 | rib_queue_init(&zebrad); |
7cdb1a84 MS |
3312 | |
3313 | /* Init dataplane, and register for results */ | |
3314 | pthread_mutex_init(&dplane_mutex, NULL); | |
3315 | TAILQ_INIT(&rib_dplane_q); | |
3316 | zebra_dplane_init(); | |
3317 | dplane_results_register(rib_dplane_results); | |
718e3744 | 3318 | } |
0915bb0c AS |
3319 | |
3320 | /* | |
3321 | * vrf_id_get_next | |
3322 | * | |
3323 | * Get the first vrf id that is greater than the given vrf id if any. | |
3324 | * | |
3325 | * Returns TRUE if a vrf id was found, FALSE otherwise. | |
3326 | */ | |
d62a17ae | 3327 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 3328 | { |
d62a17ae | 3329 | struct vrf *vrf; |
b72ede27 | 3330 | |
d62a17ae | 3331 | vrf = vrf_lookup_by_id(vrf_id); |
3332 | if (vrf) { | |
3333 | vrf = RB_NEXT(vrf_id_head, vrf); | |
3334 | if (vrf) { | |
3335 | *next_id_p = vrf->vrf_id; | |
3336 | return 1; | |
3337 | } | |
3338 | } | |
0915bb0c | 3339 | |
d62a17ae | 3340 | return 0; |
0915bb0c AS |
3341 | } |
3342 | ||
3343 | /* | |
3344 | * rib_tables_iter_next | |
3345 | * | |
3346 | * Returns the next table in the iteration. | |
3347 | */ | |
d62a17ae | 3348 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
3349 | { | |
3350 | struct route_table *table; | |
3351 | ||
3352 | /* | |
3353 | * Array that helps us go over all AFI/SAFI combinations via one | |
3354 | * index. | |
3355 | */ | |
3356 | static struct { | |
3357 | afi_t afi; | |
3358 | safi_t safi; | |
3359 | } afi_safis[] = { | |
3360 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
3361 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
3362 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
3363 | }; | |
3364 | ||
3365 | table = NULL; | |
3366 | ||
3367 | switch (iter->state) { | |
3368 | ||
3369 | case RIB_TABLES_ITER_S_INIT: | |
3370 | iter->vrf_id = VRF_DEFAULT; | |
3371 | iter->afi_safi_ix = -1; | |
3372 | ||
3373 | /* Fall through */ | |
3374 | ||
3375 | case RIB_TABLES_ITER_S_ITERATING: | |
3376 | iter->afi_safi_ix++; | |
3377 | while (1) { | |
3378 | ||
3379 | while (iter->afi_safi_ix | |
3380 | < (int)ZEBRA_NUM_OF(afi_safis)) { | |
3381 | table = zebra_vrf_table( | |
3382 | afi_safis[iter->afi_safi_ix].afi, | |
3383 | afi_safis[iter->afi_safi_ix].safi, | |
3384 | iter->vrf_id); | |
3385 | if (table) | |
3386 | break; | |
3387 | ||
3388 | iter->afi_safi_ix++; | |
3389 | } | |
3390 | ||
3391 | /* | |
3392 | * Found another table in this vrf. | |
3393 | */ | |
3394 | if (table) | |
3395 | break; | |
3396 | ||
3397 | /* | |
3398 | * Done with all tables in the current vrf, go to the | |
3399 | * next | |
3400 | * one. | |
3401 | */ | |
3402 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
3403 | break; | |
3404 | ||
3405 | iter->afi_safi_ix = 0; | |
3406 | } | |
0915bb0c | 3407 | |
0915bb0c AS |
3408 | break; |
3409 | ||
d62a17ae | 3410 | case RIB_TABLES_ITER_S_DONE: |
3411 | return NULL; | |
0915bb0c AS |
3412 | } |
3413 | ||
d62a17ae | 3414 | if (table) |
3415 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
3416 | else | |
3417 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 3418 | |
d62a17ae | 3419 | return table; |
0915bb0c | 3420 | } |