]>
Commit | Line | Data |
---|---|---|
718e3744 | 1 | /* Routing Information Base. |
2 | * Copyright (C) 1997, 98, 99, 2001 Kunihiro Ishiguro | |
3 | * | |
4 | * This file is part of GNU Zebra. | |
5 | * | |
6 | * GNU Zebra is free software; you can redistribute it and/or modify it | |
7 | * under the terms of the GNU General Public License as published by the | |
8 | * Free Software Foundation; either version 2, or (at your option) any | |
9 | * later version. | |
10 | * | |
11 | * GNU Zebra is distributed in the hope that it will be useful, but | |
12 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | * General Public License for more details. | |
15 | * | |
896014f4 DL |
16 | * You should have received a copy of the GNU General Public License along |
17 | * with this program; see the file COPYING; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
718e3744 | 19 | */ |
20 | ||
21 | #include <zebra.h> | |
22 | ||
718e3744 | 23 | #include "command.h" |
43e52561 QY |
24 | #include "if.h" |
25 | #include "linklist.h" | |
718e3744 | 26 | #include "log.h" |
43e52561 QY |
27 | #include "memory.h" |
28 | #include "mpls.h" | |
29 | #include "nexthop.h" | |
30 | #include "prefix.h" | |
7514fb77 PJ |
31 | #include "prefix.h" |
32 | #include "routemap.h" | |
43e52561 | 33 | #include "sockunion.h" |
05737783 | 34 | #include "srcdest_table.h" |
43e52561 QY |
35 | #include "table.h" |
36 | #include "thread.h" | |
37 | #include "vrf.h" | |
38 | #include "workqueue.h" | |
718e3744 | 39 | |
89272910 | 40 | #include "zebra/zebra_router.h" |
43e52561 QY |
41 | #include "zebra/connected.h" |
42 | #include "zebra/debug.h" | |
43 | #include "zebra/interface.h" | |
44 | #include "zebra/redistribute.h" | |
718e3744 | 45 | #include "zebra/rib.h" |
46 | #include "zebra/rt.h" | |
43e52561 QY |
47 | #include "zebra/zapi_msg.h" |
48 | #include "zebra/zebra_errors.h" | |
49 | #include "zebra/zebra_memory.h" | |
7c551956 | 50 | #include "zebra/zebra_ns.h" |
fb018d25 | 51 | #include "zebra/zebra_rnh.h" |
43e52561 QY |
52 | #include "zebra/zebra_routemap.h" |
53 | #include "zebra/zebra_vrf.h" | |
6134fd82 | 54 | #include "zebra/zebra_vxlan.h" |
7cdb1a84 MS |
55 | #include "zebra/zapi_msg.h" |
56 | #include "zebra/zebra_dplane.h" | |
57 | ||
58 | /* | |
59 | * Event, list, and mutex for delivery of dataplane results | |
60 | */ | |
61 | static pthread_mutex_t dplane_mutex; | |
62 | static struct thread *t_dplane; | |
25779064 | 63 | static struct dplane_ctx_q rib_dplane_q; |
718e3744 | 64 | |
d62a17ae | 65 | DEFINE_HOOK(rib_update, (struct route_node * rn, const char *reason), |
66 | (rn, reason)) | |
4f8ea50c | 67 | |
6baf7bb8 DS |
68 | /* Should we allow non Quagga processes to delete our routes */ |
69 | extern int allow_delete; | |
70 | ||
718e3744 | 71 | /* Each route type's string and default distance value. */ |
d62a17ae | 72 | static const struct { |
73 | int key; | |
8dc7a759 | 74 | uint8_t distance; |
73547a75 | 75 | uint8_t meta_q_map; |
d62a17ae | 76 | } route_info[ZEBRA_ROUTE_MAX] = { |
73547a75 DS |
77 | [ZEBRA_ROUTE_SYSTEM] = {ZEBRA_ROUTE_SYSTEM, 0, 4}, |
78 | [ZEBRA_ROUTE_KERNEL] = {ZEBRA_ROUTE_KERNEL, 0, 0}, | |
79 | [ZEBRA_ROUTE_CONNECT] = {ZEBRA_ROUTE_CONNECT, 0, 0}, | |
80 | [ZEBRA_ROUTE_STATIC] = {ZEBRA_ROUTE_STATIC, 1, 1}, | |
81 | [ZEBRA_ROUTE_RIP] = {ZEBRA_ROUTE_RIP, 120, 2}, | |
82 | [ZEBRA_ROUTE_RIPNG] = {ZEBRA_ROUTE_RIPNG, 120, 2}, | |
83 | [ZEBRA_ROUTE_OSPF] = {ZEBRA_ROUTE_OSPF, 110, 2}, | |
84 | [ZEBRA_ROUTE_OSPF6] = {ZEBRA_ROUTE_OSPF6, 110, 2}, | |
85 | [ZEBRA_ROUTE_ISIS] = {ZEBRA_ROUTE_ISIS, 115, 2}, | |
86 | [ZEBRA_ROUTE_BGP] = {ZEBRA_ROUTE_BGP, 20 /* IBGP is 200. */, 3}, | |
87 | [ZEBRA_ROUTE_PIM] = {ZEBRA_ROUTE_PIM, 255, 4}, | |
88 | [ZEBRA_ROUTE_EIGRP] = {ZEBRA_ROUTE_EIGRP, 90, 2}, | |
89 | [ZEBRA_ROUTE_NHRP] = {ZEBRA_ROUTE_NHRP, 10, 2}, | |
90 | [ZEBRA_ROUTE_HSLS] = {ZEBRA_ROUTE_HSLS, 255, 4}, | |
91 | [ZEBRA_ROUTE_OLSR] = {ZEBRA_ROUTE_OLSR, 255, 4}, | |
92 | [ZEBRA_ROUTE_TABLE] = {ZEBRA_ROUTE_TABLE, 150, 1}, | |
93 | [ZEBRA_ROUTE_LDP] = {ZEBRA_ROUTE_LDP, 150, 4}, | |
94 | [ZEBRA_ROUTE_VNC] = {ZEBRA_ROUTE_VNC, 20, 3}, | |
95 | [ZEBRA_ROUTE_VNC_DIRECT] = {ZEBRA_ROUTE_VNC_DIRECT, 20, 3}, | |
96 | [ZEBRA_ROUTE_VNC_DIRECT_RH] = {ZEBRA_ROUTE_VNC_DIRECT_RH, 20, 3}, | |
97 | [ZEBRA_ROUTE_BGP_DIRECT] = {ZEBRA_ROUTE_BGP_DIRECT, 20, 3}, | |
98 | [ZEBRA_ROUTE_BGP_DIRECT_EXT] = {ZEBRA_ROUTE_BGP_DIRECT_EXT, 20, 3}, | |
99 | [ZEBRA_ROUTE_BABEL] = {ZEBRA_ROUTE_BABEL, 100, 2}, | |
100 | [ZEBRA_ROUTE_SHARP] = {ZEBRA_ROUTE_SHARP, 150, 4}, | |
98156652 | 101 | [ZEBRA_ROUTE_PBR] = {ZEBRA_ROUTE_PBR, 200, 4}, |
42d96b73 | 102 | [ZEBRA_ROUTE_BFD] = {ZEBRA_ROUTE_BFD, 255, 4}, |
eab7b6e3 | 103 | [ZEBRA_ROUTE_OPENFABRIC] = {ZEBRA_ROUTE_OPENFABRIC, 115, 2}, |
303b93cd | 104 | [ZEBRA_ROUTE_VRRP] = {ZEBRA_ROUTE_VRRP, 255, 4} |
d6abd8b0 | 105 | /* Any new route type added to zebra, should be mirrored here */ |
c710b277 | 106 | |
d62a17ae | 107 | /* no entry/default: 150 */ |
718e3744 | 108 | }; |
6b0655a2 | 109 | |
4623d897 DL |
110 | /* RPF lookup behaviour */ |
111 | static enum multicast_mode ipv4_multicast_mode = MCAST_NO_CONFIG; | |
112 | ||
6c4f4e6e | 113 | |
d62a17ae | 114 | static void __attribute__((format(printf, 5, 6))) |
115 | _rnode_zlog(const char *_func, vrf_id_t vrf_id, struct route_node *rn, | |
116 | int priority, const char *msgfmt, ...) | |
2263a412 | 117 | { |
d62a17ae | 118 | char buf[SRCDEST2STR_BUFFER + sizeof(" (MRIB)")]; |
119 | char msgbuf[512]; | |
120 | va_list ap; | |
2263a412 | 121 | |
d62a17ae | 122 | va_start(ap, msgfmt); |
123 | vsnprintf(msgbuf, sizeof(msgbuf), msgfmt, ap); | |
124 | va_end(ap); | |
2263a412 | 125 | |
d62a17ae | 126 | if (rn) { |
127 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
128 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
cb653491 | 129 | |
d62a17ae | 130 | if (info->safi == SAFI_MULTICAST) |
131 | strcat(buf, " (MRIB)"); | |
132 | } else { | |
133 | snprintf(buf, sizeof(buf), "{(route_node *) NULL}"); | |
134 | } | |
2263a412 | 135 | |
d62a17ae | 136 | zlog(priority, "%s: %d:%s: %s", _func, vrf_id, buf, msgbuf); |
2263a412 DL |
137 | } |
138 | ||
d62a17ae | 139 | #define rnode_debug(node, vrf_id, ...) \ |
2263a412 | 140 | _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__) |
d62a17ae | 141 | #define rnode_info(node, ...) \ |
2263a412 DL |
142 | _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__) |
143 | ||
fd289fc8 | 144 | uint8_t route_distance(int type) |
40c7bdb0 | 145 | { |
fd289fc8 | 146 | uint8_t distance; |
40c7bdb0 | 147 | |
d62a17ae | 148 | if ((unsigned)type >= array_size(route_info)) |
149 | distance = 150; | |
150 | else | |
151 | distance = route_info[type].distance; | |
40c7bdb0 | 152 | |
d62a17ae | 153 | return distance; |
40c7bdb0 | 154 | } |
155 | ||
d7c0a89a | 156 | int is_zebra_valid_kernel_table(uint32_t table_id) |
7a4bb9c5 | 157 | { |
8f500a1c | 158 | #ifdef linux |
d62a17ae | 159 | if ((table_id == RT_TABLE_UNSPEC) || (table_id == RT_TABLE_LOCAL) |
160 | || (table_id == RT_TABLE_COMPAT)) | |
161 | return 0; | |
8f500a1c RW |
162 | #endif |
163 | ||
d62a17ae | 164 | return 1; |
7a4bb9c5 DS |
165 | } |
166 | ||
d7c0a89a | 167 | int is_zebra_main_routing_table(uint32_t table_id) |
7a4bb9c5 | 168 | { |
c447ad08 | 169 | if (table_id == RT_TABLE_MAIN) |
d62a17ae | 170 | return 1; |
171 | return 0; | |
7a4bb9c5 DS |
172 | } |
173 | ||
86391e56 | 174 | int zebra_check_addr(const struct prefix *p) |
0aabccc0 | 175 | { |
d62a17ae | 176 | if (p->family == AF_INET) { |
d7c0a89a | 177 | uint32_t addr; |
0aabccc0 | 178 | |
d62a17ae | 179 | addr = p->u.prefix4.s_addr; |
180 | addr = ntohl(addr); | |
0aabccc0 | 181 | |
d62a17ae | 182 | if (IPV4_NET127(addr) || IN_CLASSD(addr) |
183 | || IPV4_LINKLOCAL(addr)) | |
184 | return 0; | |
185 | } | |
186 | if (p->family == AF_INET6) { | |
187 | if (IN6_IS_ADDR_LOOPBACK(&p->u.prefix6)) | |
188 | return 0; | |
189 | if (IN6_IS_ADDR_LINKLOCAL(&p->u.prefix6)) | |
190 | return 0; | |
191 | } | |
192 | return 1; | |
0aabccc0 DD |
193 | } |
194 | ||
fa713d9e | 195 | /* Add nexthop to the end of a rib node's nexthop list */ |
d62a17ae | 196 | void route_entry_nexthop_add(struct route_entry *re, struct nexthop *nexthop) |
fa713d9e | 197 | { |
7ee30f28 | 198 | nexthop_add(&re->ng.nexthop, nexthop); |
d62a17ae | 199 | re->nexthop_num++; |
718e3744 | 200 | } |
201 | ||
6e26278c | 202 | |
6e26278c DS |
203 | /** |
204 | * copy_nexthop - copy a nexthop to the rib structure. | |
205 | */ | |
d62a17ae | 206 | void route_entry_copy_nexthops(struct route_entry *re, struct nexthop *nh) |
6e26278c | 207 | { |
7ee30f28 DS |
208 | assert(!re->ng.nexthop); |
209 | copy_nexthops(&re->ng.nexthop, nh, NULL); | |
d62a17ae | 210 | for (struct nexthop *nexthop = nh; nexthop; nexthop = nexthop->next) |
211 | re->nexthop_num++; | |
6e26278c DS |
212 | } |
213 | ||
718e3744 | 214 | /* Delete specified nexthop from the list. */ |
d62a17ae | 215 | void route_entry_nexthop_delete(struct route_entry *re, struct nexthop *nexthop) |
718e3744 | 216 | { |
d62a17ae | 217 | if (nexthop->next) |
218 | nexthop->next->prev = nexthop->prev; | |
219 | if (nexthop->prev) | |
220 | nexthop->prev->next = nexthop->next; | |
221 | else | |
7ee30f28 | 222 | re->ng.nexthop = nexthop->next; |
d62a17ae | 223 | re->nexthop_num--; |
718e3744 | 224 | } |
225 | ||
fa713d9e | 226 | |
d62a17ae | 227 | struct nexthop *route_entry_nexthop_ifindex_add(struct route_entry *re, |
4a7371e9 DS |
228 | ifindex_t ifindex, |
229 | vrf_id_t nh_vrf_id) | |
718e3744 | 230 | { |
d62a17ae | 231 | struct nexthop *nexthop; |
718e3744 | 232 | |
d62a17ae | 233 | nexthop = nexthop_new(); |
234 | nexthop->type = NEXTHOP_TYPE_IFINDEX; | |
235 | nexthop->ifindex = ifindex; | |
4a7371e9 | 236 | nexthop->vrf_id = nh_vrf_id; |
718e3744 | 237 | |
d62a17ae | 238 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 239 | |
d62a17ae | 240 | return nexthop; |
718e3744 | 241 | } |
242 | ||
d62a17ae | 243 | struct nexthop *route_entry_nexthop_ipv4_add(struct route_entry *re, |
244 | struct in_addr *ipv4, | |
4a7371e9 DS |
245 | struct in_addr *src, |
246 | vrf_id_t nh_vrf_id) | |
718e3744 | 247 | { |
d62a17ae | 248 | struct nexthop *nexthop; |
718e3744 | 249 | |
d62a17ae | 250 | nexthop = nexthop_new(); |
251 | nexthop->type = NEXTHOP_TYPE_IPV4; | |
4a7371e9 | 252 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 253 | nexthop->gate.ipv4 = *ipv4; |
254 | if (src) | |
255 | nexthop->src.ipv4 = *src; | |
718e3744 | 256 | |
d62a17ae | 257 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 258 | |
d62a17ae | 259 | return nexthop; |
718e3744 | 260 | } |
261 | ||
d62a17ae | 262 | struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re, |
263 | struct in_addr *ipv4, | |
264 | struct in_addr *src, | |
4a7371e9 DS |
265 | ifindex_t ifindex, |
266 | vrf_id_t nh_vrf_id) | |
718e3744 | 267 | { |
d62a17ae | 268 | struct nexthop *nexthop; |
269 | struct interface *ifp; | |
718e3744 | 270 | |
d62a17ae | 271 | nexthop = nexthop_new(); |
4a7371e9 | 272 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 273 | nexthop->type = NEXTHOP_TYPE_IPV4_IFINDEX; |
274 | nexthop->gate.ipv4 = *ipv4; | |
275 | if (src) | |
276 | nexthop->src.ipv4 = *src; | |
277 | nexthop->ifindex = ifindex; | |
4a7371e9 | 278 | ifp = if_lookup_by_index(nexthop->ifindex, nh_vrf_id); |
d62a17ae | 279 | /*Pending: need to think if null ifp here is ok during bootup? |
280 | There was a crash because ifp here was coming to be NULL */ | |
281 | if (ifp) | |
2b83602b | 282 | if (connected_is_unnumbered(ifp)) |
d62a17ae | 283 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK); |
718e3744 | 284 | |
d62a17ae | 285 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 286 | |
d62a17ae | 287 | return nexthop; |
718e3744 | 288 | } |
289 | ||
d62a17ae | 290 | struct nexthop *route_entry_nexthop_ipv6_add(struct route_entry *re, |
4a7371e9 DS |
291 | struct in6_addr *ipv6, |
292 | vrf_id_t nh_vrf_id) | |
718e3744 | 293 | { |
d62a17ae | 294 | struct nexthop *nexthop; |
718e3744 | 295 | |
d62a17ae | 296 | nexthop = nexthop_new(); |
4a7371e9 | 297 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 298 | nexthop->type = NEXTHOP_TYPE_IPV6; |
299 | nexthop->gate.ipv6 = *ipv6; | |
718e3744 | 300 | |
d62a17ae | 301 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 302 | |
d62a17ae | 303 | return nexthop; |
718e3744 | 304 | } |
305 | ||
d62a17ae | 306 | struct nexthop *route_entry_nexthop_ipv6_ifindex_add(struct route_entry *re, |
307 | struct in6_addr *ipv6, | |
4a7371e9 DS |
308 | ifindex_t ifindex, |
309 | vrf_id_t nh_vrf_id) | |
718e3744 | 310 | { |
d62a17ae | 311 | struct nexthop *nexthop; |
718e3744 | 312 | |
d62a17ae | 313 | nexthop = nexthop_new(); |
4a7371e9 | 314 | nexthop->vrf_id = nh_vrf_id; |
d62a17ae | 315 | nexthop->type = NEXTHOP_TYPE_IPV6_IFINDEX; |
316 | nexthop->gate.ipv6 = *ipv6; | |
317 | nexthop->ifindex = ifindex; | |
718e3744 | 318 | |
d62a17ae | 319 | route_entry_nexthop_add(re, nexthop); |
718e3744 | 320 | |
d62a17ae | 321 | return nexthop; |
718e3744 | 322 | } |
718e3744 | 323 | |
a8309422 | 324 | struct nexthop *route_entry_nexthop_blackhole_add(struct route_entry *re, |
60466a63 | 325 | enum blackhole_type bh_type) |
595db7f1 | 326 | { |
d62a17ae | 327 | struct nexthop *nexthop; |
595db7f1 | 328 | |
d62a17ae | 329 | nexthop = nexthop_new(); |
4a7371e9 | 330 | nexthop->vrf_id = VRF_DEFAULT; |
d62a17ae | 331 | nexthop->type = NEXTHOP_TYPE_BLACKHOLE; |
a8309422 | 332 | nexthop->bh_type = bh_type; |
595db7f1 | 333 | |
d62a17ae | 334 | route_entry_nexthop_add(re, nexthop); |
595db7f1 | 335 | |
d62a17ae | 336 | return nexthop; |
595db7f1 | 337 | } |
338 | ||
86391e56 | 339 | static void nexthop_set_resolved(afi_t afi, const struct nexthop *newhop, |
d62a17ae | 340 | struct nexthop *nexthop) |
4491a88f | 341 | { |
d62a17ae | 342 | struct nexthop *resolved_hop; |
4491a88f | 343 | |
d62a17ae | 344 | resolved_hop = nexthop_new(); |
345 | SET_FLAG(resolved_hop->flags, NEXTHOP_FLAG_ACTIVE); | |
d855d11f | 346 | |
4a7371e9 | 347 | resolved_hop->vrf_id = nexthop->vrf_id; |
d855d11f RW |
348 | switch (newhop->type) { |
349 | case NEXTHOP_TYPE_IPV4: | |
350 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
351 | /* If the resolving route specifies a gateway, use it */ | |
d62a17ae | 352 | resolved_hop->type = newhop->type; |
353 | resolved_hop->gate.ipv4 = newhop->gate.ipv4; | |
4491a88f | 354 | |
d62a17ae | 355 | if (newhop->ifindex) { |
356 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
357 | resolved_hop->ifindex = newhop->ifindex; | |
d62a17ae | 358 | } |
d855d11f RW |
359 | break; |
360 | case NEXTHOP_TYPE_IPV6: | |
361 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
d62a17ae | 362 | resolved_hop->type = newhop->type; |
363 | resolved_hop->gate.ipv6 = newhop->gate.ipv6; | |
364 | ||
365 | if (newhop->ifindex) { | |
366 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
367 | resolved_hop->ifindex = newhop->ifindex; | |
368 | } | |
d855d11f RW |
369 | break; |
370 | case NEXTHOP_TYPE_IFINDEX: | |
371 | /* If the resolving route is an interface route, | |
372 | * it means the gateway we are looking up is connected | |
373 | * to that interface. (The actual network is _not_ onlink). | |
374 | * Therefore, the resolved route should have the original | |
375 | * gateway as nexthop as it is directly connected. | |
376 | * | |
377 | * On Linux, we have to set the onlink netlink flag because | |
378 | * otherwise, the kernel won't accept the route. | |
379 | */ | |
d62a17ae | 380 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; |
381 | if (afi == AFI_IP) { | |
382 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
383 | resolved_hop->gate.ipv4 = nexthop->gate.ipv4; | |
384 | } else if (afi == AFI_IP6) { | |
385 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
386 | resolved_hop->gate.ipv6 = nexthop->gate.ipv6; | |
387 | } | |
388 | resolved_hop->ifindex = newhop->ifindex; | |
d855d11f RW |
389 | break; |
390 | case NEXTHOP_TYPE_BLACKHOLE: | |
59693377 DS |
391 | resolved_hop->type = NEXTHOP_TYPE_BLACKHOLE; |
392 | resolved_hop->bh_type = nexthop->bh_type; | |
d855d11f | 393 | break; |
59693377 | 394 | } |
d855d11f | 395 | |
effcfaeb DS |
396 | if (newhop->flags & NEXTHOP_FLAG_ONLINK) |
397 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
398 | ||
f674dfe2 RW |
399 | /* Copy labels of the resolved route */ |
400 | if (newhop->nh_label) | |
401 | nexthop_add_labels(resolved_hop, newhop->nh_label_type, | |
402 | newhop->nh_label->num_labels, | |
403 | &newhop->nh_label->label[0]); | |
404 | ||
d62a17ae | 405 | resolved_hop->rparent = nexthop; |
406 | nexthop_add(&nexthop->resolved, resolved_hop); | |
4491a88f DS |
407 | } |
408 | ||
99eabcec DS |
409 | /* |
410 | * Given a nexthop we need to properly recursively resolve | |
411 | * the route. As such, do a table lookup to find and match | |
412 | * if at all possible. Set the nexthop->ifindex as appropriate | |
413 | */ | |
d62a17ae | 414 | static int nexthop_active(afi_t afi, struct route_entry *re, |
99eabcec | 415 | struct nexthop *nexthop, |
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 | ||
99eabcec DS |
431 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE); |
432 | nexthops_free(nexthop->resolved); | |
433 | nexthop->resolved = NULL; | |
434 | re->nexthop_mtu = 0; | |
18ff3edd | 435 | |
058c16b7 DS |
436 | /* |
437 | * If the kernel has sent us a route, then | |
438 | * by golly gee whiz it's a good route. | |
439 | */ | |
440 | if (re->type == ZEBRA_ROUTE_KERNEL || | |
441 | re->type == ZEBRA_ROUTE_SYSTEM) | |
442 | return 1; | |
443 | ||
d62a17ae | 444 | /* |
445 | * Check to see if we should trust the passed in information | |
446 | * for UNNUMBERED interfaces as that we won't find the GW | |
447 | * address in the routing table. | |
2b83602b | 448 | * This check should suffice to handle IPv4 or IPv6 routes |
449 | * sourced from EVPN routes which are installed with the | |
450 | * next hop as the remote VTEP IP. | |
d62a17ae | 451 | */ |
452 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK)) { | |
4a7371e9 | 453 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); |
fe85601c DS |
454 | if (!ifp) { |
455 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
456 | zlog_debug( | |
457 | "\t%s: Onlink and interface: %u[%u] does not exist", | |
458 | __PRETTY_FUNCTION__, nexthop->ifindex, | |
459 | nexthop->vrf_id); | |
460 | return 0; | |
461 | } | |
462 | if (connected_is_unnumbered(ifp)) { | |
d62a17ae | 463 | if (if_is_operative(ifp)) |
464 | return 1; | |
34815ea3 DS |
465 | else { |
466 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
467 | zlog_debug( | |
468 | "\t%s: Onlink and interface %s is not operative", | |
469 | __PRETTY_FUNCTION__, ifp->name); | |
d62a17ae | 470 | return 0; |
34815ea3 | 471 | } |
fe85601c DS |
472 | } |
473 | if (!if_is_operative(ifp)) { | |
34815ea3 DS |
474 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
475 | zlog_debug( | |
476 | "\t%s: Interface %s is not unnumbered", | |
4bb55bbe | 477 | __PRETTY_FUNCTION__, ifp->name); |
d62a17ae | 478 | return 0; |
34815ea3 | 479 | } |
16814f96 | 480 | } |
718e3744 | 481 | |
d62a17ae | 482 | /* Make lookup prefix. */ |
483 | memset(&p, 0, sizeof(struct prefix)); | |
484 | switch (afi) { | |
485 | case AFI_IP: | |
486 | p.family = AF_INET; | |
487 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
488 | p.u.prefix4 = nexthop->gate.ipv4; | |
489 | break; | |
490 | case AFI_IP6: | |
491 | p.family = AF_INET6; | |
492 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
493 | p.u.prefix6 = nexthop->gate.ipv6; | |
494 | break; | |
495 | default: | |
496 | assert(afi != AFI_IP && afi != AFI_IP6); | |
497 | break; | |
718e3744 | 498 | } |
d62a17ae | 499 | /* Lookup table. */ |
4a7371e9 | 500 | table = zebra_vrf_table(afi, SAFI_UNICAST, nexthop->vrf_id); |
34815ea3 DS |
501 | if (!table) { |
502 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
503 | zlog_debug("\t%s: Table not found", | |
504 | __PRETTY_FUNCTION__); | |
d62a17ae | 505 | return 0; |
34815ea3 | 506 | } |
d62a17ae | 507 | |
508 | rn = route_node_match(table, (struct prefix *)&p); | |
509 | while (rn) { | |
510 | route_unlock_node(rn); | |
511 | ||
fd7fd9e5 DS |
512 | /* Lookup should halt if we've matched against ourselves ('top', |
513 | * if specified) - i.e., we cannot have a nexthop NH1 is | |
514 | * resolved by a route NH1. The exception is if the route is a | |
515 | * host route. | |
516 | */ | |
517 | if (top && rn == top) | |
996c9314 | 518 | if (((afi == AFI_IP) && (rn->p.prefixlen != 32)) |
34815ea3 DS |
519 | || ((afi == AFI_IP6) && (rn->p.prefixlen != 128))) { |
520 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
521 | zlog_debug( | |
522 | "\t%s: Matched against ourself and prefix length is not max bit length", | |
523 | __PRETTY_FUNCTION__); | |
fd7fd9e5 | 524 | return 0; |
34815ea3 | 525 | } |
d62a17ae | 526 | |
527 | /* Pick up selected route. */ | |
528 | /* However, do not resolve over default route unless explicitly | |
529 | * allowed. */ | |
530 | if (is_default_prefix(&rn->p) | |
34815ea3 DS |
531 | && !rnh_resolve_via_default(p.family)) { |
532 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
533 | zlog_debug( | |
534 | "\t:%s: Resolved against default route", | |
535 | __PRETTY_FUNCTION__); | |
d62a17ae | 536 | return 0; |
34815ea3 | 537 | } |
d62a17ae | 538 | |
5f7a4718 | 539 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
540 | if (dest && dest->selected_fib |
541 | && !CHECK_FLAG(dest->selected_fib->status, | |
542 | ROUTE_ENTRY_REMOVED) | |
543 | && dest->selected_fib->type != ZEBRA_ROUTE_TABLE) | |
5f7a4718 | 544 | match = dest->selected_fib; |
48a53dc7 | 545 | |
d62a17ae | 546 | /* If there is no selected route or matched route is EGP, go up |
547 | tree. */ | |
548 | if (!match) { | |
549 | do { | |
550 | rn = rn->parent; | |
551 | } while (rn && rn->info == NULL); | |
552 | if (rn) | |
553 | route_lock_node(rn); | |
c87bdd2b | 554 | |
d62a17ae | 555 | continue; |
556 | } | |
557 | ||
d62a17ae | 558 | if (match->type == ZEBRA_ROUTE_CONNECT) { |
559 | /* Directly point connected route. */ | |
7ee30f28 | 560 | newhop = match->ng.nexthop; |
d62a17ae | 561 | if (newhop) { |
562 | if (nexthop->type == NEXTHOP_TYPE_IPV4 | |
563 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
564 | nexthop->ifindex = newhop->ifindex; | |
565 | } | |
566 | return 1; | |
4e8b02f4 | 567 | } else if (CHECK_FLAG(re->flags, ZEBRA_FLAG_ALLOW_RECURSION)) { |
d62a17ae | 568 | resolved = 0; |
7ee30f28 | 569 | for (ALL_NEXTHOPS(match->ng, newhop)) { |
677c1dd5 DS |
570 | if (!CHECK_FLAG(match->status, |
571 | ROUTE_ENTRY_INSTALLED)) | |
a8309422 DL |
572 | continue; |
573 | if (CHECK_FLAG(newhop->flags, | |
574 | NEXTHOP_FLAG_RECURSIVE)) | |
575 | continue; | |
576 | ||
99eabcec DS |
577 | SET_FLAG(nexthop->flags, |
578 | NEXTHOP_FLAG_RECURSIVE); | |
579 | SET_FLAG(re->status, | |
580 | ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
581 | nexthop_set_resolved(afi, newhop, nexthop); | |
a8309422 DL |
582 | resolved = 1; |
583 | } | |
99eabcec | 584 | if (resolved) |
43e31305 | 585 | re->nexthop_mtu = match->mtu; |
34815ea3 DS |
586 | if (!resolved && IS_ZEBRA_DEBUG_RIB_DETAILED) |
587 | zlog_debug("\t%s: Recursion failed to find", | |
588 | __PRETTY_FUNCTION__); | |
d62a17ae | 589 | return resolved; |
590 | } else if (re->type == ZEBRA_ROUTE_STATIC) { | |
591 | resolved = 0; | |
7ee30f28 | 592 | for (ALL_NEXTHOPS(match->ng, newhop)) { |
677c1dd5 DS |
593 | if (!CHECK_FLAG(match->status, |
594 | ROUTE_ENTRY_INSTALLED)) | |
a8309422 | 595 | continue; |
6c47d399 DS |
596 | if (CHECK_FLAG(newhop->flags, |
597 | NEXTHOP_FLAG_RECURSIVE)) | |
598 | continue; | |
a8309422 | 599 | |
99eabcec DS |
600 | SET_FLAG(nexthop->flags, |
601 | NEXTHOP_FLAG_RECURSIVE); | |
602 | nexthop_set_resolved(afi, newhop, nexthop); | |
a8309422 DL |
603 | resolved = 1; |
604 | } | |
99eabcec | 605 | if (resolved) |
d62a17ae | 606 | re->nexthop_mtu = match->mtu; |
34815ea3 DS |
607 | |
608 | if (!resolved && IS_ZEBRA_DEBUG_RIB_DETAILED) | |
609 | zlog_debug( | |
610 | "\t%s: Static route unable to resolve", | |
611 | __PRETTY_FUNCTION__); | |
d62a17ae | 612 | return resolved; |
613 | } else { | |
7939ff76 DS |
614 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
615 | zlog_debug("\t%s: Route Type %s has not turned on recursion", | |
616 | __PRETTY_FUNCTION__, | |
617 | zebra_route_string(re->type)); | |
618 | if (re->type == ZEBRA_ROUTE_BGP && | |
619 | !CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
620 | zlog_debug("\tEBGP: see \"disable-ebgp-connected-route-check\" or \"disable-connected-check\""); | |
621 | } | |
d62a17ae | 622 | return 0; |
623 | } | |
718e3744 | 624 | } |
7939ff76 DS |
625 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
626 | zlog_debug("\t%s: Nexthop did not lookup in table", | |
627 | __PRETTY_FUNCTION__); | |
d62a17ae | 628 | return 0; |
718e3744 | 629 | } |
718e3744 | 630 | |
d62a17ae | 631 | struct route_entry *rib_match(afi_t afi, safi_t safi, vrf_id_t vrf_id, |
632 | union g_addr *addr, struct route_node **rn_out) | |
633 | { | |
634 | struct prefix p; | |
635 | struct route_table *table; | |
636 | struct route_node *rn; | |
5f7a4718 | 637 | struct route_entry *match = NULL; |
d62a17ae | 638 | |
639 | /* Lookup table. */ | |
640 | table = zebra_vrf_table(afi, safi, vrf_id); | |
641 | if (!table) | |
642 | return 0; | |
643 | ||
644 | memset(&p, 0, sizeof(struct prefix)); | |
645 | p.family = afi; | |
646 | if (afi == AFI_IP) { | |
647 | p.u.prefix4 = addr->ipv4; | |
648 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
649 | } else { | |
650 | p.u.prefix6 = addr->ipv6; | |
651 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
652 | } | |
718e3744 | 653 | |
d62a17ae | 654 | rn = route_node_match(table, (struct prefix *)&p); |
718e3744 | 655 | |
d62a17ae | 656 | while (rn) { |
5f7a4718 DS |
657 | rib_dest_t *dest; |
658 | ||
d62a17ae | 659 | route_unlock_node(rn); |
718e3744 | 660 | |
5f7a4718 | 661 | dest = rib_dest_from_rnode(rn); |
996c9314 LB |
662 | if (dest && dest->selected_fib |
663 | && !CHECK_FLAG(dest->selected_fib->status, | |
664 | ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 665 | match = dest->selected_fib; |
718e3744 | 666 | |
d62a17ae | 667 | /* If there is no selected route or matched route is EGP, go up |
668 | tree. */ | |
669 | if (!match) { | |
670 | do { | |
671 | rn = rn->parent; | |
672 | } while (rn && rn->info == NULL); | |
673 | if (rn) | |
674 | route_lock_node(rn); | |
675 | } else { | |
676 | if (match->type != ZEBRA_ROUTE_CONNECT) { | |
677c1dd5 DS |
677 | if (!CHECK_FLAG(match->status, |
678 | ROUTE_ENTRY_INSTALLED)) | |
d62a17ae | 679 | return NULL; |
680 | } | |
681 | ||
682 | if (rn_out) | |
683 | *rn_out = rn; | |
684 | return match; | |
685 | } | |
16814f96 | 686 | } |
d62a17ae | 687 | return NULL; |
688 | } | |
689 | ||
690 | struct route_entry *rib_match_ipv4_multicast(vrf_id_t vrf_id, | |
691 | struct in_addr addr, | |
692 | struct route_node **rn_out) | |
693 | { | |
694 | struct route_entry *re = NULL, *mre = NULL, *ure = NULL; | |
695 | struct route_node *m_rn = NULL, *u_rn = NULL; | |
696 | union g_addr gaddr = {.ipv4 = addr}; | |
697 | ||
698 | switch (ipv4_multicast_mode) { | |
699 | case MCAST_MRIB_ONLY: | |
700 | return rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
701 | rn_out); | |
702 | case MCAST_URIB_ONLY: | |
703 | return rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, rn_out); | |
704 | case MCAST_NO_CONFIG: | |
705 | case MCAST_MIX_MRIB_FIRST: | |
706 | re = mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, | |
707 | &m_rn); | |
708 | if (!mre) | |
709 | re = ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, | |
710 | &gaddr, &u_rn); | |
711 | break; | |
712 | case MCAST_MIX_DISTANCE: | |
713 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
714 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
715 | if (mre && ure) | |
716 | re = ure->distance < mre->distance ? ure : mre; | |
717 | else if (mre) | |
718 | re = mre; | |
719 | else if (ure) | |
720 | re = ure; | |
721 | break; | |
722 | case MCAST_MIX_PFXLEN: | |
723 | mre = rib_match(AFI_IP, SAFI_MULTICAST, vrf_id, &gaddr, &m_rn); | |
724 | ure = rib_match(AFI_IP, SAFI_UNICAST, vrf_id, &gaddr, &u_rn); | |
725 | if (mre && ure) | |
726 | re = u_rn->p.prefixlen > m_rn->p.prefixlen ? ure : mre; | |
727 | else if (mre) | |
728 | re = mre; | |
729 | else if (ure) | |
730 | re = ure; | |
731 | break; | |
718e3744 | 732 | } |
33550aa8 | 733 | |
d62a17ae | 734 | if (rn_out) |
735 | *rn_out = (re == mre) ? m_rn : u_rn; | |
718e3744 | 736 | |
d62a17ae | 737 | if (IS_ZEBRA_DEBUG_RIB) { |
738 | char buf[BUFSIZ]; | |
739 | inet_ntop(AF_INET, &addr, buf, BUFSIZ); | |
4623d897 | 740 | |
32391aff DS |
741 | zlog_debug("%s: %s: vrf: %u found %s, using %s", |
742 | __func__, buf, vrf_id, | |
d62a17ae | 743 | mre ? (ure ? "MRIB+URIB" : "MRIB") |
744 | : ure ? "URIB" : "nothing", | |
745 | re == ure ? "URIB" : re == mre ? "MRIB" : "none"); | |
746 | } | |
747 | return re; | |
4623d897 DL |
748 | } |
749 | ||
d62a17ae | 750 | void multicast_mode_ipv4_set(enum multicast_mode mode) |
4623d897 | 751 | { |
d62a17ae | 752 | if (IS_ZEBRA_DEBUG_RIB) |
753 | zlog_debug("%s: multicast lookup mode set (%d)", __func__, | |
754 | mode); | |
755 | ipv4_multicast_mode = mode; | |
4623d897 DL |
756 | } |
757 | ||
d62a17ae | 758 | enum multicast_mode multicast_mode_ipv4_get(void) |
4623d897 | 759 | { |
d62a17ae | 760 | return ipv4_multicast_mode; |
4623d897 DL |
761 | } |
762 | ||
d62a17ae | 763 | struct route_entry *rib_lookup_ipv4(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
718e3744 | 764 | { |
d62a17ae | 765 | struct route_table *table; |
766 | struct route_node *rn; | |
5f7a4718 | 767 | struct route_entry *match = NULL; |
5f7a4718 | 768 | rib_dest_t *dest; |
718e3744 | 769 | |
d62a17ae | 770 | /* Lookup table. */ |
771 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
772 | if (!table) | |
773 | return 0; | |
718e3744 | 774 | |
d62a17ae | 775 | rn = route_node_lookup(table, (struct prefix *)p); |
718e3744 | 776 | |
d62a17ae | 777 | /* No route for this prefix. */ |
778 | if (!rn) | |
779 | return NULL; | |
718e3744 | 780 | |
d62a17ae | 781 | /* Unlock node. */ |
782 | route_unlock_node(rn); | |
5f7a4718 | 783 | dest = rib_dest_from_rnode(rn); |
718e3744 | 784 | |
996c9314 LB |
785 | if (dest && dest->selected_fib |
786 | && !CHECK_FLAG(dest->selected_fib->status, ROUTE_ENTRY_REMOVED)) | |
5f7a4718 | 787 | match = dest->selected_fib; |
718e3744 | 788 | |
d62a17ae | 789 | if (!match) |
790 | return NULL; | |
718e3744 | 791 | |
d62a17ae | 792 | if (match->type == ZEBRA_ROUTE_CONNECT) |
793 | return match; | |
f9e1b38e | 794 | |
677c1dd5 DS |
795 | if (CHECK_FLAG(match->status, ROUTE_ENTRY_INSTALLED)) |
796 | return match; | |
718e3744 | 797 | |
d62a17ae | 798 | return NULL; |
718e3744 | 799 | } |
800 | ||
dc95824a DO |
801 | /* This function verifies reachability of one given nexthop, which can be |
802 | * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored | |
99eabcec DS |
803 | * in nexthop->flags field. The nexthop->ifindex will be updated |
804 | * appropriately as well. An existing route map can turn | |
805 | * (otherwise active) nexthop into inactive, but not vice versa. | |
dc95824a DO |
806 | * |
807 | * The return value is the final value of 'ACTIVE' flag. | |
808 | */ | |
d62a17ae | 809 | static unsigned nexthop_active_check(struct route_node *rn, |
810 | struct route_entry *re, | |
99eabcec | 811 | struct nexthop *nexthop) |
d62a17ae | 812 | { |
813 | struct interface *ifp; | |
814 | route_map_result_t ret = RMAP_MATCH; | |
815 | int family; | |
816 | char buf[SRCDEST2STR_BUFFER]; | |
86391e56 | 817 | const struct prefix *p, *src_p; |
ac6eebce | 818 | struct zebra_vrf *zvrf; |
819 | ||
d62a17ae | 820 | srcdest_rnode_prefixes(rn, &p, &src_p); |
821 | ||
822 | if (rn->p.family == AF_INET) | |
823 | family = AFI_IP; | |
824 | else if (rn->p.family == AF_INET6) | |
825 | family = AFI_IP6; | |
826 | else | |
827 | family = 0; | |
828 | switch (nexthop->type) { | |
829 | case NEXTHOP_TYPE_IFINDEX: | |
4a7371e9 | 830 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); |
d62a17ae | 831 | if (ifp && if_is_operative(ifp)) |
832 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
833 | else | |
834 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
835 | break; | |
836 | case NEXTHOP_TYPE_IPV4: | |
837 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
838 | family = AFI_IP; | |
99eabcec | 839 | if (nexthop_active(AFI_IP, re, nexthop, rn)) |
d62a17ae | 840 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
841 | else | |
842 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
843 | break; | |
844 | case NEXTHOP_TYPE_IPV6: | |
845 | family = AFI_IP6; | |
99eabcec | 846 | if (nexthop_active(AFI_IP6, re, nexthop, rn)) |
d62a17ae | 847 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
848 | else | |
849 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
850 | break; | |
851 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
852 | /* RFC 5549, v4 prefix with v6 NH */ | |
853 | if (rn->p.family != AF_INET) | |
854 | family = AFI_IP6; | |
855 | if (IN6_IS_ADDR_LINKLOCAL(&nexthop->gate.ipv6)) { | |
007dbee6 | 856 | ifp = if_lookup_by_index(nexthop->ifindex, |
4a7371e9 | 857 | nexthop->vrf_id); |
d62a17ae | 858 | if (ifp && if_is_operative(ifp)) |
859 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
860 | else | |
861 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
862 | } else { | |
99eabcec | 863 | if (nexthop_active(AFI_IP6, re, nexthop, rn)) |
d62a17ae | 864 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
865 | else | |
866 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
867 | } | |
868 | break; | |
869 | case NEXTHOP_TYPE_BLACKHOLE: | |
870 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
871 | break; | |
872 | default: | |
873 | break; | |
874 | } | |
34815ea3 DS |
875 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) { |
876 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
877 | zlog_debug("\t%s: Unable to find a active nexthop", | |
878 | __PRETTY_FUNCTION__); | |
d62a17ae | 879 | return 0; |
34815ea3 | 880 | } |
d62a17ae | 881 | |
882 | /* XXX: What exactly do those checks do? Do we support | |
ac6eebce | 883 | * e.g. IPv4 routes with IPv6 nexthops or vice versa? |
884 | */ | |
d62a17ae | 885 | if (RIB_SYSTEM_ROUTE(re) || (family == AFI_IP && p->family != AF_INET) |
886 | || (family == AFI_IP6 && p->family != AF_INET6)) | |
887 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
888 | ||
889 | /* The original code didn't determine the family correctly | |
890 | * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi | |
891 | * from the rib_table_info in those cases. | |
892 | * Possibly it may be better to use only the rib_table_info | |
893 | * in every case. | |
894 | */ | |
895 | if (!family) { | |
896 | rib_table_info_t *info; | |
897 | ||
898 | info = srcdest_rnode_table_info(rn); | |
899 | family = info->afi; | |
718e3744 | 900 | } |
c52ef59f | 901 | |
d62a17ae | 902 | memset(&nexthop->rmap_src.ipv6, 0, sizeof(union g_addr)); |
903 | ||
ac6eebce | 904 | zvrf = zebra_vrf_lookup_by_id(nexthop->vrf_id); |
905 | if (!zvrf) { | |
906 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
907 | zlog_debug("\t%s: zvrf is NULL", __PRETTY_FUNCTION__); | |
908 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
909 | } | |
910 | ||
d62a17ae | 911 | /* It'll get set if required inside */ |
ac6eebce | 912 | ret = zebra_route_map_check(family, re->type, re->instance, p, |
913 | nexthop, zvrf, re->tag); | |
d62a17ae | 914 | if (ret == RMAP_DENYMATCH) { |
915 | if (IS_ZEBRA_DEBUG_RIB) { | |
916 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
917 | zlog_debug( | |
918 | "%u:%s: Filtering out with NH out %s due to route map", | |
919 | re->vrf_id, buf, | |
99b9d960 | 920 | ifindex2ifname(nexthop->ifindex, |
4a7371e9 | 921 | nexthop->vrf_id)); |
d62a17ae | 922 | } |
923 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
518f0eb1 | 924 | } |
d62a17ae | 925 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
718e3744 | 926 | } |
927 | ||
99eabcec DS |
928 | /* |
929 | * Iterate over all nexthops of the given RIB entry and refresh their | |
f0f77c9a DS |
930 | * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any |
931 | * nexthop is found to toggle the ACTIVE flag, the whole re structure | |
99eabcec | 932 | * is flagged with ROUTE_ENTRY_CHANGED. |
03e232a4 DO |
933 | * |
934 | * Return value is the new number of active nexthops. | |
935 | */ | |
99eabcec | 936 | static int nexthop_active_update(struct route_node *rn, struct route_entry *re) |
d62a17ae | 937 | { |
938 | struct nexthop *nexthop; | |
939 | union g_addr prev_src; | |
80ad0418 | 940 | unsigned int prev_active, new_active; |
d62a17ae | 941 | ifindex_t prev_index; |
ac6eebce | 942 | |
d62a17ae | 943 | re->nexthop_active_num = 0; |
944 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
945 | ||
7ee30f28 | 946 | for (nexthop = re->ng.nexthop; nexthop; nexthop = nexthop->next) { |
d62a17ae | 947 | /* No protocol daemon provides src and so we're skipping |
948 | * tracking it */ | |
949 | prev_src = nexthop->rmap_src; | |
950 | prev_active = CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
951 | prev_index = nexthop->ifindex; | |
220f0f42 DS |
952 | /* |
953 | * We need to respect the multipath_num here | |
954 | * as that what we should be able to install from | |
955 | * a multipath perpsective should not be a data plane | |
956 | * decision point. | |
957 | */ | |
99eabcec | 958 | new_active = nexthop_active_check(rn, re, nexthop); |
b3f2b590 DS |
959 | if (new_active |
960 | && re->nexthop_active_num >= zrouter.multipath_num) { | |
220f0f42 DS |
961 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
962 | new_active = 0; | |
963 | } | |
964 | if (new_active) | |
d62a17ae | 965 | re->nexthop_active_num++; |
966 | /* Don't allow src setting on IPv6 addr for now */ | |
967 | if (prev_active != new_active || prev_index != nexthop->ifindex | |
968 | || ((nexthop->type >= NEXTHOP_TYPE_IFINDEX | |
969 | && nexthop->type < NEXTHOP_TYPE_IPV6) | |
970 | && prev_src.ipv4.s_addr | |
971 | != nexthop->rmap_src.ipv4.s_addr) | |
972 | || ((nexthop->type >= NEXTHOP_TYPE_IPV6 | |
973 | && nexthop->type < NEXTHOP_TYPE_BLACKHOLE) | |
974 | && !(IPV6_ADDR_SAME(&prev_src.ipv6, | |
6883bf8d DS |
975 | &nexthop->rmap_src.ipv6))) |
976 | || CHECK_FLAG(re->status, ROUTE_ENTRY_LABELS_CHANGED)) { | |
d62a17ae | 977 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); |
978 | SET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
979 | } | |
980 | } | |
981 | ||
d62a17ae | 982 | return re->nexthop_active_num; |
718e3744 | 983 | } |
6baeb988 | 984 | |
a64448ba DS |
985 | /* |
986 | * Is this RIB labeled-unicast? It must be of type BGP and all paths | |
987 | * (nexthops) must have a label. | |
988 | */ | |
d62a17ae | 989 | int zebra_rib_labeled_unicast(struct route_entry *re) |
a64448ba | 990 | { |
d62a17ae | 991 | struct nexthop *nexthop = NULL; |
a64448ba | 992 | |
d62a17ae | 993 | if (re->type != ZEBRA_ROUTE_BGP) |
994 | return 0; | |
a64448ba | 995 | |
7ee30f28 | 996 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d62a17ae | 997 | if (!nexthop->nh_label || !nexthop->nh_label->num_labels) |
998 | return 0; | |
6b0655a2 | 999 | |
d62a17ae | 1000 | return 1; |
a64448ba | 1001 | } |
718e3744 | 1002 | |
6ae24471 DS |
1003 | /* Update flag indicates whether this is a "replace" or not. Currently, this |
1004 | * is only used for IPv4. | |
1005 | */ | |
0c555cc6 DS |
1006 | void rib_install_kernel(struct route_node *rn, struct route_entry *re, |
1007 | struct route_entry *old) | |
718e3744 | 1008 | { |
d62a17ae | 1009 | struct nexthop *nexthop; |
1010 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 1011 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
97f5b441 MS |
1012 | const struct prefix *p, *src_p; |
1013 | enum zebra_dplane_result ret; | |
1014 | ||
1015 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
718e3744 | 1016 | |
d62a17ae | 1017 | srcdest_rnode_prefixes(rn, &p, &src_p); |
416ec78d | 1018 | |
d62a17ae | 1019 | if (info->safi != SAFI_UNICAST) { |
7ee30f28 | 1020 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d62a17ae | 1021 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 1022 | return; |
25b9cb0c DL |
1023 | } else { |
1024 | struct nexthop *prev; | |
1025 | ||
7ee30f28 | 1026 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
25b9cb0c | 1027 | UNSET_FLAG (nexthop->flags, NEXTHOP_FLAG_DUPLICATE); |
7ee30f28 | 1028 | for (ALL_NEXTHOPS(re->ng, prev)) { |
25b9cb0c DL |
1029 | if (prev == nexthop) |
1030 | break; | |
996c9314 LB |
1031 | if (nexthop_same_firsthop(nexthop, prev)) { |
1032 | SET_FLAG(nexthop->flags, | |
1033 | NEXTHOP_FLAG_DUPLICATE); | |
25b9cb0c DL |
1034 | break; |
1035 | } | |
1036 | } | |
1037 | } | |
d62a17ae | 1038 | } |
718e3744 | 1039 | |
2063a814 DS |
1040 | /* |
1041 | * If this is a replace to a new RE let the originator of the RE | |
1042 | * know that they've lost | |
1043 | */ | |
9a9f8926 | 1044 | if (old && (old != re) && (old->type != re->type)) |
28610f7e | 1045 | zsend_route_notify_owner(old, p, ZAPI_ROUTE_BETTER_ADMIN_WON); |
25b9cb0c | 1046 | |
97f5b441 MS |
1047 | /* Update fib selection */ |
1048 | dest->selected_fib = re; | |
1049 | ||
d62a17ae | 1050 | /* |
1051 | * Make sure we update the FPM any time we send new information to | |
1052 | * the kernel. | |
1053 | */ | |
1054 | hook_call(rib_update, rn, "installing in kernel"); | |
97f5b441 MS |
1055 | |
1056 | /* Send add or update */ | |
8263d1d0 | 1057 | if (old) |
97f5b441 | 1058 | ret = dplane_route_update(rn, re, old); |
5709131c | 1059 | else |
97f5b441 | 1060 | ret = dplane_route_add(rn, re); |
97f5b441 MS |
1061 | |
1062 | switch (ret) { | |
ea1c14f6 | 1063 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
60f98b23 DS |
1064 | SET_FLAG(re->status, ROUTE_ENTRY_QUEUED); |
1065 | if (old) | |
1066 | SET_FLAG(old->status, ROUTE_ENTRY_QUEUED); | |
97f5b441 MS |
1067 | if (zvrf) |
1068 | zvrf->installs_queued++; | |
1e885672 | 1069 | break; |
ea1c14f6 | 1070 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
1071 | { |
1072 | char str[SRCDEST2STR_BUFFER]; | |
1073 | ||
1074 | srcdest_rnode2str(rn, str, sizeof(str)); | |
1075 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
1076 | "%u:%s: Failed to enqueue dataplane install", | |
1077 | re->vrf_id, str); | |
1e885672 | 1078 | break; |
97f5b441 | 1079 | } |
ea1c14f6 | 1080 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
97f5b441 MS |
1081 | if (zvrf) |
1082 | zvrf->installs++; | |
1e885672 DS |
1083 | break; |
1084 | } | |
d62a17ae | 1085 | |
0c555cc6 | 1086 | return; |
718e3744 | 1087 | } |
1088 | ||
1089 | /* Uninstall the route from kernel. */ | |
0c555cc6 | 1090 | void rib_uninstall_kernel(struct route_node *rn, struct route_entry *re) |
718e3744 | 1091 | { |
d62a17ae | 1092 | struct nexthop *nexthop; |
1093 | rib_table_info_t *info = srcdest_rnode_table_info(rn); | |
d62a17ae | 1094 | struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id); |
05737783 | 1095 | |
d62a17ae | 1096 | if (info->safi != SAFI_UNICAST) { |
677c1dd5 | 1097 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
7ee30f28 | 1098 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
d6792f9d | 1099 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
0c555cc6 | 1100 | return; |
d62a17ae | 1101 | } |
416ec78d | 1102 | |
d62a17ae | 1103 | /* |
1104 | * Make sure we update the FPM any time we send new information to | |
97f5b441 | 1105 | * the dataplane. |
d62a17ae | 1106 | */ |
1107 | hook_call(rib_update, rn, "uninstalling from kernel"); | |
97f5b441 MS |
1108 | |
1109 | switch (dplane_route_delete(rn, re)) { | |
ea1c14f6 | 1110 | case ZEBRA_DPLANE_REQUEST_QUEUED: |
97f5b441 MS |
1111 | if (zvrf) |
1112 | zvrf->removals_queued++; | |
1e885672 | 1113 | break; |
ea1c14f6 | 1114 | case ZEBRA_DPLANE_REQUEST_FAILURE: |
97f5b441 MS |
1115 | { |
1116 | char str[SRCDEST2STR_BUFFER]; | |
1117 | ||
1118 | srcdest_rnode2str(rn, str, sizeof(str)); | |
1119 | flog_err(EC_ZEBRA_DP_INSTALL_FAIL, | |
1120 | "%u:%s: Failed to enqueue dataplane uninstall", | |
1121 | re->vrf_id, str); | |
1e885672 | 1122 | break; |
97f5b441 | 1123 | } |
ea1c14f6 | 1124 | case ZEBRA_DPLANE_REQUEST_SUCCESS: |
1e885672 DS |
1125 | if (zvrf) |
1126 | zvrf->removals++; | |
1127 | break; | |
1128 | } | |
718e3744 | 1129 | |
0c555cc6 | 1130 | return; |
718e3744 | 1131 | } |
1132 | ||
1133 | /* Uninstall the route from kernel. */ | |
d62a17ae | 1134 | static void rib_uninstall(struct route_node *rn, struct route_entry *re) |
718e3744 | 1135 | { |
d62a17ae | 1136 | rib_table_info_t *info = srcdest_rnode_table_info(rn); |
5f7a4718 | 1137 | rib_dest_t *dest = rib_dest_from_rnode(rn); |
97f5b441 | 1138 | struct nexthop *nexthop; |
416ec78d | 1139 | |
5f7a4718 | 1140 | if (dest && dest->selected_fib == re) { |
d62a17ae | 1141 | if (info->safi == SAFI_UNICAST) |
1142 | hook_call(rib_update, rn, "rib_uninstall"); | |
5adc2528 | 1143 | |
d62a17ae | 1144 | /* If labeled-unicast route, uninstall transit LSP. */ |
1145 | if (zebra_rib_labeled_unicast(re)) | |
1146 | zebra_mpls_lsp_uninstall(info->zvrf, rn, re); | |
7cdb1a84 | 1147 | |
3cdba47a | 1148 | rib_uninstall_kernel(rn, re); |
97f5b441 MS |
1149 | |
1150 | dest->selected_fib = NULL; | |
1151 | ||
1152 | for (ALL_NEXTHOPS(re->ng, nexthop)) | |
1153 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
d62a17ae | 1154 | } |
446bb95e | 1155 | |
d62a17ae | 1156 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { |
86391e56 MS |
1157 | const struct prefix *p, *src_p; |
1158 | ||
d62a17ae | 1159 | srcdest_rnode_prefixes(rn, &p, &src_p); |
05737783 | 1160 | |
d62a17ae | 1161 | redistribute_delete(p, src_p, re); |
1162 | UNSET_FLAG(re->flags, ZEBRA_FLAG_SELECTED); | |
1163 | } | |
718e3744 | 1164 | } |
1165 | ||
9fd92e3c AS |
1166 | /* |
1167 | * rib_can_delete_dest | |
1168 | * | |
1169 | * Returns TRUE if the given dest can be deleted from the table. | |
1170 | */ | |
d62a17ae | 1171 | static int rib_can_delete_dest(rib_dest_t *dest) |
9fd92e3c | 1172 | { |
045207e2 | 1173 | if (re_list_first(&dest->routes)) { |
d62a17ae | 1174 | return 0; |
1175 | } | |
9fd92e3c | 1176 | |
c86ba6c2 DS |
1177 | /* |
1178 | * Unresolved rnh's are stored on the default route's list | |
1179 | * | |
1180 | * dest->rnode can also be the source prefix node in an | |
1181 | * ipv6 sourcedest table. Fortunately the prefix of a | |
1182 | * source prefix node can never be the default prefix. | |
1183 | */ | |
1184 | if (is_default_prefix(&dest->rnode->p)) | |
1185 | return 0; | |
1186 | ||
d62a17ae | 1187 | /* |
1188 | * Don't delete the dest if we have to update the FPM about this | |
1189 | * prefix. | |
1190 | */ | |
1191 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_FPM) | |
1192 | || CHECK_FLAG(dest->flags, RIB_DEST_SENT_TO_FPM)) | |
1193 | return 0; | |
5adc2528 | 1194 | |
d62a17ae | 1195 | return 1; |
9fd92e3c AS |
1196 | } |
1197 | ||
699dae23 DS |
1198 | void zebra_rib_evaluate_rn_nexthops(struct route_node *rn, uint32_t seq) |
1199 | { | |
1200 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
699dae23 DS |
1201 | struct rnh *rnh; |
1202 | ||
1203 | /* | |
50872b08 | 1204 | * We are storing the rnh's associated withb |
699dae23 DS |
1205 | * the tracked nexthop as a list of the rn's. |
1206 | * Unresolved rnh's are placed at the top | |
1207 | * of the tree list.( 0.0.0.0/0 for v4 and 0::0/0 for v6 ) | |
1208 | * As such for each rn we need to walk up the tree | |
1209 | * and see if any rnh's need to see if they | |
1210 | * would match a more specific route | |
1211 | */ | |
1212 | while (rn) { | |
50872b08 DS |
1213 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
1214 | char buf[PREFIX_STRLEN]; | |
1215 | ||
1216 | zlog_debug("%s: %s Being examined for Nexthop Tracking", | |
1217 | __PRETTY_FUNCTION__, | |
1218 | srcdest_rnode2str(rn, buf, sizeof(buf))); | |
1219 | } | |
699dae23 DS |
1220 | if (!dest) { |
1221 | rn = rn->parent; | |
1222 | if (rn) | |
1223 | dest = rib_dest_from_rnode(rn); | |
1224 | continue; | |
1225 | } | |
1226 | /* | |
1227 | * If we have any rnh's stored in the nht list | |
1228 | * then we know that this route node was used for | |
1229 | * nht resolution and as such we need to call the | |
1230 | * nexthop tracking evaluation code | |
1231 | */ | |
81fddbe7 | 1232 | frr_each (rnh_list, &dest->nht, rnh) { |
699dae23 DS |
1233 | struct zebra_vrf *zvrf = |
1234 | zebra_vrf_lookup_by_id(rnh->vrf_id); | |
1235 | struct prefix *p = &rnh->node->p; | |
1236 | ||
50872b08 | 1237 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) { |
699dae23 DS |
1238 | char buf1[PREFIX_STRLEN]; |
1239 | char buf2[PREFIX_STRLEN]; | |
1240 | ||
1241 | zlog_debug("%u:%s has Nexthop(%s) depending on it, evaluating %u:%u", | |
1242 | zvrf->vrf->vrf_id, | |
50872b08 | 1243 | srcdest_rnode2str(rn, buf1, |
699dae23 DS |
1244 | sizeof(buf1)), |
1245 | prefix2str(p, buf2, sizeof(buf2)), | |
1246 | seq, rnh->seqno); | |
1247 | } | |
1248 | ||
1249 | /* | |
1250 | * If we have evaluated this node on this pass | |
1251 | * already, due to following the tree up | |
1252 | * then we know that we can move onto the next | |
1253 | * rnh to process. | |
1254 | * | |
1255 | * Additionally we call zebra_evaluate_rnh | |
1256 | * when we gc the dest. In this case we know | |
1257 | * that there must be no other re's where | |
1258 | * we were originally as such we know that | |
1259 | * that sequence number is ok to respect. | |
1260 | */ | |
50872b08 DS |
1261 | if (rnh->seqno == seq) { |
1262 | if (IS_ZEBRA_DEBUG_NHT_DETAILED) | |
1263 | zlog_debug( | |
1264 | "\tNode processed and moved already"); | |
699dae23 | 1265 | continue; |
50872b08 | 1266 | } |
699dae23 DS |
1267 | |
1268 | rnh->seqno = seq; | |
1269 | zebra_evaluate_rnh(zvrf, family2afi(p->family), 0, | |
1270 | rnh->type, p); | |
1271 | } | |
1272 | ||
1273 | rn = rn->parent; | |
1274 | if (rn) | |
1275 | dest = rib_dest_from_rnode(rn); | |
1276 | } | |
1277 | } | |
1278 | ||
9fd92e3c AS |
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 | |
699dae23 DS |
1305 | zebra_rib_evaluate_rn_nexthops(rn, zebra_router_get_next_sequence()); |
1306 | ||
d62a17ae | 1307 | dest->rnode = NULL; |
aa57abfb | 1308 | rnh_list_fini(&dest->nht); |
d62a17ae | 1309 | XFREE(MTYPE_RIB_DEST, dest); |
1310 | rn->info = NULL; | |
9fd92e3c | 1311 | |
d62a17ae | 1312 | /* |
1313 | * Release the one reference that we keep on the route node. | |
1314 | */ | |
1315 | route_unlock_node(rn); | |
1316 | return 1; | |
9fd92e3c AS |
1317 | } |
1318 | ||
d62a17ae | 1319 | static void rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1320 | struct route_entry *new) | |
3e5c6e00 | 1321 | { |
d62a17ae | 1322 | hook_call(rib_update, rn, "new route selected"); |
3e5c6e00 | 1323 | |
d62a17ae | 1324 | /* Update real nexthop. This may actually determine if nexthop is active |
1325 | * or not. */ | |
6883bf8d | 1326 | if (!nexthop_group_active_nexthop_num(&new->ng)) { |
d62a17ae | 1327 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
1328 | return; | |
1329 | } | |
3e5c6e00 | 1330 | |
d62a17ae | 1331 | if (IS_ZEBRA_DEBUG_RIB) { |
1332 | char buf[SRCDEST2STR_BUFFER]; | |
1333 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
1334 | zlog_debug("%u:%s: Adding route rn %p, re %p (%s)", |
1335 | zvrf_id(zvrf), buf, rn, new, | |
1336 | zebra_route_string(new->type)); | |
d62a17ae | 1337 | } |
3e5c6e00 | 1338 | |
d62a17ae | 1339 | /* If labeled-unicast route, install transit LSP. */ |
1340 | if (zebra_rib_labeled_unicast(new)) | |
1341 | zebra_mpls_lsp_install(zvrf, rn, new); | |
a64448ba | 1342 | |
3cdba47a | 1343 | rib_install_kernel(rn, new, NULL); |
3e5c6e00 | 1344 | |
d62a17ae | 1345 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1346 | } |
1347 | ||
d62a17ae | 1348 | static void rib_process_del_fib(struct zebra_vrf *zvrf, struct route_node *rn, |
1349 | struct route_entry *old) | |
3e5c6e00 | 1350 | { |
d62a17ae | 1351 | hook_call(rib_update, rn, "removing existing route"); |
3e5c6e00 | 1352 | |
d62a17ae | 1353 | /* Uninstall from kernel. */ |
1354 | if (IS_ZEBRA_DEBUG_RIB) { | |
1355 | char buf[SRCDEST2STR_BUFFER]; | |
1356 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
1357 | zlog_debug("%u:%s: Deleting route rn %p, re %p (%s)", |
1358 | zvrf_id(zvrf), buf, rn, old, | |
1359 | zebra_route_string(old->type)); | |
d62a17ae | 1360 | } |
3e5c6e00 | 1361 | |
d62a17ae | 1362 | /* If labeled-unicast route, uninstall transit LSP. */ |
1363 | if (zebra_rib_labeled_unicast(old)) | |
1364 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1365 | ||
3cdba47a | 1366 | rib_uninstall_kernel(rn, old); |
d62a17ae | 1367 | |
1368 | /* Update nexthop for route, reset changed flag. */ | |
ef57f35f DL |
1369 | /* Note: this code also handles the Linux case when an interface goes |
1370 | * down, causing the kernel to delete routes without sending DELROUTE | |
1371 | * notifications | |
1372 | */ | |
6883bf8d | 1373 | if (RIB_KERNEL_ROUTE(old)) |
212df1de PG |
1374 | SET_FLAG(old->status, ROUTE_ENTRY_REMOVED); |
1375 | else | |
1376 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1377 | } |
1378 | ||
1379 | static void rib_process_update_fib(struct zebra_vrf *zvrf, | |
1380 | struct route_node *rn, | |
1381 | struct route_entry *old, | |
1382 | struct route_entry *new) | |
1383 | { | |
d62a17ae | 1384 | int nh_active = 0; |
d62a17ae | 1385 | |
1386 | /* | |
1387 | * We have to install or update if a new route has been selected or | |
1388 | * something has changed. | |
1389 | */ | |
1390 | if (new != old || CHECK_FLAG(new->status, ROUTE_ENTRY_CHANGED)) { | |
1391 | hook_call(rib_update, rn, "updating existing route"); | |
1392 | ||
1393 | /* Update the nexthop; we could determine here that nexthop is | |
1394 | * inactive. */ | |
6883bf8d | 1395 | if (nexthop_group_active_nexthop_num(&new->ng)) |
d62a17ae | 1396 | nh_active = 1; |
1397 | ||
1398 | /* If nexthop is active, install the selected route, if | |
1399 | * appropriate. If | |
1400 | * the install succeeds, cleanup flags for prior route, if | |
1401 | * different from | |
1402 | * newly selected. | |
1403 | */ | |
1404 | if (nh_active) { | |
1405 | if (IS_ZEBRA_DEBUG_RIB) { | |
1406 | char buf[SRCDEST2STR_BUFFER]; | |
1407 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1408 | if (new != old) | |
1409 | zlog_debug( | |
2da33d6b | 1410 | "%u:%s: Updating route rn %p, re %p (%s) old %p (%s)", |
d62a17ae | 1411 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
1412 | zebra_route_string(new->type), |
1413 | old, | |
1414 | zebra_route_string(old->type)); | |
d62a17ae | 1415 | else |
1416 | zlog_debug( | |
2da33d6b | 1417 | "%u:%s: Updating route rn %p, re %p (%s)", |
d62a17ae | 1418 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 1419 | zebra_route_string(new->type)); |
d62a17ae | 1420 | } |
1421 | ||
1422 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1423 | if (zebra_rib_labeled_unicast(old)) | |
1424 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1425 | ||
3cdba47a DS |
1426 | /* |
1427 | * Non-system route should be installed. | |
1428 | * If labeled-unicast route, install transit | |
1429 | * LSP. | |
1430 | */ | |
1431 | if (zebra_rib_labeled_unicast(new)) | |
1432 | zebra_mpls_lsp_install(zvrf, rn, new); | |
d62a17ae | 1433 | |
3cdba47a | 1434 | rib_install_kernel(rn, new, old); |
d62a17ae | 1435 | } |
a64448ba | 1436 | |
d62a17ae | 1437 | /* |
1438 | * If nexthop for selected route is not active or install | |
1439 | * failed, we | |
1440 | * may need to uninstall and delete for redistribution. | |
1441 | */ | |
ed216282 | 1442 | if (!nh_active) { |
d62a17ae | 1443 | if (IS_ZEBRA_DEBUG_RIB) { |
1444 | char buf[SRCDEST2STR_BUFFER]; | |
1445 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
1446 | if (new != old) | |
1447 | zlog_debug( | |
2da33d6b | 1448 | "%u:%s: Deleting route rn %p, re %p (%s) old %p (%s) - nexthop inactive", |
d62a17ae | 1449 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b DS |
1450 | zebra_route_string(new->type), |
1451 | old, | |
1452 | zebra_route_string(old->type)); | |
d62a17ae | 1453 | else |
1454 | zlog_debug( | |
2da33d6b | 1455 | "%u:%s: Deleting route rn %p, re %p (%s) - nexthop inactive", |
d62a17ae | 1456 | zvrf_id(zvrf), buf, rn, new, |
2da33d6b | 1457 | zebra_route_string(new->type)); |
d62a17ae | 1458 | } |
1459 | ||
1460 | /* If labeled-unicast route, uninstall transit LSP. */ | |
1461 | if (zebra_rib_labeled_unicast(old)) | |
1462 | zebra_mpls_lsp_uninstall(zvrf, rn, old); | |
1463 | ||
3cdba47a | 1464 | rib_uninstall_kernel(rn, old); |
d62a17ae | 1465 | } |
1466 | } else { | |
1467 | /* | |
1468 | * Same route selected; check if in the FIB and if not, | |
3cdba47a DS |
1469 | * re-install. This is housekeeping code to deal with |
1470 | * race conditions in kernel with linux netlink reporting | |
1471 | * interface up before IPv4 or IPv6 protocol is ready | |
d62a17ae | 1472 | * to add routes. |
1473 | */ | |
3cdba47a DS |
1474 | if (!CHECK_FLAG(new->status, ROUTE_ENTRY_INSTALLED) || |
1475 | RIB_SYSTEM_ROUTE(new)) | |
677c1dd5 | 1476 | rib_install_kernel(rn, new, NULL); |
d62a17ae | 1477 | } |
1478 | ||
1479 | /* Update prior route. */ | |
6883bf8d | 1480 | if (new != old) |
d62a17ae | 1481 | UNSET_FLAG(old->status, ROUTE_ENTRY_CHANGED); |
3e5c6e00 | 1482 | |
d62a17ae | 1483 | /* Clear changed flag. */ |
1484 | UNSET_FLAG(new->status, ROUTE_ENTRY_CHANGED); | |
3e5c6e00 | 1485 | } |
1486 | ||
d62a17ae | 1487 | /* Check if 'alternate' RIB entry is better than 'current'. */ |
1488 | static struct route_entry *rib_choose_best(struct route_entry *current, | |
1489 | struct route_entry *alternate) | |
1490 | { | |
1491 | if (current == NULL) | |
1492 | return alternate; | |
1493 | ||
1494 | /* filter route selection in following order: | |
1495 | * - connected beats other types | |
fec4ca19 | 1496 | * - if both connected, loopback or vrf wins |
d62a17ae | 1497 | * - lower distance beats higher |
1498 | * - lower metric beats higher for equal distance | |
1499 | * - last, hence oldest, route wins tie break. | |
1500 | */ | |
1501 | ||
fec4ca19 DS |
1502 | /* Connected routes. Check to see if either are a vrf |
1503 | * or loopback interface. If not, pick the last connected | |
d62a17ae | 1504 | * route of the set of lowest metric connected routes. |
1505 | */ | |
1506 | if (alternate->type == ZEBRA_ROUTE_CONNECT) { | |
fec4ca19 DS |
1507 | if (current->type != ZEBRA_ROUTE_CONNECT) |
1508 | return alternate; | |
1509 | ||
1510 | /* both are connected. are either loop or vrf? */ | |
1511 | struct nexthop *nexthop = NULL; | |
1512 | ||
1513 | for (ALL_NEXTHOPS(alternate->ng, nexthop)) { | |
1514 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1515 | nexthop->ifindex, alternate->vrf_id))) | |
1516 | return alternate; | |
1517 | } | |
1518 | ||
1519 | for (ALL_NEXTHOPS(current->ng, nexthop)) { | |
1520 | if (if_is_loopback_or_vrf(if_lookup_by_index( | |
1521 | nexthop->ifindex, current->vrf_id))) | |
1522 | return current; | |
1523 | } | |
1524 | ||
1525 | /* Neither are loop or vrf so pick best metric */ | |
1526 | if (alternate->metric <= current->metric) | |
d62a17ae | 1527 | return alternate; |
1528 | ||
1529 | return current; | |
1530 | } | |
3e5c6e00 | 1531 | |
d62a17ae | 1532 | if (current->type == ZEBRA_ROUTE_CONNECT) |
1533 | return current; | |
3e5c6e00 | 1534 | |
d62a17ae | 1535 | /* higher distance loses */ |
1536 | if (alternate->distance < current->distance) | |
1537 | return alternate; | |
1538 | if (current->distance < alternate->distance) | |
1539 | return current; | |
3e5c6e00 | 1540 | |
d62a17ae | 1541 | /* metric tie-breaks equal distance */ |
1542 | if (alternate->metric <= current->metric) | |
1543 | return alternate; | |
3e5c6e00 | 1544 | |
d62a17ae | 1545 | return current; |
3e5c6e00 | 1546 | } |
1547 | ||
d62a17ae | 1548 | /* Core function for processing routing information base. */ |
1549 | static void rib_process(struct route_node *rn) | |
1550 | { | |
1551 | struct route_entry *re; | |
1552 | struct route_entry *next; | |
1553 | struct route_entry *old_selected = NULL; | |
1554 | struct route_entry *new_selected = NULL; | |
1555 | struct route_entry *old_fib = NULL; | |
1556 | struct route_entry *new_fib = NULL; | |
1557 | struct route_entry *best = NULL; | |
1558 | char buf[SRCDEST2STR_BUFFER]; | |
1559 | rib_dest_t *dest; | |
1560 | struct zebra_vrf *zvrf = NULL; | |
86391e56 MS |
1561 | const struct prefix *p, *src_p; |
1562 | ||
d62a17ae | 1563 | srcdest_rnode_prefixes(rn, &p, &src_p); |
1564 | vrf_id_t vrf_id = VRF_UNKNOWN; | |
1565 | ||
1566 | assert(rn); | |
1567 | ||
1568 | dest = rib_dest_from_rnode(rn); | |
1569 | if (dest) { | |
1570 | zvrf = rib_dest_vrf(dest); | |
1571 | vrf_id = zvrf_id(zvrf); | |
1572 | } | |
bab85d4f | 1573 | |
d62a17ae | 1574 | if (IS_ZEBRA_DEBUG_RIB) |
1575 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
bab85d4f | 1576 | |
d62a17ae | 1577 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1578 | zlog_debug("%u:%s: Processing rn %p", vrf_id, buf, rn); | |
bab85d4f | 1579 | |
607425e5 DS |
1580 | /* |
1581 | * we can have rn's that have a NULL info pointer | |
1582 | * (dest). As such let's not let the deref happen | |
1583 | * additionally we know RNODE_FOREACH_RE_SAFE | |
1584 | * will not iterate so we are ok. | |
1585 | */ | |
1586 | if (dest) | |
1587 | old_fib = dest->selected_fib; | |
5f7a4718 | 1588 | |
a2addae8 | 1589 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1590 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1591 | zlog_debug( | |
2da33d6b DS |
1592 | "%u:%s: Examine re %p (%s) status %x flags %x dist %d metric %d", |
1593 | vrf_id, buf, re, zebra_route_string(re->type), | |
1594 | re->status, re->flags, re->distance, | |
1595 | re->metric); | |
d62a17ae | 1596 | |
1597 | UNSET_FLAG(re->status, ROUTE_ENTRY_NEXTHOPS_CHANGED); | |
1598 | ||
1599 | /* Currently selected re. */ | |
1600 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED)) { | |
1601 | assert(old_selected == NULL); | |
1602 | old_selected = re; | |
1603 | } | |
bab85d4f | 1604 | |
d62a17ae | 1605 | /* Skip deleted entries from selection */ |
1606 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1607 | continue; | |
1608 | ||
1609 | /* Skip unreachable nexthop. */ | |
1610 | /* This first call to nexthop_active_update is merely to | |
6883bf8d DS |
1611 | * determine if there's any change to nexthops associated |
1612 | * with this RIB entry. Now, rib_process() can be invoked due | |
1613 | * to an external event such as link down or due to | |
1614 | * next-hop-tracking evaluation. In the latter case, | |
d62a17ae | 1615 | * a decision has already been made that the NHs have changed. |
6883bf8d DS |
1616 | * So, no need to invoke a potentially expensive call again. |
1617 | * Further, since the change might be in a recursive NH which | |
1618 | * is not caught in the nexthop_active_update() code. Thus, we | |
1619 | * might miss changes to recursive NHs. | |
d62a17ae | 1620 | */ |
6883bf8d | 1621 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED) |
99eabcec | 1622 | && !nexthop_active_update(rn, re)) { |
d62a17ae | 1623 | if (re->type == ZEBRA_ROUTE_TABLE) { |
1624 | /* XXX: HERE BE DRAGONS!!!!! | |
1625 | * In all honesty, I have not yet figured out | |
6883bf8d DS |
1626 | * what this part does or why the |
1627 | * ROUTE_ENTRY_CHANGED test above is correct | |
d62a17ae | 1628 | * or why we need to delete a route here, and |
6883bf8d DS |
1629 | * also not whether this concerns both selected |
1630 | * and fib route, or only selected | |
1631 | * or only fib | |
1632 | * | |
1633 | * This entry was denied by the 'ip protocol | |
1634 | * table' route-map, we need to delete it */ | |
d62a17ae | 1635 | if (re != old_selected) { |
1636 | if (IS_ZEBRA_DEBUG_RIB) | |
1637 | zlog_debug( | |
32391aff | 1638 | "%s: %u:%s: imported via import-table but denied " |
d62a17ae | 1639 | "by the ip protocol table route-map", |
32391aff | 1640 | __func__, vrf_id, buf); |
d62a17ae | 1641 | rib_unlink(rn, re); |
1642 | } else | |
1643 | SET_FLAG(re->status, | |
1644 | ROUTE_ENTRY_REMOVED); | |
1645 | } | |
1646 | ||
1647 | continue; | |
1648 | } | |
bab85d4f | 1649 | |
d62a17ae | 1650 | /* Infinite distance. */ |
1651 | if (re->distance == DISTANCE_INFINITY) { | |
1652 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1653 | continue; | |
1654 | } | |
bab85d4f | 1655 | |
d62a17ae | 1656 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_FIB_OVERRIDE)) { |
1657 | best = rib_choose_best(new_fib, re); | |
1658 | if (new_fib && best != new_fib) | |
1659 | UNSET_FLAG(new_fib->status, | |
1660 | ROUTE_ENTRY_CHANGED); | |
1661 | new_fib = best; | |
1662 | } else { | |
1663 | best = rib_choose_best(new_selected, re); | |
1664 | if (new_selected && best != new_selected) | |
1665 | UNSET_FLAG(new_selected->status, | |
1666 | ROUTE_ENTRY_CHANGED); | |
1667 | new_selected = best; | |
1668 | } | |
1669 | if (best != re) | |
1670 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
1671 | } /* RNODE_FOREACH_RE */ | |
1672 | ||
1673 | /* If no FIB override route, use the selected route also for FIB */ | |
1674 | if (new_fib == NULL) | |
1675 | new_fib = new_selected; | |
1676 | ||
1677 | /* After the cycle is finished, the following pointers will be set: | |
1678 | * old_selected --- RE entry currently having SELECTED | |
1679 | * new_selected --- RE entry that is newly SELECTED | |
1680 | * old_fib --- RE entry currently in kernel FIB | |
1681 | * new_fib --- RE entry that is newly to be in kernel FIB | |
1682 | * | |
1683 | * new_selected will get SELECTED flag, and is going to be redistributed | |
1684 | * the zclients. new_fib (which can be new_selected) will be installed | |
1685 | * in kernel. | |
1686 | */ | |
1687 | ||
1688 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
1689 | zlog_debug( | |
1690 | "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p", | |
1691 | vrf_id, buf, (void *)old_selected, (void *)new_selected, | |
1692 | (void *)old_fib, (void *)new_fib); | |
1693 | } | |
446bb95e | 1694 | |
d62a17ae | 1695 | /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if |
1696 | * fib == selected */ | |
9d303b37 DL |
1697 | bool selected_changed = new_selected && CHECK_FLAG(new_selected->status, |
1698 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 1699 | |
1700 | /* Update fib according to selection results */ | |
1701 | if (new_fib && old_fib) | |
1702 | rib_process_update_fib(zvrf, rn, old_fib, new_fib); | |
1703 | else if (new_fib) | |
1704 | rib_process_add_fib(zvrf, rn, new_fib); | |
1705 | else if (old_fib) | |
1706 | rib_process_del_fib(zvrf, rn, old_fib); | |
1707 | ||
8cb41cd6 | 1708 | /* Update SELECTED entry */ |
d62a17ae | 1709 | if (old_selected != new_selected || selected_changed) { |
93bdadae | 1710 | |
6883bf8d | 1711 | if (new_selected && new_selected != new_fib) |
d62a17ae | 1712 | UNSET_FLAG(new_selected->status, ROUTE_ENTRY_CHANGED); |
41ec9222 | 1713 | |
3cdba47a | 1714 | if (new_selected) |
5af4b346 MS |
1715 | SET_FLAG(new_selected->flags, ZEBRA_FLAG_SELECTED); |
1716 | ||
d62a17ae | 1717 | if (old_selected) { |
1718 | if (!new_selected) | |
1719 | redistribute_delete(p, src_p, old_selected); | |
1720 | if (old_selected != new_selected) | |
1721 | UNSET_FLAG(old_selected->flags, | |
1722 | ZEBRA_FLAG_SELECTED); | |
f857321e | 1723 | } |
d62a17ae | 1724 | } |
3e5c6e00 | 1725 | |
d62a17ae | 1726 | /* Remove all RE entries queued for removal */ |
a2addae8 | 1727 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 1728 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { |
1729 | if (IS_ZEBRA_DEBUG_RIB) { | |
1730 | rnode_debug(rn, vrf_id, "rn %p, removing re %p", | |
1731 | (void *)rn, (void *)re); | |
1732 | } | |
1733 | rib_unlink(rn, re); | |
1734 | } | |
1735 | } | |
4d38fdb4 | 1736 | |
d62a17ae | 1737 | /* |
1738 | * Check if the dest can be deleted now. | |
1739 | */ | |
1740 | rib_gc_dest(rn); | |
e96f9203 DO |
1741 | } |
1742 | ||
a1494c25 DS |
1743 | static void zebra_rib_evaluate_mpls(struct route_node *rn) |
1744 | { | |
1745 | rib_dest_t *dest = rib_dest_from_rnode(rn); | |
1746 | struct zebra_vrf *zvrf = vrf_info_lookup(VRF_DEFAULT); | |
1747 | ||
1748 | if (!dest) | |
1749 | return; | |
1750 | ||
1751 | if (CHECK_FLAG(dest->flags, RIB_DEST_UPDATE_LSPS)) { | |
1752 | if (IS_ZEBRA_DEBUG_MPLS) | |
1753 | zlog_debug( | |
1754 | "%u: Scheduling all LSPs upon RIB completion", | |
1755 | zvrf_id(zvrf)); | |
1756 | zebra_mpls_lsp_schedule(zvrf); | |
1757 | mpls_unmark_lsps_for_processing(rn); | |
1758 | } | |
1759 | } | |
1760 | ||
e5ac2adf MS |
1761 | /* |
1762 | * Utility to match route with dplane context data | |
1763 | */ | |
1764 | static bool rib_route_match_ctx(const struct route_entry *re, | |
25779064 MS |
1765 | const struct zebra_dplane_ctx *ctx, |
1766 | bool is_update) | |
e5ac2adf MS |
1767 | { |
1768 | bool result = false; | |
1769 | ||
1770 | if (is_update) { | |
1771 | /* | |
1772 | * In 'update' case, we test info about the 'previous' or | |
1773 | * 'old' route | |
1774 | */ | |
1775 | if ((re->type == dplane_ctx_get_old_type(ctx)) && | |
1776 | (re->instance == dplane_ctx_get_old_instance(ctx))) { | |
1777 | result = true; | |
1778 | ||
1779 | /* TODO -- we're using this extra test, but it's not | |
1780 | * exactly clear why. | |
1781 | */ | |
1782 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1783 | (re->distance != dplane_ctx_get_old_distance(ctx) || | |
1784 | re->tag != dplane_ctx_get_old_tag(ctx))) { | |
1785 | result = false; | |
1786 | } | |
1787 | } | |
1788 | ||
1789 | } else { | |
1790 | /* | |
1791 | * Ordinary, single-route case using primary context info | |
1792 | */ | |
1793 | if ((dplane_ctx_get_op(ctx) != DPLANE_OP_ROUTE_DELETE) && | |
1794 | CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
1795 | /* Skip route that's been deleted */ | |
1796 | goto done; | |
1797 | } | |
1798 | ||
1799 | if ((re->type == dplane_ctx_get_type(ctx)) && | |
1800 | (re->instance == dplane_ctx_get_instance(ctx))) { | |
1801 | result = true; | |
1802 | ||
1803 | /* TODO -- we're using this extra test, but it's not | |
1804 | * exactly clear why. | |
1805 | */ | |
1806 | if (re->type == ZEBRA_ROUTE_STATIC && | |
1807 | (re->distance != dplane_ctx_get_distance(ctx) || | |
1808 | re->tag != dplane_ctx_get_tag(ctx))) { | |
1809 | result = false; | |
1810 | } | |
1811 | } | |
1812 | } | |
1813 | ||
1814 | done: | |
1815 | ||
1816 | return (result); | |
1817 | } | |
1818 | ||
3cdba47a DS |
1819 | static void zebra_rib_fixup_system(struct route_node *rn) |
1820 | { | |
1821 | struct route_entry *re; | |
1822 | ||
1823 | RNODE_FOREACH_RE(rn, re) { | |
1824 | struct nexthop *nhop; | |
1825 | ||
1826 | if (!RIB_SYSTEM_ROUTE(re)) | |
1827 | continue; | |
1828 | ||
1829 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
1830 | continue; | |
1831 | ||
1832 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
1833 | ||
1834 | for (ALL_NEXTHOPS(re->ng, nhop)) { | |
1835 | if (CHECK_FLAG(nhop->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1836 | continue; | |
1837 | ||
1838 | SET_FLAG(nhop->flags, NEXTHOP_FLAG_FIB); | |
1839 | } | |
1840 | } | |
1841 | } | |
1842 | ||
78bf56b0 MS |
1843 | /* |
1844 | * Update a route from a dplane context. This consolidates common code | |
1845 | * that can be used in processing of results from FIB updates, and in | |
1846 | * async notification processing. | |
1847 | */ | |
1848 | static int rib_update_re_from_ctx(struct route_entry *re, | |
1849 | struct route_node *rn, | |
1850 | const struct zebra_dplane_ctx *ctx) | |
1851 | { | |
1852 | int result = 0; | |
1853 | struct nexthop *nexthop, *ctx_nexthop; | |
1854 | const struct prefix *dest_pfx, *src_pfx; | |
1855 | ||
1856 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); | |
1857 | ||
1858 | /* Update zebra nexthop FIB flag for each | |
1859 | * nexthop that was installed. | |
1860 | */ | |
5695d9ac | 1861 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
78bf56b0 | 1862 | |
5695d9ac | 1863 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE)) |
78bf56b0 MS |
1864 | continue; |
1865 | ||
5695d9ac MS |
1866 | for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), |
1867 | ctx_nexthop)) { | |
1868 | ||
78bf56b0 MS |
1869 | if (nexthop_same(ctx_nexthop, nexthop)) |
1870 | break; | |
1871 | } | |
1872 | ||
5695d9ac MS |
1873 | /* If the FIB doesn't know about the nexthop, |
1874 | * it's not installed | |
1875 | */ | |
1876 | if (ctx_nexthop == NULL) { | |
1877 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
78bf56b0 | 1878 | continue; |
5695d9ac | 1879 | } |
78bf56b0 | 1880 | |
5695d9ac MS |
1881 | if (CHECK_FLAG(ctx_nexthop->flags, NEXTHOP_FLAG_FIB)) |
1882 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); | |
78bf56b0 | 1883 | else |
5695d9ac | 1884 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
78bf56b0 MS |
1885 | } |
1886 | ||
1887 | /* Redistribute */ | |
1888 | redistribute_update(dest_pfx, src_pfx, re, NULL); | |
1889 | ||
1890 | return result; | |
1891 | } | |
1892 | ||
e5ac2adf | 1893 | /* |
fad4d69c MS |
1894 | * Helper to locate a zebra route-node from a dplane context. This is used |
1895 | * when processing dplane results, e.g. Note well: the route-node is returned | |
1896 | * with a ref held - route_unlock_node() must be called eventually. | |
e5ac2adf | 1897 | */ |
fad4d69c MS |
1898 | static struct route_node * |
1899 | rib_find_rn_from_ctx(const struct zebra_dplane_ctx *ctx) | |
e5ac2adf MS |
1900 | { |
1901 | struct route_table *table = NULL; | |
1902 | struct route_node *rn = NULL; | |
e5ac2adf MS |
1903 | const struct prefix *dest_pfx, *src_pfx; |
1904 | ||
1905 | /* Locate rn and re(s) from ctx */ | |
1906 | ||
1907 | table = zebra_vrf_table_with_table_id(dplane_ctx_get_afi(ctx), | |
1908 | dplane_ctx_get_safi(ctx), | |
1909 | dplane_ctx_get_vrf(ctx), | |
1910 | dplane_ctx_get_table(ctx)); | |
1911 | if (table == NULL) { | |
1912 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fad4d69c | 1913 | zlog_debug("Failed to find route for ctx: no table for afi %d, safi %d, vrf %u", |
e5ac2adf MS |
1914 | dplane_ctx_get_afi(ctx), |
1915 | dplane_ctx_get_safi(ctx), | |
1916 | dplane_ctx_get_vrf(ctx)); | |
1917 | } | |
1918 | goto done; | |
1919 | } | |
1920 | ||
fad4d69c MS |
1921 | dest_pfx = dplane_ctx_get_dest(ctx); |
1922 | src_pfx = dplane_ctx_get_src(ctx); | |
1923 | ||
1924 | rn = srcdest_rnode_get(table, dest_pfx, | |
1925 | src_pfx ? (struct prefix_ipv6 *)src_pfx : NULL); | |
1926 | ||
1927 | done: | |
1928 | return rn; | |
1929 | } | |
1930 | ||
1931 | ||
97f5b441 | 1932 | |
fad4d69c MS |
1933 | /* |
1934 | * Route-update results processing after async dataplane update. | |
1935 | */ | |
1936 | static void rib_process_result(struct zebra_dplane_ctx *ctx) | |
1937 | { | |
1938 | struct zebra_vrf *zvrf = NULL; | |
1939 | struct route_node *rn = NULL; | |
1940 | struct route_entry *re = NULL, *old_re = NULL, *rib; | |
1941 | bool is_update = false; | |
1942 | char dest_str[PREFIX_STRLEN] = ""; | |
1943 | enum dplane_op_e op; | |
1944 | enum zebra_dplane_result status; | |
1945 | const struct prefix *dest_pfx, *src_pfx; | |
1946 | uint32_t seq; | |
1947 | ||
1948 | zvrf = vrf_info_lookup(dplane_ctx_get_vrf(ctx)); | |
e5ac2adf MS |
1949 | dest_pfx = dplane_ctx_get_dest(ctx); |
1950 | ||
1951 | /* Note well: only capturing the prefix string if debug is enabled here; | |
1952 | * unconditional log messages will have to generate the string. | |
1953 | */ | |
5709131c | 1954 | if (IS_ZEBRA_DEBUG_DPLANE) |
e5ac2adf | 1955 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); |
e5ac2adf | 1956 | |
fad4d69c MS |
1957 | /* Locate rn and re(s) from ctx */ |
1958 | rn = rib_find_rn_from_ctx(ctx); | |
e5ac2adf MS |
1959 | if (rn == NULL) { |
1960 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
fe2c53d4 | 1961 | zlog_debug("Failed to process dplane results: no route for %u:%s", |
e5ac2adf MS |
1962 | dplane_ctx_get_vrf(ctx), dest_str); |
1963 | } | |
1964 | goto done; | |
1965 | } | |
1966 | ||
1967 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); | |
1968 | ||
1969 | op = dplane_ctx_get_op(ctx); | |
1970 | status = dplane_ctx_get_status(ctx); | |
1971 | ||
c831033f | 1972 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
f183e380 | 1973 | zlog_debug("%u:%s Processing dplane ctx %p, op %s result %s", |
e5ac2adf | 1974 | dplane_ctx_get_vrf(ctx), dest_str, ctx, |
f183e380 | 1975 | dplane_op2str(op), dplane_res2str(status)); |
e5ac2adf | 1976 | |
e5ac2adf MS |
1977 | /* |
1978 | * Update is a bit of a special case, where we may have both old and new | |
1979 | * routes to post-process. | |
1980 | */ | |
1981 | is_update = dplane_ctx_is_update(ctx); | |
1982 | ||
1983 | /* | |
1984 | * Take a pass through the routes, look for matches with the context | |
1985 | * info. | |
1986 | */ | |
1987 | RNODE_FOREACH_RE(rn, rib) { | |
1988 | ||
1989 | if (re == NULL) { | |
5709131c | 1990 | if (rib_route_match_ctx(rib, ctx, false)) |
e5ac2adf | 1991 | re = rib; |
e5ac2adf MS |
1992 | } |
1993 | ||
1994 | /* Check for old route match */ | |
1995 | if (is_update && (old_re == NULL)) { | |
5709131c | 1996 | if (rib_route_match_ctx(rib, ctx, true /*is_update*/)) |
e5ac2adf | 1997 | old_re = rib; |
e5ac2adf MS |
1998 | } |
1999 | ||
2000 | /* Have we found the routes we need to work on? */ | |
5709131c | 2001 | if (re && ((!is_update || old_re))) |
e5ac2adf | 2002 | break; |
e5ac2adf MS |
2003 | } |
2004 | ||
699dae23 DS |
2005 | seq = dplane_ctx_get_seq(ctx); |
2006 | ||
e5ac2adf MS |
2007 | /* |
2008 | * Check sequence number(s) to detect stale results before continuing | |
2009 | */ | |
60f98b23 | 2010 | if (re) { |
699dae23 | 2011 | if (re->dplane_sequence != seq) { |
60f98b23 DS |
2012 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) |
2013 | zlog_debug("%u:%s Stale dplane result for re %p", | |
2014 | dplane_ctx_get_vrf(ctx), | |
2015 | dest_str, re); | |
2016 | } else | |
2017 | UNSET_FLAG(re->status, ROUTE_ENTRY_QUEUED); | |
e5ac2adf MS |
2018 | } |
2019 | ||
60f98b23 DS |
2020 | if (old_re) { |
2021 | if (old_re->dplane_sequence != dplane_ctx_get_old_seq(ctx)) { | |
2022 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
2023 | zlog_debug("%u:%s Stale dplane result for old_re %p", | |
2024 | dplane_ctx_get_vrf(ctx), | |
2025 | dest_str, old_re); | |
2026 | } else | |
9ef0c6ba | 2027 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_QUEUED); |
e5ac2adf MS |
2028 | } |
2029 | ||
12e7fe3a | 2030 | switch (op) { |
12e7fe3a DS |
2031 | case DPLANE_OP_ROUTE_INSTALL: |
2032 | case DPLANE_OP_ROUTE_UPDATE: | |
2033 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
677c1dd5 DS |
2034 | if (re) { |
2035 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
2036 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
2037 | } | |
b9f0e5ee DS |
2038 | /* |
2039 | * On an update operation from the same route type | |
2040 | * context retrieval currently has no way to know | |
2041 | * which was the old and which was the new. | |
2042 | * So don't unset our flags that we just set. | |
2043 | * We know redistribution is ok because the | |
2044 | * old_re in this case is used for nothing | |
2045 | * more than knowing whom to contact if necessary. | |
2046 | */ | |
2047 | if (old_re && old_re != re) { | |
677c1dd5 DS |
2048 | UNSET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
2049 | UNSET_FLAG(old_re->status, | |
2050 | ROUTE_ENTRY_INSTALLED); | |
2051 | } | |
12e7fe3a | 2052 | |
78bf56b0 MS |
2053 | /* Update zebra route based on the results in |
2054 | * the context struct. This also triggers | |
2055 | * redistribution for the route. | |
2056 | */ | |
2057 | if (re) | |
2058 | rib_update_re_from_ctx(re, rn, ctx); | |
f183e380 | 2059 | |
3cdba47a DS |
2060 | /* |
2061 | * System routes are weird in that they | |
2062 | * allow multiple to be installed that match | |
2063 | * to the same prefix, so after we get the | |
2064 | * result we need to clean them up so that | |
2065 | * we can actually use them. | |
2066 | */ | |
2067 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
2068 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
2069 | zebra_rib_fixup_system(rn); | |
2070 | ||
f4c6e2a8 | 2071 | if (zvrf) |
12e7fe3a | 2072 | zvrf->installs++; |
f183e380 | 2073 | |
12e7fe3a | 2074 | /* Notify route owner */ |
677c1dd5 | 2075 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_INSTALLED); |
e5ac2adf | 2076 | |
12e7fe3a | 2077 | } else { |
7a230a9d | 2078 | if (re) { |
677c1dd5 | 2079 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); |
7a230a9d DS |
2080 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
2081 | } if (old_re) | |
677c1dd5 DS |
2082 | SET_FLAG(old_re->status, ROUTE_ENTRY_FAILED); |
2083 | if (re) | |
2084 | zsend_route_notify_owner(re, dest_pfx, | |
2085 | ZAPI_ROUTE_FAIL_INSTALL); | |
97f5b441 | 2086 | |
12e7fe3a DS |
2087 | zlog_warn("%u:%s: Route install failed", |
2088 | dplane_ctx_get_vrf(ctx), | |
2089 | prefix2str(dest_pfx, | |
2090 | dest_str, sizeof(dest_str))); | |
2091 | } | |
2092 | break; | |
2093 | case DPLANE_OP_ROUTE_DELETE: | |
677c1dd5 DS |
2094 | if (re) |
2095 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
2096 | /* |
2097 | * In the delete case, the zebra core datastructs were | |
2098 | * updated (or removed) at the time the delete was issued, | |
2099 | * so we're just notifying the route owner. | |
e5ac2adf | 2100 | */ |
12e7fe3a | 2101 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { |
677c1dd5 DS |
2102 | if (re) { |
2103 | UNSET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); | |
2104 | UNSET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
2105 | } | |
12e7fe3a | 2106 | zsend_route_notify_owner_ctx(ctx, ZAPI_ROUTE_REMOVED); |
e5ac2adf | 2107 | |
12e7fe3a DS |
2108 | if (zvrf) |
2109 | zvrf->removals++; | |
2110 | } else { | |
677c1dd5 DS |
2111 | if (re) |
2112 | SET_FLAG(re->status, ROUTE_ENTRY_FAILED); | |
12e7fe3a DS |
2113 | zsend_route_notify_owner_ctx(ctx, |
2114 | ZAPI_ROUTE_REMOVE_FAIL); | |
e5ac2adf | 2115 | |
12e7fe3a DS |
2116 | zlog_warn("%u:%s: Route Deletion failure", |
2117 | dplane_ctx_get_vrf(ctx), | |
2118 | prefix2str(dest_pfx, | |
2119 | dest_str, sizeof(dest_str))); | |
2120 | } | |
3cdba47a DS |
2121 | |
2122 | /* | |
2123 | * System routes are weird in that they | |
2124 | * allow multiple to be installed that match | |
2125 | * to the same prefix, so after we get the | |
2126 | * result we need to clean them up so that | |
2127 | * we can actually use them. | |
2128 | */ | |
2129 | if ((re && RIB_SYSTEM_ROUTE(re)) || | |
2130 | (old_re && RIB_SYSTEM_ROUTE(old_re))) | |
2131 | zebra_rib_fixup_system(rn); | |
12e7fe3a | 2132 | break; |
d37f4d6c MS |
2133 | default: |
2134 | break; | |
e5ac2adf | 2135 | } |
699dae23 DS |
2136 | |
2137 | zebra_rib_evaluate_rn_nexthops(rn, seq); | |
a1494c25 | 2138 | zebra_rib_evaluate_mpls(rn); |
e5ac2adf MS |
2139 | done: |
2140 | ||
5f27bcba DS |
2141 | if (rn) |
2142 | route_unlock_node(rn); | |
2143 | ||
e5ac2adf MS |
2144 | /* Return context to dataplane module */ |
2145 | dplane_ctx_fini(&ctx); | |
2146 | } | |
2147 | ||
54818e3b MS |
2148 | /* |
2149 | * Handle notification from async dataplane: the dataplane has detected | |
2150 | * some change to a route, and notifies zebra so that the control plane | |
2151 | * can reflect that change. | |
2152 | */ | |
2153 | static void rib_process_dplane_notify(struct zebra_dplane_ctx *ctx) | |
2154 | { | |
2155 | struct route_node *rn = NULL; | |
2156 | struct route_entry *re = NULL, *rib; | |
2157 | char dest_str[PREFIX_STRLEN] = ""; | |
2158 | const struct prefix *dest_pfx, *src_pfx; | |
2159 | bool changed_p = false; | |
2160 | ||
2161 | dest_pfx = dplane_ctx_get_dest(ctx); | |
2162 | ||
2163 | /* Note well: only capturing the prefix string if debug is enabled here; | |
2164 | * unconditional log messages will have to generate the string. | |
2165 | */ | |
2166 | if (IS_ZEBRA_DEBUG_DPLANE) | |
2167 | prefix2str(dest_pfx, dest_str, sizeof(dest_str)); | |
2168 | ||
2169 | /* Locate rn and re(s) from ctx */ | |
2170 | rn = rib_find_rn_from_ctx(ctx); | |
2171 | if (rn == NULL) { | |
2172 | if (IS_ZEBRA_DEBUG_DPLANE) { | |
2173 | zlog_debug("Failed to process dplane notification: no route for %u:%s", | |
2174 | dplane_ctx_get_vrf(ctx), dest_str); | |
2175 | } | |
2176 | goto done; | |
2177 | } | |
2178 | ||
2179 | route_unlock_node(rn); | |
2180 | ||
2181 | srcdest_rnode_prefixes(rn, &dest_pfx, &src_pfx); | |
2182 | ||
2183 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
2184 | zlog_debug("%u:%s Processing dplane notif ctx %p", | |
2185 | dplane_ctx_get_vrf(ctx), dest_str, ctx); | |
2186 | ||
2187 | /* | |
2188 | * Take a pass through the routes, look for matches with the context | |
2189 | * info. | |
2190 | */ | |
2191 | RNODE_FOREACH_RE(rn, rib) { | |
2192 | ||
2193 | if (re == NULL) { | |
2194 | if (rib_route_match_ctx(rib, ctx, false)) | |
2195 | re = rib; | |
2196 | } | |
2197 | ||
2198 | /* Have we found the route we need to work on? */ | |
2199 | if (re) | |
2200 | break; | |
2201 | } | |
2202 | ||
2203 | /* No match? Nothing we can do */ | |
2204 | if (re == NULL) { | |
2205 | if (IS_ZEBRA_DEBUG_DPLANE) | |
2206 | zlog_debug("Unable to process dplane notification: no entry for %u:%s, type %d", | |
2207 | dplane_ctx_get_vrf(ctx), dest_str, | |
2208 | dplane_ctx_get_type(ctx)); | |
2209 | ||
2210 | goto done; | |
2211 | } | |
2212 | ||
2213 | /* Update zebra's nexthop FIB flags based on the context struct's | |
2214 | * nexthops. | |
2215 | */ | |
2216 | rib_update_re_from_ctx(re, rn, ctx); | |
2217 | ||
2218 | /* TODO -- we'd like to know about this possibility? */ | |
2219 | if (!changed_p) { | |
2220 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL) | |
2221 | zlog_debug("%u:%s No change from dplane notification", | |
2222 | dplane_ctx_get_vrf(ctx), dest_str); | |
2223 | } | |
2224 | ||
2225 | done: | |
2226 | /* Return context to dataplane module */ | |
2227 | dplane_ctx_fini(&ctx); | |
2228 | } | |
2229 | ||
5110a0c6 | 2230 | /* Take a list of route_node structs and return 1, if there was a record |
d62a17ae | 2231 | * picked from it and processed by rib_process(). Don't process more, |
5110a0c6 | 2232 | * than one RN record; operate only in the specified sub-queue. |
e96f9203 | 2233 | */ |
d7c0a89a | 2234 | static unsigned int process_subq(struct list *subq, uint8_t qindex) |
e96f9203 | 2235 | { |
d62a17ae | 2236 | struct listnode *lnode = listhead(subq); |
2237 | struct route_node *rnode; | |
2238 | rib_dest_t *dest; | |
2239 | struct zebra_vrf *zvrf = NULL; | |
5110a0c6 | 2240 | |
d62a17ae | 2241 | if (!lnode) |
2242 | return 0; | |
5110a0c6 | 2243 | |
d62a17ae | 2244 | rnode = listgetdata(lnode); |
2245 | dest = rib_dest_from_rnode(rnode); | |
2246 | if (dest) | |
2247 | zvrf = rib_dest_vrf(dest); | |
41ec9222 | 2248 | |
d62a17ae | 2249 | rib_process(rnode); |
5110a0c6 | 2250 | |
d62a17ae | 2251 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { |
2252 | char buf[SRCDEST2STR_BUFFER]; | |
2253 | srcdest_rnode2str(rnode, buf, sizeof(buf)); | |
2254 | zlog_debug("%u:%s: rn %p dequeued from sub-queue %u", | |
2255 | zvrf ? zvrf_id(zvrf) : 0, buf, rnode, qindex); | |
2256 | } | |
41ec9222 | 2257 | |
d62a17ae | 2258 | if (rnode->info) |
2259 | UNSET_FLAG(rib_dest_from_rnode(rnode)->flags, | |
2260 | RIB_ROUTE_QUEUED(qindex)); | |
9fd92e3c | 2261 | |
67b9467f | 2262 | #if 0 |
5110a0c6 SH |
2263 | else |
2264 | { | |
2265 | zlog_debug ("%s: called for route_node (%p, %d) with no ribs", | |
2266 | __func__, rnode, rnode->lock); | |
2267 | zlog_backtrace(LOG_DEBUG); | |
2268 | } | |
67b9467f | 2269 | #endif |
d62a17ae | 2270 | route_unlock_node(rnode); |
2271 | list_delete_node(subq, lnode); | |
2272 | return 1; | |
e96f9203 DO |
2273 | } |
2274 | ||
a1494c25 | 2275 | |
fb018d25 | 2276 | /* |
f183e380 | 2277 | * Perform next-hop tracking processing after RIB updates. |
fb018d25 | 2278 | */ |
f183e380 | 2279 | static void do_nht_processing(void) |
fb018d25 | 2280 | { |
fb018d25 DS |
2281 | } |
2282 | ||
e96f9203 | 2283 | /* Dispatch the meta queue by picking, processing and unlocking the next RN from |
d62a17ae | 2284 | * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and |
2285 | * data | |
e96f9203 DO |
2286 | * is pointed to the meta queue structure. |
2287 | */ | |
d62a17ae | 2288 | static wq_item_status meta_queue_process(struct work_queue *dummy, void *data) |
e96f9203 | 2289 | { |
d62a17ae | 2290 | struct meta_queue *mq = data; |
2291 | unsigned i; | |
91f16812 MS |
2292 | uint32_t queue_len, queue_limit; |
2293 | ||
2294 | /* Ensure there's room for more dataplane updates */ | |
2295 | queue_limit = dplane_get_in_queue_limit(); | |
2296 | queue_len = dplane_get_in_queue_len(); | |
2297 | if (queue_len > queue_limit) { | |
2298 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
8b962e77 MS |
2299 | zlog_debug("rib queue: dplane queue len %u, limit %u, retrying", |
2300 | queue_len, queue_limit); | |
91f16812 MS |
2301 | |
2302 | /* Ensure that the meta-queue is actually enqueued */ | |
489a9614 | 2303 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2304 | work_queue_add(zrouter.ribq, zrouter.mq); |
91f16812 MS |
2305 | |
2306 | return WQ_QUEUE_BLOCKED; | |
2307 | } | |
5110a0c6 | 2308 | |
d62a17ae | 2309 | for (i = 0; i < MQ_SIZE; i++) |
2310 | if (process_subq(mq->subq[i], i)) { | |
2311 | mq->size--; | |
2312 | break; | |
2313 | } | |
2314 | return mq->size ? WQ_REQUEUE : WQ_SUCCESS; | |
e96f9203 DO |
2315 | } |
2316 | ||
f52ed677 DS |
2317 | |
2318 | /* | |
2319 | * Look into the RN and queue it into the highest priority queue | |
2320 | * at this point in time for processing. | |
2321 | * | |
2322 | * We will enqueue a route node only once per invocation. | |
2323 | * | |
2324 | * There are two possibilities here that should be kept in mind. | |
2325 | * If the original invocation has not been pulled off for processing | |
2326 | * yet, A subsuquent invocation can have a route entry with a better | |
2327 | * meta queue index value and we can have a situation where | |
2328 | * we might have the same node enqueued 2 times. Not necessarily | |
2329 | * an optimal situation but it should be ok. | |
2330 | * | |
2331 | * The other possibility is that the original invocation has not | |
2332 | * been pulled off for processing yet, A subsusquent invocation | |
2333 | * doesn't have a route_entry with a better meta-queue and the | |
2334 | * original metaqueue index value will win and we'll end up with | |
2335 | * the route node enqueued once. | |
e96f9203 | 2336 | */ |
d62a17ae | 2337 | static void rib_meta_queue_add(struct meta_queue *mq, struct route_node *rn) |
e96f9203 | 2338 | { |
f52ed677 DS |
2339 | struct route_entry *re = NULL, *curr_re = NULL; |
2340 | uint8_t qindex = MQ_SIZE, curr_qindex = MQ_SIZE; | |
5110a0c6 | 2341 | |
f52ed677 DS |
2342 | RNODE_FOREACH_RE (rn, curr_re) { |
2343 | curr_qindex = route_info[curr_re->type].meta_q_map; | |
d62a17ae | 2344 | |
f52ed677 DS |
2345 | if (curr_qindex <= qindex) { |
2346 | re = curr_re; | |
2347 | qindex = curr_qindex; | |
d62a17ae | 2348 | } |
f52ed677 | 2349 | } |
5110a0c6 | 2350 | |
f52ed677 DS |
2351 | if (!re) |
2352 | return; | |
5110a0c6 | 2353 | |
f52ed677 DS |
2354 | /* Invariant: at this point we always have rn->info set. */ |
2355 | if (CHECK_FLAG(rib_dest_from_rnode(rn)->flags, | |
2356 | RIB_ROUTE_QUEUED(qindex))) { | |
d62a17ae | 2357 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2358 | rnode_debug(rn, re->vrf_id, | |
f52ed677 | 2359 | "rn %p is already queued in sub-queue %u", |
d62a17ae | 2360 | (void *)rn, qindex); |
f52ed677 | 2361 | return; |
d62a17ae | 2362 | } |
f52ed677 DS |
2363 | |
2364 | SET_FLAG(rib_dest_from_rnode(rn)->flags, RIB_ROUTE_QUEUED(qindex)); | |
2365 | listnode_add(mq->subq[qindex], rn); | |
2366 | route_lock_node(rn); | |
2367 | mq->size++; | |
2368 | ||
2369 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2370 | rnode_debug(rn, re->vrf_id, "queued rn %p into sub-queue %u", | |
2371 | (void *)rn, qindex); | |
4d38fdb4 | 2372 | } |
2373 | ||
6d691129 | 2374 | /* Add route_node to work queue and schedule processing */ |
d62a17ae | 2375 | void rib_queue_add(struct route_node *rn) |
4d38fdb4 | 2376 | { |
d62a17ae | 2377 | assert(rn); |
fc328ac9 | 2378 | |
d62a17ae | 2379 | /* Pointless to queue a route_node with no RIB entries to add or remove |
2380 | */ | |
2381 | if (!rnode_to_ribs(rn)) { | |
2382 | zlog_debug("%s: called for route_node (%p, %d) with no ribs", | |
2383 | __func__, (void *)rn, rn->lock); | |
2384 | zlog_backtrace(LOG_DEBUG); | |
2385 | return; | |
2386 | } | |
4d38fdb4 | 2387 | |
489a9614 | 2388 | if (zrouter.ribq == NULL) { |
e914ccbe | 2389 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2390 | "%s: work_queue does not exist!", __func__); |
d62a17ae | 2391 | return; |
2392 | } | |
2393 | ||
2394 | /* | |
2395 | * The RIB queue should normally be either empty or holding the only | |
2396 | * work_queue_item element. In the latter case this element would | |
2397 | * hold a pointer to the meta queue structure, which must be used to | |
2398 | * actually queue the route nodes to process. So create the MQ | |
2399 | * holder, if necessary, then push the work into it in any case. | |
2400 | * This semantics was introduced after 0.99.9 release. | |
2401 | */ | |
489a9614 | 2402 | if (work_queue_empty(zrouter.ribq)) |
ea45a4e7 | 2403 | work_queue_add(zrouter.ribq, zrouter.mq); |
e96f9203 | 2404 | |
ea45a4e7 | 2405 | rib_meta_queue_add(zrouter.mq, rn); |
fc328ac9 | 2406 | |
d62a17ae | 2407 | return; |
4d38fdb4 | 2408 | } |
2409 | ||
5110a0c6 SH |
2410 | /* Create new meta queue. |
2411 | A destructor function doesn't seem to be necessary here. | |
2412 | */ | |
d62a17ae | 2413 | static struct meta_queue *meta_queue_new(void) |
e96f9203 | 2414 | { |
d62a17ae | 2415 | struct meta_queue *new; |
2416 | unsigned i; | |
5110a0c6 | 2417 | |
d62a17ae | 2418 | new = XCALLOC(MTYPE_WORK_QUEUE, sizeof(struct meta_queue)); |
e96f9203 | 2419 | |
d62a17ae | 2420 | for (i = 0; i < MQ_SIZE; i++) { |
2421 | new->subq[i] = list_new(); | |
2422 | assert(new->subq[i]); | |
2423 | } | |
5110a0c6 | 2424 | |
d62a17ae | 2425 | return new; |
e96f9203 DO |
2426 | } |
2427 | ||
d62a17ae | 2428 | void meta_queue_free(struct meta_queue *mq) |
5a8dfcd8 | 2429 | { |
d62a17ae | 2430 | unsigned i; |
5a8dfcd8 | 2431 | |
d62a17ae | 2432 | for (i = 0; i < MQ_SIZE; i++) |
6a154c88 | 2433 | list_delete(&mq->subq[i]); |
5a8dfcd8 | 2434 | |
d62a17ae | 2435 | XFREE(MTYPE_WORK_QUEUE, mq); |
5a8dfcd8 RW |
2436 | } |
2437 | ||
4d38fdb4 | 2438 | /* initialise zebra rib work queue */ |
2561d12e | 2439 | static void rib_queue_init(void) |
4d38fdb4 | 2440 | { |
489a9614 DS |
2441 | if (!(zrouter.ribq = work_queue_new(zrouter.master, |
2442 | "route_node processing"))) { | |
e914ccbe | 2443 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2444 | "%s: could not initialise work queue!", __func__); |
d62a17ae | 2445 | return; |
2446 | } | |
4d38fdb4 | 2447 | |
d62a17ae | 2448 | /* fill in the work queue spec */ |
489a9614 DS |
2449 | zrouter.ribq->spec.workfunc = &meta_queue_process; |
2450 | zrouter.ribq->spec.errorfunc = NULL; | |
46a4e345 | 2451 | zrouter.ribq->spec.completion_func = NULL; |
d62a17ae | 2452 | /* XXX: TODO: These should be runtime configurable via vty */ |
489a9614 DS |
2453 | zrouter.ribq->spec.max_retries = 3; |
2454 | zrouter.ribq->spec.hold = ZEBRA_RIB_PROCESS_HOLD_TIME; | |
2455 | zrouter.ribq->spec.retry = ZEBRA_RIB_PROCESS_RETRY_TIME; | |
d62a17ae | 2456 | |
ea45a4e7 | 2457 | if (!(zrouter.mq = meta_queue_new())) { |
e914ccbe | 2458 | flog_err(EC_ZEBRA_WQ_NONEXISTENT, |
1c50c1c0 | 2459 | "%s: could not initialise meta queue!", __func__); |
d62a17ae | 2460 | return; |
2461 | } | |
2462 | return; | |
718e3744 | 2463 | } |
2464 | ||
434434f7 DS |
2465 | rib_dest_t *zebra_rib_create_dest(struct route_node *rn) |
2466 | { | |
2467 | rib_dest_t *dest; | |
2468 | ||
2469 | dest = XCALLOC(MTYPE_RIB_DEST, sizeof(rib_dest_t)); | |
aa57abfb | 2470 | rnh_list_init(&dest->nht); |
434434f7 DS |
2471 | route_lock_node(rn); /* rn route table reference */ |
2472 | rn->info = dest; | |
2473 | dest->rnode = rn; | |
2474 | ||
2475 | return dest; | |
2476 | } | |
2477 | ||
6d691129 PJ |
2478 | /* RIB updates are processed via a queue of pointers to route_nodes. |
2479 | * | |
2480 | * The queue length is bounded by the maximal size of the routing table, | |
2481 | * as a route_node will not be requeued, if already queued. | |
2482 | * | |
f0f77c9a DS |
2483 | * REs are submitted via rib_addnode or rib_delnode which set minimal |
2484 | * state, or static_install_route (when an existing RE is updated) | |
3c0755dc | 2485 | * and then submit route_node to queue for best-path selection later. |
f0f77c9a | 2486 | * Order of add/delete state changes are preserved for any given RE. |
6d691129 | 2487 | * |
f0f77c9a | 2488 | * Deleted REs are reaped during best-path selection. |
6d691129 PJ |
2489 | * |
2490 | * rib_addnode | |
f0f77c9a DS |
2491 | * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with |
2492 | * |-------->| | best RE, if required | |
3c0755dc PJ |
2493 | * | | |
2494 | * static_install->|->rib_addqueue...... -> rib_process | |
2495 | * | | | |
2496 | * |-------->| |-> rib_unlink | |
f0f77c9a DS |
2497 | * |-> set ROUTE_ENTRY_REMOVE | |
2498 | * rib_delnode (RE freed) | |
6d691129 | 2499 | * |
9fd92e3c AS |
2500 | * The 'info' pointer of a route_node points to a rib_dest_t |
2501 | * ('dest'). Queueing state for a route_node is kept on the dest. The | |
2502 | * dest is created on-demand by rib_link() and is kept around at least | |
2503 | * as long as there are ribs hanging off it (@see rib_gc_dest()). | |
d62a17ae | 2504 | * |
6d691129 PJ |
2505 | * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code): |
2506 | * | |
2507 | * - route_nodes: refcounted by: | |
9fd92e3c AS |
2508 | * - dest attached to route_node: |
2509 | * - managed by: rib_link/rib_gc_dest | |
6d691129 PJ |
2510 | * - route_node processing queue |
2511 | * - managed by: rib_addqueue, rib_process. | |
2512 | * | |
2513 | */ | |
d62a17ae | 2514 | |
f0f77c9a | 2515 | /* Add RE to head of the route node. */ |
d62a17ae | 2516 | static void rib_link(struct route_node *rn, struct route_entry *re, int process) |
2517 | { | |
d62a17ae | 2518 | rib_dest_t *dest; |
2519 | afi_t afi; | |
2520 | const char *rmap_name; | |
9fd92e3c | 2521 | |
d62a17ae | 2522 | assert(re && rn); |
9fd92e3c | 2523 | |
d62a17ae | 2524 | dest = rib_dest_from_rnode(rn); |
2525 | if (!dest) { | |
2526 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
2527 | rnode_debug(rn, re->vrf_id, "rn %p adding dest", rn); | |
7a4bb9c5 | 2528 | |
434434f7 | 2529 | dest = zebra_rib_create_dest(rn); |
d62a17ae | 2530 | } |
2263a412 | 2531 | |
045207e2 | 2532 | re_list_add_head(&dest->routes, re); |
d62a17ae | 2533 | |
2534 | afi = (rn->p.family == AF_INET) | |
2535 | ? AFI_IP | |
2536 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2537 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2538 | rmap_name = zebra_get_import_table_route_map(afi, re->table); | |
2539 | zebra_add_import_table_entry(rn, re, rmap_name); | |
2540 | } else if (process) | |
2541 | rib_queue_add(rn); | |
2542 | } | |
2543 | ||
7e24fdf3 DS |
2544 | static void rib_addnode(struct route_node *rn, |
2545 | struct route_entry *re, int process) | |
d62a17ae | 2546 | { |
2547 | /* RE node has been un-removed before route-node is processed. | |
2548 | * route_node must hence already be on the queue for processing.. | |
2549 | */ | |
2550 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) { | |
2551 | if (IS_ZEBRA_DEBUG_RIB) | |
2552 | rnode_debug(rn, re->vrf_id, "rn %p, un-removed re %p", | |
2553 | (void *)rn, (void *)re); | |
2554 | ||
2555 | UNSET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2556 | return; | |
2557 | } | |
2558 | rib_link(rn, re, process); | |
6d691129 PJ |
2559 | } |
2560 | ||
9fd92e3c AS |
2561 | /* |
2562 | * rib_unlink | |
2563 | * | |
2564 | * Detach a rib structure from a route_node. | |
2565 | * | |
2566 | * Note that a call to rib_unlink() should be followed by a call to | |
2567 | * rib_gc_dest() at some point. This allows a rib_dest_t that is no | |
2568 | * longer required to be deleted. | |
2569 | */ | |
d62a17ae | 2570 | void rib_unlink(struct route_node *rn, struct route_entry *re) |
6d691129 | 2571 | { |
d62a17ae | 2572 | rib_dest_t *dest; |
9fd92e3c | 2573 | |
d62a17ae | 2574 | assert(rn && re); |
6d691129 | 2575 | |
d62a17ae | 2576 | if (IS_ZEBRA_DEBUG_RIB) |
2577 | rnode_debug(rn, re->vrf_id, "rn %p, re %p", (void *)rn, | |
2578 | (void *)re); | |
6d691129 | 2579 | |
d62a17ae | 2580 | dest = rib_dest_from_rnode(rn); |
6d691129 | 2581 | |
045207e2 | 2582 | re_list_del(&dest->routes, re); |
7a4bb9c5 | 2583 | |
2eb07de3 DS |
2584 | if (dest->selected_fib == re) |
2585 | dest->selected_fib = NULL; | |
2586 | ||
7ee30f28 | 2587 | nexthops_free(re->ng.nexthop); |
d62a17ae | 2588 | XFREE(MTYPE_RE, re); |
2589 | } | |
2590 | ||
2591 | void rib_delnode(struct route_node *rn, struct route_entry *re) | |
2592 | { | |
2593 | afi_t afi; | |
2594 | ||
2595 | if (IS_ZEBRA_DEBUG_RIB) | |
2596 | rnode_debug(rn, re->vrf_id, "rn %p, re %p, removing", | |
2597 | (void *)rn, (void *)re); | |
2598 | SET_FLAG(re->status, ROUTE_ENTRY_REMOVED); | |
2599 | ||
2600 | afi = (rn->p.family == AF_INET) | |
2601 | ? AFI_IP | |
2602 | : (rn->p.family == AF_INET6) ? AFI_IP6 : AFI_MAX; | |
2603 | if (is_zebra_import_table_enabled(afi, re->table)) { | |
2604 | zebra_del_import_table_entry(rn, re); | |
2605 | /* Just clean up if non main table */ | |
2606 | if (IS_ZEBRA_DEBUG_RIB) { | |
2607 | char buf[SRCDEST2STR_BUFFER]; | |
2608 | srcdest_rnode2str(rn, buf, sizeof(buf)); | |
2da33d6b DS |
2609 | zlog_debug("%u:%s: Freeing route rn %p, re %p (%s)", |
2610 | re->vrf_id, buf, rn, re, | |
2611 | zebra_route_string(re->type)); | |
d62a17ae | 2612 | } |
7a4bb9c5 | 2613 | |
d62a17ae | 2614 | rib_unlink(rn, re); |
2615 | } else { | |
2616 | rib_queue_add(rn); | |
2617 | } | |
718e3744 | 2618 | } |
2619 | ||
f0f77c9a | 2620 | /* This function dumps the contents of a given RE entry into |
dc95824a DO |
2621 | * standard debug log. Calling function name and IP prefix in |
2622 | * question are passed as 1st and 2nd arguments. | |
2623 | */ | |
2624 | ||
d62a17ae | 2625 | void _route_entry_dump(const char *func, union prefixconstptr pp, |
2626 | union prefixconstptr src_pp, | |
2627 | const struct route_entry *re) | |
2628 | { | |
d62a17ae | 2629 | const struct prefix *src_p = src_pp.p; |
2630 | bool is_srcdst = src_p && src_p->prefixlen; | |
2631 | char straddr[PREFIX_STRLEN]; | |
2632 | char srcaddr[PREFIX_STRLEN]; | |
2633 | struct nexthop *nexthop; | |
2634 | ||
2635 | zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func, | |
2636 | (const void *)re, prefix2str(pp, straddr, sizeof(straddr)), | |
2637 | is_srcdst ? " from " : "", | |
2638 | is_srcdst ? prefix2str(src_pp, srcaddr, sizeof(srcaddr)) | |
2639 | : "", | |
2640 | re->vrf_id); | |
cc54cfee RW |
2641 | zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d", |
2642 | func, (unsigned long)re->uptime, re->type, re->instance, | |
2643 | re->table); | |
d62a17ae | 2644 | zlog_debug( |
2645 | "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u", | |
2646 | func, re->metric, re->mtu, re->distance, re->flags, re->status); | |
2647 | zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func, | |
2648 | re->nexthop_num, re->nexthop_active_num); | |
2649 | ||
7ee30f28 | 2650 | for (ALL_NEXTHOPS(re->ng, nexthop)) { |
2d68a0f2 DS |
2651 | struct interface *ifp; |
2652 | struct vrf *vrf = vrf_lookup_by_id(nexthop->vrf_id); | |
2653 | ||
2654 | switch (nexthop->type) { | |
2655 | case NEXTHOP_TYPE_BLACKHOLE: | |
2656 | sprintf(straddr, "Blackhole"); | |
2657 | break; | |
2658 | case NEXTHOP_TYPE_IFINDEX: | |
2659 | ifp = if_lookup_by_index(nexthop->ifindex, | |
2660 | nexthop->vrf_id); | |
2661 | sprintf(straddr, "%s", ifp ? ifp->name : "Unknown"); | |
2662 | break; | |
2663 | case NEXTHOP_TYPE_IPV4: | |
2664 | /* fallthrough */ | |
2665 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
2666 | inet_ntop(AF_INET, &nexthop->gate, straddr, | |
2667 | INET6_ADDRSTRLEN); | |
2668 | break; | |
2669 | case NEXTHOP_TYPE_IPV6: | |
2670 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
2671 | inet_ntop(AF_INET6, &nexthop->gate, straddr, | |
2672 | INET6_ADDRSTRLEN); | |
2673 | break; | |
2674 | } | |
df38b099 DS |
2675 | zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s%s%s%s", |
2676 | func, (nexthop->rparent ? " NH" : "NH"), straddr, | |
2d68a0f2 DS |
2677 | nexthop->ifindex, vrf ? vrf->name : "Unknown", |
2678 | nexthop->vrf_id, | |
d62a17ae | 2679 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE) |
2680 | ? "ACTIVE " | |
2681 | : ""), | |
677c1dd5 | 2682 | (CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED) |
d62a17ae | 2683 | ? "FIB " |
2684 | : ""), | |
2685 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE) | |
df38b099 DS |
2686 | ? "RECURSIVE " |
2687 | : ""), | |
2688 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK) | |
2689 | ? "ONLINK " | |
2690 | : ""), | |
2691 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_MATCHED) | |
2692 | ? "MATCHED " | |
2693 | : ""), | |
2694 | (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_DUPLICATE) | |
2695 | ? "DUPLICATE " | |
d62a17ae | 2696 | : "")); |
2697 | } | |
2698 | zlog_debug("%s: dump complete", func); | |
dc95824a DO |
2699 | } |
2700 | ||
2701 | /* This is an exported helper to rtm_read() to dump the strange | |
f0f77c9a | 2702 | * RE entry found by rib_lookup_ipv4_route() |
dc95824a DO |
2703 | */ |
2704 | ||
d62a17ae | 2705 | void rib_lookup_and_dump(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2706 | { | |
2707 | struct route_table *table; | |
2708 | struct route_node *rn; | |
2709 | struct route_entry *re; | |
2710 | char prefix_buf[INET_ADDRSTRLEN]; | |
2711 | ||
2712 | /* Lookup table. */ | |
2713 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
2714 | if (!table) { | |
e914ccbe | 2715 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2716 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2717 | vrf_id); | |
d62a17ae | 2718 | return; |
2719 | } | |
2720 | ||
2721 | /* Scan the RIB table for exactly matching RE entry. */ | |
2722 | rn = route_node_lookup(table, (struct prefix *)p); | |
2723 | ||
2724 | /* No route for this prefix. */ | |
2725 | if (!rn) { | |
32391aff | 2726 | zlog_debug("%s:%u lookup failed for %s", __func__, vrf_id, |
d62a17ae | 2727 | prefix2str((struct prefix *)p, prefix_buf, |
2728 | sizeof(prefix_buf))); | |
2729 | return; | |
2730 | } | |
2731 | ||
2732 | /* Unlock node. */ | |
2733 | route_unlock_node(rn); | |
2734 | ||
2735 | /* let's go */ | |
a2addae8 | 2736 | RNODE_FOREACH_RE (rn, re) { |
32391aff DS |
2737 | zlog_debug("%s:%u rn %p, re %p: %s, %s", |
2738 | __func__, vrf_id, | |
2739 | (void *)rn, (void *)re, | |
d62a17ae | 2740 | (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED) |
2741 | ? "removed" | |
2742 | : "NOT removed"), | |
2743 | (CHECK_FLAG(re->flags, ZEBRA_FLAG_SELECTED) | |
2744 | ? "selected" | |
2745 | : "NOT selected")); | |
2746 | route_entry_dump(p, NULL, re); | |
2747 | } | |
dc95824a DO |
2748 | } |
2749 | ||
20e5ff0a DO |
2750 | /* Check if requested address assignment will fail due to another |
2751 | * route being installed by zebra in FIB already. Take necessary | |
2752 | * actions, if needed: remove such a route from FIB and deSELECT | |
f0f77c9a | 2753 | * corresponding RE entry. Then put affected RN into RIBQ head. |
20e5ff0a | 2754 | */ |
d62a17ae | 2755 | void rib_lookup_and_pushup(struct prefix_ipv4 *p, vrf_id_t vrf_id) |
2756 | { | |
2757 | struct route_table *table; | |
2758 | struct route_node *rn; | |
5f7a4718 | 2759 | rib_dest_t *dest; |
d62a17ae | 2760 | |
2761 | if (NULL == (table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id))) { | |
e914ccbe | 2762 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, |
1c50c1c0 QY |
2763 | "%s:%u zebra_vrf_table() returned NULL", __func__, |
2764 | vrf_id); | |
d62a17ae | 2765 | return; |
2766 | } | |
2767 | ||
2768 | /* No matches would be the simplest case. */ | |
2769 | if (NULL == (rn = route_node_lookup(table, (struct prefix *)p))) | |
2770 | return; | |
2771 | ||
2772 | /* Unlock node. */ | |
2773 | route_unlock_node(rn); | |
2774 | ||
5f7a4718 | 2775 | dest = rib_dest_from_rnode(rn); |
d62a17ae | 2776 | /* Check all RE entries. In case any changes have to be done, requeue |
2777 | * the RN into RIBQ head. If the routing message about the new connected | |
2778 | * route (generated by the IP address we are going to assign very soon) | |
2779 | * comes before the RIBQ is processed, the new RE entry will join | |
2780 | * RIBQ record already on head. This is necessary for proper | |
2781 | * revalidation | |
2782 | * of the rest of the RE. | |
2783 | */ | |
3cdba47a | 2784 | if (dest->selected_fib) { |
5f7a4718 DS |
2785 | if (IS_ZEBRA_DEBUG_RIB) { |
2786 | char buf[PREFIX_STRLEN]; | |
2787 | ||
2788 | zlog_debug("%u:%s: freeing way for connected prefix", | |
2789 | dest->selected_fib->vrf_id, | |
2790 | prefix2str(&rn->p, buf, sizeof(buf))); | |
2791 | route_entry_dump(&rn->p, NULL, dest->selected_fib); | |
d62a17ae | 2792 | } |
5f7a4718 | 2793 | rib_uninstall(rn, dest->selected_fib); |
d62a17ae | 2794 | rib_queue_add(rn); |
ade4a886 | 2795 | } |
20e5ff0a DO |
2796 | } |
2797 | ||
d62a17ae | 2798 | int rib_add_multipath(afi_t afi, safi_t safi, struct prefix *p, |
2799 | struct prefix_ipv6 *src_p, struct route_entry *re) | |
718e3744 | 2800 | { |
d62a17ae | 2801 | struct route_table *table; |
2802 | struct route_node *rn; | |
40ecd8e4 | 2803 | struct route_entry *same = NULL; |
d62a17ae | 2804 | int ret = 0; |
b4c034b0 | 2805 | |
d62a17ae | 2806 | if (!re) |
2807 | return 0; | |
b4c034b0 | 2808 | |
1f610a1f | 2809 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
05737783 | 2810 | |
d62a17ae | 2811 | /* Lookup table. */ |
7865c65d RW |
2812 | table = zebra_vrf_table_with_table_id(afi, safi, re->vrf_id, re->table); |
2813 | if (!table) { | |
2814 | XFREE(MTYPE_RE, re); | |
d62a17ae | 2815 | return 0; |
7865c65d | 2816 | } |
cddf391b | 2817 | |
d62a17ae | 2818 | /* Make it sure prefixlen is applied to the prefix. */ |
2819 | apply_mask(p); | |
2820 | if (src_p) | |
2821 | apply_mask_ipv6(src_p); | |
718e3744 | 2822 | |
d62a17ae | 2823 | /* Set default distance by route type. */ |
2824 | if (re->distance == 0) { | |
0492eea0 | 2825 | re->distance = route_distance(re->type); |
718e3744 | 2826 | |
d62a17ae | 2827 | /* iBGP distance is 200. */ |
2828 | if (re->type == ZEBRA_ROUTE_BGP | |
2829 | && CHECK_FLAG(re->flags, ZEBRA_FLAG_IBGP)) | |
2830 | re->distance = 200; | |
2831 | } | |
718e3744 | 2832 | |
d62a17ae | 2833 | /* Lookup route node.*/ |
2834 | rn = srcdest_rnode_get(table, p, src_p); | |
718e3744 | 2835 | |
40ecd8e4 DS |
2836 | /* |
2837 | * If same type of route are installed, treat it as a implicit | |
2838 | * withdraw. | |
2839 | * If the user has specified the No route replace semantics | |
2840 | * for the install don't do a route replace. | |
2841 | */ | |
a2addae8 | 2842 | RNODE_FOREACH_RE (rn, same) { |
d62a17ae | 2843 | if (CHECK_FLAG(same->status, ROUTE_ENTRY_REMOVED)) |
2844 | continue; | |
41ec9222 | 2845 | |
eb327fa5 RW |
2846 | if (same->type != re->type) |
2847 | continue; | |
2848 | if (same->instance != re->instance) | |
2849 | continue; | |
996c9314 LB |
2850 | if (same->type == ZEBRA_ROUTE_KERNEL |
2851 | && same->metric != re->metric) | |
eb327fa5 | 2852 | continue; |
40ecd8e4 DS |
2853 | |
2854 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && | |
2855 | same->distance != re->distance) | |
2856 | continue; | |
2857 | ||
844b3a87 | 2858 | /* |
40ecd8e4 DS |
2859 | * We should allow duplicate connected routes |
2860 | * because of IPv6 link-local routes and unnumbered | |
2861 | * interfaces on Linux. | |
844b3a87 RW |
2862 | */ |
2863 | if (same->type != ZEBRA_ROUTE_CONNECT) | |
d62a17ae | 2864 | break; |
2865 | } | |
718e3744 | 2866 | |
cf363e1b MS |
2867 | /* If this route is kernel/connected route, notify the dataplane. */ |
2868 | if (RIB_SYSTEM_ROUTE(re)) { | |
2869 | /* Notify dataplane */ | |
2870 | dplane_sys_route_add(rn, re); | |
2871 | } | |
2872 | ||
d62a17ae | 2873 | /* Link new re to node.*/ |
2874 | if (IS_ZEBRA_DEBUG_RIB) { | |
2da33d6b DS |
2875 | rnode_debug(rn, re->vrf_id, |
2876 | "Inserting route rn %p, re %p (%s) existing %p", | |
2877 | rn, re, zebra_route_string(re->type), same); | |
718e3744 | 2878 | |
d62a17ae | 2879 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2880 | route_entry_dump(p, src_p, re); | |
718e3744 | 2881 | } |
6883bf8d DS |
2882 | |
2883 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
d62a17ae | 2884 | rib_addnode(rn, re, 1); |
2885 | ret = 1; | |
6b0655a2 | 2886 | |
d62a17ae | 2887 | /* Free implicit route.*/ |
2888 | if (same) { | |
2889 | rib_delnode(rn, same); | |
2890 | ret = -1; | |
2891 | } | |
718e3744 | 2892 | |
d62a17ae | 2893 | route_unlock_node(rn); |
2894 | return ret; | |
2895 | } | |
2896 | ||
2897 | void rib_delete(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, | |
d7c0a89a | 2898 | unsigned short instance, int flags, struct prefix *p, |
fd36be7e | 2899 | struct prefix_ipv6 *src_p, const struct nexthop *nh, |
40ecd8e4 DS |
2900 | uint32_t table_id, uint32_t metric, uint8_t distance, |
2901 | bool fromkernel) | |
d62a17ae | 2902 | { |
2903 | struct route_table *table; | |
2904 | struct route_node *rn; | |
2905 | struct route_entry *re; | |
2906 | struct route_entry *fib = NULL; | |
2907 | struct route_entry *same = NULL; | |
fd36be7e | 2908 | struct nexthop *rtnh; |
d62a17ae | 2909 | char buf2[INET6_ADDRSTRLEN]; |
5f7a4718 | 2910 | rib_dest_t *dest; |
d62a17ae | 2911 | |
1f610a1f | 2912 | assert(!src_p || !src_p->prefixlen || afi == AFI_IP6); |
d62a17ae | 2913 | |
2914 | /* Lookup table. */ | |
2915 | table = zebra_vrf_table_with_table_id(afi, safi, vrf_id, table_id); | |
2916 | if (!table) | |
2917 | return; | |
2918 | ||
2919 | /* Apply mask. */ | |
2920 | apply_mask(p); | |
2921 | if (src_p) | |
2922 | apply_mask_ipv6(src_p); | |
2923 | ||
2924 | /* Lookup route node. */ | |
2925 | rn = srcdest_rnode_lookup(table, p, src_p); | |
2926 | if (!rn) { | |
2927 | char dst_buf[PREFIX_STRLEN], src_buf[PREFIX_STRLEN]; | |
2928 | ||
2929 | prefix2str(p, dst_buf, sizeof(dst_buf)); | |
2930 | if (src_p && src_p->prefixlen) | |
2931 | prefix2str(src_p, src_buf, sizeof(src_buf)); | |
2932 | else | |
2933 | src_buf[0] = '\0'; | |
2934 | ||
2935 | if (IS_ZEBRA_DEBUG_RIB) | |
2936 | zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id, | |
2937 | dst_buf, | |
2938 | (src_buf[0] != '\0') ? " from " : "", | |
2939 | src_buf); | |
2940 | return; | |
2941 | } | |
718e3744 | 2942 | |
5f7a4718 DS |
2943 | dest = rib_dest_from_rnode(rn); |
2944 | fib = dest->selected_fib; | |
2945 | ||
d62a17ae | 2946 | /* Lookup same type route. */ |
a2addae8 | 2947 | RNODE_FOREACH_RE (rn, re) { |
d62a17ae | 2948 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
2949 | continue; | |
2950 | ||
d62a17ae | 2951 | if (re->type != type) |
2952 | continue; | |
2953 | if (re->instance != instance) | |
2954 | continue; | |
40ecd8e4 DS |
2955 | if (CHECK_FLAG(re->flags, ZEBRA_FLAG_RR_USE_DISTANCE) && |
2956 | distance != re->distance) | |
2957 | continue; | |
2958 | ||
996c9314 | 2959 | if (re->type == ZEBRA_ROUTE_KERNEL && re->metric != metric) |
f19435a8 | 2960 | continue; |
7ee30f28 | 2961 | if (re->type == ZEBRA_ROUTE_CONNECT && (rtnh = re->ng.nexthop) |
fd36be7e DL |
2962 | && rtnh->type == NEXTHOP_TYPE_IFINDEX && nh) { |
2963 | if (rtnh->ifindex != nh->ifindex) | |
d62a17ae | 2964 | continue; |
d62a17ae | 2965 | same = re; |
2966 | break; | |
2967 | } | |
2968 | /* Make sure that the route found has the same gateway. */ | |
2969 | else { | |
fd36be7e | 2970 | if (nh == NULL) { |
d62a17ae | 2971 | same = re; |
2972 | break; | |
2973 | } | |
7ee30f28 | 2974 | for (ALL_NEXTHOPS(re->ng, rtnh)) |
fd36be7e | 2975 | if (nexthop_same_no_recurse(rtnh, nh)) { |
d62a17ae | 2976 | same = re; |
2977 | break; | |
2978 | } | |
2979 | if (same) | |
2980 | break; | |
2981 | } | |
2982 | } | |
2983 | /* If same type of route can't be found and this message is from | |
2984 | kernel. */ | |
2985 | if (!same) { | |
5dfeba19 DS |
2986 | /* |
2987 | * In the past(HA!) we could get here because | |
2988 | * we were receiving a route delete from the | |
2989 | * kernel and we're not marking the proto | |
2990 | * as coming from it's appropriate originator. | |
2991 | * Now that we are properly noticing the fact | |
2992 | * that the kernel has deleted our route we | |
2993 | * are not going to get called in this path | |
2994 | * I am going to leave this here because | |
2995 | * this might still work this way on non-linux | |
2996 | * platforms as well as some weird state I have | |
2997 | * not properly thought of yet. | |
2998 | * If we can show that this code path is | |
2999 | * dead then we can remove it. | |
3000 | */ | |
b8faa875 | 3001 | if (fib && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE)) { |
d62a17ae | 3002 | if (IS_ZEBRA_DEBUG_RIB) { |
2da33d6b DS |
3003 | rnode_debug(rn, vrf_id, |
3004 | "rn %p, re %p (%s) was deleted from kernel, adding", | |
3005 | rn, fib, | |
3006 | zebra_route_string(fib->type)); | |
d62a17ae | 3007 | } |
3008 | if (allow_delete) { | |
677c1dd5 | 3009 | UNSET_FLAG(fib->status, ROUTE_ENTRY_INSTALLED); |
d62a17ae | 3010 | /* Unset flags. */ |
7ee30f28 | 3011 | for (rtnh = fib->ng.nexthop; rtnh; |
fd36be7e DL |
3012 | rtnh = rtnh->next) |
3013 | UNSET_FLAG(rtnh->flags, | |
d62a17ae | 3014 | NEXTHOP_FLAG_FIB); |
3015 | ||
ed216282 DS |
3016 | /* |
3017 | * This is a non FRR route | |
3018 | * as such we should mark | |
3019 | * it as deleted | |
3020 | */ | |
5f7a4718 | 3021 | dest->selected_fib = NULL; |
d62a17ae | 3022 | } else { |
3023 | /* This means someone else, other than Zebra, | |
3024 | * has deleted | |
3025 | * a Zebra router from the kernel. We will add | |
3026 | * it back */ | |
3027 | rib_install_kernel(rn, fib, NULL); | |
3028 | } | |
3029 | } else { | |
3030 | if (IS_ZEBRA_DEBUG_RIB) { | |
fd36be7e | 3031 | if (nh) |
d62a17ae | 3032 | rnode_debug( |
3033 | rn, vrf_id, | |
3034 | "via %s ifindex %d type %d " | |
3035 | "doesn't exist in rib", | |
36228974 | 3036 | inet_ntop(afi2family(afi), |
3037 | &nh->gate, buf2, | |
3038 | sizeof(buf2)), | |
3039 | nh->ifindex, type); | |
d62a17ae | 3040 | else |
3041 | rnode_debug( | |
3042 | rn, vrf_id, | |
fd36be7e DL |
3043 | "type %d doesn't exist in rib", |
3044 | type); | |
d62a17ae | 3045 | } |
3046 | route_unlock_node(rn); | |
3047 | return; | |
3048 | } | |
3049 | } | |
718e3744 | 3050 | |
5dfeba19 | 3051 | if (same) { |
996c9314 LB |
3052 | if (fromkernel && CHECK_FLAG(flags, ZEBRA_FLAG_SELFROUTE) |
3053 | && !allow_delete) { | |
5dfeba19 DS |
3054 | rib_install_kernel(rn, same, NULL); |
3055 | route_unlock_node(rn); | |
3056 | ||
3057 | return; | |
3058 | } | |
6134fd82 | 3059 | |
2b83602b | 3060 | /* Special handling for IPv4 or IPv6 routes sourced from |
3061 | * EVPN - the nexthop (and associated MAC) need to be | |
3062 | * uninstalled if no more refs. | |
3063 | */ | |
90264d64 | 3064 | if (CHECK_FLAG(flags, ZEBRA_FLAG_EVPN_ROUTE)) { |
6134fd82 | 3065 | struct nexthop *tmp_nh; |
3066 | ||
7ee30f28 | 3067 | for (ALL_NEXTHOPS(re->ng, tmp_nh)) { |
6134fd82 | 3068 | struct ipaddr vtep_ip; |
3069 | ||
3070 | memset(&vtep_ip, 0, sizeof(struct ipaddr)); | |
1ec31309 | 3071 | if (afi == AFI_IP) { |
3072 | vtep_ip.ipa_type = IPADDR_V4; | |
3073 | memcpy(&(vtep_ip.ipaddr_v4), | |
3074 | &(tmp_nh->gate.ipv4), | |
3075 | sizeof(struct in_addr)); | |
3076 | } else { | |
3077 | vtep_ip.ipa_type = IPADDR_V6; | |
3078 | memcpy(&(vtep_ip.ipaddr_v6), | |
3079 | &(tmp_nh->gate.ipv6), | |
3080 | sizeof(struct in6_addr)); | |
3081 | } | |
a317a9b9 | 3082 | zebra_vxlan_evpn_vrf_route_del(re->vrf_id, |
6134fd82 | 3083 | &vtep_ip, p); |
3084 | } | |
3085 | } | |
cf363e1b MS |
3086 | |
3087 | /* Notify dplane if system route changes */ | |
3088 | if (RIB_SYSTEM_ROUTE(re)) | |
3089 | dplane_sys_route_del(rn, same); | |
3090 | ||
d62a17ae | 3091 | rib_delnode(rn, same); |
5dfeba19 | 3092 | } |
05737783 | 3093 | |
d62a17ae | 3094 | route_unlock_node(rn); |
3095 | return; | |
3096 | } | |
718e3744 | 3097 | |
718e3744 | 3098 | |
d7c0a89a QY |
3099 | int rib_add(afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, |
3100 | unsigned short instance, int flags, struct prefix *p, | |
3101 | struct prefix_ipv6 *src_p, const struct nexthop *nh, | |
3102 | uint32_t table_id, uint32_t metric, uint32_t mtu, uint8_t distance, | |
3103 | route_tag_t tag) | |
d62a17ae | 3104 | { |
3105 | struct route_entry *re; | |
66af6845 | 3106 | struct nexthop *nexthop; |
718e3744 | 3107 | |
66af6845 | 3108 | /* Allocate new route_entry structure. */ |
d62a17ae | 3109 | re = XCALLOC(MTYPE_RE, sizeof(struct route_entry)); |
d62a17ae | 3110 | re->type = type; |
3111 | re->instance = instance; | |
3112 | re->distance = distance; | |
3113 | re->flags = flags; | |
3114 | re->metric = metric; | |
3115 | re->mtu = mtu; | |
3116 | re->table = table_id; | |
3117 | re->vrf_id = vrf_id; | |
3118 | re->nexthop_num = 0; | |
98572489 | 3119 | re->uptime = monotime(NULL); |
4e40b6d6 | 3120 | re->tag = tag; |
d62a17ae | 3121 | |
66af6845 RW |
3122 | /* Add nexthop. */ |
3123 | nexthop = nexthop_new(); | |
3124 | *nexthop = *nh; | |
3125 | route_entry_nexthop_add(re, nexthop); | |
718e3744 | 3126 | |
66af6845 | 3127 | return rib_add_multipath(afi, safi, p, src_p, re); |
718e3744 | 3128 | } |
3129 | ||
1c848137 | 3130 | /* Schedule routes of a particular table (address-family) based on event. */ |
d5b8c216 | 3131 | void rib_update_table(struct route_table *table, rib_update_event_t event) |
d62a17ae | 3132 | { |
3133 | struct route_node *rn; | |
3134 | struct route_entry *re, *next; | |
3135 | ||
3136 | /* Walk all routes and queue for processing, if appropriate for | |
3137 | * the trigger event. | |
3138 | */ | |
3139 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { | |
1ca60f2c DS |
3140 | /* |
3141 | * If we are looking at a route node and the node | |
3142 | * has already been queued we don't | |
3143 | * need to queue it up again | |
3144 | */ | |
996c9314 LB |
3145 | if (rn->info && CHECK_FLAG(rib_dest_from_rnode(rn)->flags, |
3146 | RIB_ROUTE_ANY_QUEUED)) | |
1ca60f2c | 3147 | continue; |
d62a17ae | 3148 | switch (event) { |
3149 | case RIB_UPDATE_IF_CHANGE: | |
3150 | /* Examine all routes that won't get processed by the | |
3151 | * protocol or | |
3152 | * triggered by nexthop evaluation (NHT). This would be | |
3153 | * system, | |
3154 | * kernel and certain static routes. Note that NHT will | |
3155 | * get | |
3156 | * triggered upon an interface event as connected routes | |
3157 | * always | |
3158 | * get queued for processing. | |
3159 | */ | |
a2addae8 | 3160 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
0a16efff DS |
3161 | struct nexthop *nh; |
3162 | ||
996c9314 LB |
3163 | if (re->type != ZEBRA_ROUTE_SYSTEM |
3164 | && re->type != ZEBRA_ROUTE_KERNEL | |
3165 | && re->type != ZEBRA_ROUTE_CONNECT | |
3166 | && re->type != ZEBRA_ROUTE_STATIC) | |
0a16efff DS |
3167 | continue; |
3168 | ||
3169 | if (re->type != ZEBRA_ROUTE_STATIC) { | |
ade4a886 DS |
3170 | SET_FLAG(re->status, |
3171 | ROUTE_ENTRY_CHANGED); | |
0a16efff DS |
3172 | rib_queue_add(rn); |
3173 | continue; | |
3174 | } | |
3175 | ||
7ee30f28 | 3176 | for (nh = re->ng.nexthop; nh; nh = nh->next) |
0a16efff DS |
3177 | if (!(nh->type == NEXTHOP_TYPE_IPV4 |
3178 | || nh->type == NEXTHOP_TYPE_IPV6)) | |
3179 | break; | |
3180 | ||
3181 | /* If we only have nexthops to a | |
3182 | * gateway, NHT will | |
3183 | * take care. | |
3184 | */ | |
ade4a886 DS |
3185 | if (nh) { |
3186 | SET_FLAG(re->status, | |
3187 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 3188 | rib_queue_add(rn); |
ade4a886 | 3189 | } |
d62a17ae | 3190 | } |
3191 | break; | |
3192 | ||
3193 | case RIB_UPDATE_RMAP_CHANGE: | |
3194 | case RIB_UPDATE_OTHER: | |
3195 | /* Right now, examine all routes. Can restrict to a | |
3196 | * protocol in | |
3197 | * some cases (TODO). | |
3198 | */ | |
ade4a886 DS |
3199 | if (rnode_to_ribs(rn)) { |
3200 | RNODE_FOREACH_RE_SAFE (rn, re, next) | |
3201 | SET_FLAG(re->status, | |
3202 | ROUTE_ENTRY_CHANGED); | |
d62a17ae | 3203 | rib_queue_add(rn); |
ade4a886 | 3204 | } |
d62a17ae | 3205 | break; |
3206 | ||
3207 | default: | |
3208 | break; | |
3209 | } | |
3210 | } | |
b84c7253 | 3211 | } |
3212 | ||
718e3744 | 3213 | /* RIB update function. */ |
d62a17ae | 3214 | void rib_update(vrf_id_t vrf_id, rib_update_event_t event) |
718e3744 | 3215 | { |
d62a17ae | 3216 | struct route_table *table; |
1c848137 | 3217 | |
d62a17ae | 3218 | /* Process routes of interested address-families. */ |
3219 | table = zebra_vrf_table(AFI_IP, SAFI_UNICAST, vrf_id); | |
d5b8c216 | 3220 | if (table) { |
3221 | if (IS_ZEBRA_DEBUG_EVENT) | |
3222 | zlog_debug("%s : AFI_IP event %d", __func__, event); | |
d62a17ae | 3223 | rib_update_table(table, event); |
d5b8c216 | 3224 | } |
718e3744 | 3225 | |
d62a17ae | 3226 | table = zebra_vrf_table(AFI_IP6, SAFI_UNICAST, vrf_id); |
d5b8c216 | 3227 | if (table) { |
3228 | if (IS_ZEBRA_DEBUG_EVENT) | |
3229 | zlog_debug("%s : AFI_IP6 event %d", __func__, event); | |
d62a17ae | 3230 | rib_update_table(table, event); |
d5b8c216 | 3231 | } |
718e3744 | 3232 | } |
3233 | ||
718e3744 | 3234 | /* Delete self installed routes after zebra is relaunched. */ |
95a29032 | 3235 | void rib_sweep_table(struct route_table *table) |
d62a17ae | 3236 | { |
3237 | struct route_node *rn; | |
3238 | struct route_entry *re; | |
3239 | struct route_entry *next; | |
915902cb | 3240 | struct nexthop *nexthop; |
d62a17ae | 3241 | |
915902cb DS |
3242 | if (!table) |
3243 | return; | |
d62a17ae | 3244 | |
915902cb | 3245 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
a2addae8 | 3246 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d4644d41 | 3247 | |
915902cb DS |
3248 | if (IS_ZEBRA_DEBUG_RIB) |
3249 | route_entry_dump(&rn->p, NULL, re); | |
3250 | ||
3251 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) | |
3252 | continue; | |
3253 | ||
3254 | if (!CHECK_FLAG(re->flags, ZEBRA_FLAG_SELFROUTE)) | |
3255 | continue; | |
3256 | ||
d4644d41 DS |
3257 | /* |
3258 | * If routes are older than startup_time then | |
3259 | * we know we read them in from the kernel. | |
3260 | * As such we can safely remove them. | |
3261 | */ | |
3262 | if (zrouter.startup_time < re->uptime) | |
3263 | continue; | |
3264 | ||
915902cb DS |
3265 | /* |
3266 | * So we are starting up and have received | |
3267 | * routes from the kernel that we have installed | |
3268 | * from a previous run of zebra but not cleaned | |
3269 | * up ( say a kill -9 ) | |
3270 | * But since we haven't actually installed | |
3271 | * them yet( we received them from the kernel ) | |
3272 | * we don't think they are active. | |
3273 | * So let's pretend they are active to actually | |
3274 | * remove them. | |
3275 | * In all honesty I'm not sure if we should | |
3276 | * mark them as active when we receive them | |
3277 | * This is startup only so probably ok. | |
3278 | * | |
3279 | * If we ever decide to move rib_sweep_table | |
3280 | * to a different spot (ie startup ) | |
3281 | * this decision needs to be revisited | |
3282 | */ | |
677c1dd5 | 3283 | SET_FLAG(re->status, ROUTE_ENTRY_INSTALLED); |
7ee30f28 | 3284 | for (ALL_NEXTHOPS(re->ng, nexthop)) |
915902cb DS |
3285 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB); |
3286 | ||
0c555cc6 DS |
3287 | rib_uninstall_kernel(rn, re); |
3288 | rib_delnode(rn, re); | |
915902cb DS |
3289 | } |
3290 | } | |
718e3744 | 3291 | } |
3292 | ||
3293 | /* Sweep all RIB tables. */ | |
d4644d41 | 3294 | int rib_sweep_route(struct thread *t) |
718e3744 | 3295 | { |
d62a17ae | 3296 | struct vrf *vrf; |
3297 | struct zebra_vrf *zvrf; | |
78104b9b | 3298 | |
a2addae8 | 3299 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
915902cb DS |
3300 | if ((zvrf = vrf->info) == NULL) |
3301 | continue; | |
3302 | ||
d62a17ae | 3303 | rib_sweep_table(zvrf->table[AFI_IP][SAFI_UNICAST]); |
3304 | rib_sweep_table(zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3305 | } | |
95a29032 | 3306 | |
89272910 | 3307 | zebra_router_sweep_route(); |
d4644d41 DS |
3308 | |
3309 | return 0; | |
718e3744 | 3310 | } |
2ea1ab1c VT |
3311 | |
3312 | /* Remove specific by protocol routes from 'table'. */ | |
d7c0a89a | 3313 | unsigned long rib_score_proto_table(uint8_t proto, unsigned short instance, |
47a08aa9 | 3314 | struct route_table *table) |
d62a17ae | 3315 | { |
3316 | struct route_node *rn; | |
3317 | struct route_entry *re; | |
3318 | struct route_entry *next; | |
3319 | unsigned long n = 0; | |
3320 | ||
3321 | if (table) | |
3322 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) | |
a2addae8 | 3323 | RNODE_FOREACH_RE_SAFE (rn, re, next) { |
d62a17ae | 3324 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_REMOVED)) |
3325 | continue; | |
3326 | if (re->type == proto | |
3327 | && re->instance == instance) { | |
3328 | rib_delnode(rn, re); | |
3329 | n++; | |
3330 | } | |
3331 | } | |
3332 | return n; | |
2ea1ab1c VT |
3333 | } |
3334 | ||
3335 | /* Remove specific by protocol routes. */ | |
d7c0a89a | 3336 | unsigned long rib_score_proto(uint8_t proto, unsigned short instance) |
2ea1ab1c | 3337 | { |
d62a17ae | 3338 | struct vrf *vrf; |
3339 | struct zebra_vrf *zvrf; | |
d8612e65 | 3340 | struct other_route_table *ort; |
d62a17ae | 3341 | unsigned long cnt = 0; |
78104b9b | 3342 | |
d8612e65 DS |
3343 | RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) { |
3344 | zvrf = vrf->info; | |
3345 | if (!zvrf) | |
3346 | continue; | |
78104b9b | 3347 | |
d8612e65 DS |
3348 | cnt += rib_score_proto_table(proto, instance, |
3349 | zvrf->table[AFI_IP][SAFI_UNICAST]) | |
3350 | + rib_score_proto_table( | |
3351 | proto, instance, | |
3352 | zvrf->table[AFI_IP6][SAFI_UNICAST]); | |
3353 | ||
81fddbe7 | 3354 | frr_each(otable, &zvrf->other_tables, ort) cnt += |
d8612e65 DS |
3355 | rib_score_proto_table(proto, instance, ort->table); |
3356 | } | |
47a08aa9 | 3357 | |
d62a17ae | 3358 | return cnt; |
2ea1ab1c VT |
3359 | } |
3360 | ||
718e3744 | 3361 | /* Close RIB and clean up kernel routes. */ |
d62a17ae | 3362 | void rib_close_table(struct route_table *table) |
718e3744 | 3363 | { |
d62a17ae | 3364 | struct route_node *rn; |
1e9f448f | 3365 | rib_table_info_t *info; |
5f7a4718 | 3366 | rib_dest_t *dest; |
718e3744 | 3367 | |
1e9f448f DS |
3368 | if (!table) |
3369 | return; | |
9fd92e3c | 3370 | |
6ca30e9e | 3371 | info = route_table_get_info(table); |
5adc2528 | 3372 | |
5f7a4718 DS |
3373 | for (rn = route_top(table); rn; rn = srcdest_route_next(rn)) { |
3374 | dest = rib_dest_from_rnode(rn); | |
1e9f448f | 3375 | |
5f7a4718 | 3376 | if (dest && dest->selected_fib) { |
1e9f448f DS |
3377 | if (info->safi == SAFI_UNICAST) |
3378 | hook_call(rib_update, rn, NULL); | |
3379 | ||
3cdba47a DS |
3380 | rib_uninstall_kernel(rn, dest->selected_fib); |
3381 | dest->selected_fib = NULL; | |
1e9f448f | 3382 | } |
5f7a4718 | 3383 | } |
718e3744 | 3384 | } |
3385 | ||
9bd9717b MS |
3386 | /* |
3387 | * Handler for async dataplane results after a pseudowire installation | |
3388 | */ | |
3389 | static int handle_pw_result(struct zebra_dplane_ctx *ctx) | |
3390 | { | |
9bd9717b MS |
3391 | struct zebra_pw *pw; |
3392 | struct zebra_vrf *vrf; | |
3393 | ||
3394 | /* The pseudowire code assumes success - we act on an error | |
3395 | * result for installation attempts here. | |
3396 | */ | |
3397 | if (dplane_ctx_get_op(ctx) != DPLANE_OP_PW_INSTALL) | |
3398 | goto done; | |
3399 | ||
3400 | if (dplane_ctx_get_status(ctx) != ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
3401 | vrf = zebra_vrf_lookup_by_id(dplane_ctx_get_vrf(ctx)); | |
3402 | pw = zebra_pw_find(vrf, dplane_ctx_get_pw_ifname(ctx)); | |
3403 | if (pw) | |
3404 | zebra_pw_install_failure(pw); | |
3405 | } | |
3406 | ||
3407 | done: | |
3408 | ||
9f2d0354 | 3409 | return 0; |
9bd9717b MS |
3410 | } |
3411 | ||
3412 | ||
7cdb1a84 | 3413 | /* |
d37f4d6c MS |
3414 | * Handle results from the dataplane system. Dequeue update context |
3415 | * structs, dispatch to appropriate internal handlers. | |
7cdb1a84 MS |
3416 | */ |
3417 | static int rib_process_dplane_results(struct thread *thread) | |
3418 | { | |
25779064 | 3419 | struct zebra_dplane_ctx *ctx; |
4c206c8f MS |
3420 | struct dplane_ctx_q ctxlist; |
3421 | ||
3422 | /* Dequeue a list of completed updates with one lock/unlock cycle */ | |
7cdb1a84 MS |
3423 | |
3424 | do { | |
4c206c8f MS |
3425 | TAILQ_INIT(&ctxlist); |
3426 | ||
7cdb1a84 MS |
3427 | /* Take lock controlling queue of results */ |
3428 | pthread_mutex_lock(&dplane_mutex); | |
3429 | { | |
d37f4d6c | 3430 | /* Dequeue list of context structs */ |
4c206c8f | 3431 | dplane_ctx_list_append(&ctxlist, &rib_dplane_q); |
7cdb1a84 MS |
3432 | } |
3433 | pthread_mutex_unlock(&dplane_mutex); | |
3434 | ||
4c206c8f MS |
3435 | /* Dequeue context block */ |
3436 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3437 | ||
3438 | /* If we've emptied the results queue, we're done */ | |
3439 | if (ctx == NULL) | |
7cdb1a84 | 3440 | break; |
7cdb1a84 | 3441 | |
4c206c8f | 3442 | while (ctx) { |
d37f4d6c MS |
3443 | switch (dplane_ctx_get_op(ctx)) { |
3444 | case DPLANE_OP_ROUTE_INSTALL: | |
3445 | case DPLANE_OP_ROUTE_UPDATE: | |
3446 | case DPLANE_OP_ROUTE_DELETE: | |
3447 | rib_process_result(ctx); | |
3448 | break; | |
3449 | ||
54818e3b MS |
3450 | case DPLANE_OP_ROUTE_NOTIFY: |
3451 | rib_process_dplane_notify(ctx); | |
3452 | break; | |
3453 | ||
d37f4d6c MS |
3454 | case DPLANE_OP_LSP_INSTALL: |
3455 | case DPLANE_OP_LSP_UPDATE: | |
3456 | case DPLANE_OP_LSP_DELETE: | |
3457 | zebra_mpls_lsp_dplane_result(ctx); | |
3458 | break; | |
3459 | ||
9bd9717b MS |
3460 | case DPLANE_OP_PW_INSTALL: |
3461 | case DPLANE_OP_PW_UNINSTALL: | |
3462 | handle_pw_result(ctx); | |
3463 | break; | |
3464 | ||
cf363e1b MS |
3465 | case DPLANE_OP_SYS_ROUTE_ADD: |
3466 | case DPLANE_OP_SYS_ROUTE_DELETE: | |
3467 | /* No further processing in zebra for these. */ | |
3468 | dplane_ctx_fini(&ctx); | |
3469 | break; | |
3470 | ||
d37f4d6c MS |
3471 | default: |
3472 | /* Don't expect this: just return the struct? */ | |
3473 | dplane_ctx_fini(&ctx); | |
3474 | break; | |
3475 | } /* Dispatch by op code */ | |
4c206c8f MS |
3476 | |
3477 | ctx = dplane_ctx_dequeue(&ctxlist); | |
3478 | } | |
3479 | ||
5709131c | 3480 | } while (1); |
7cdb1a84 | 3481 | |
f183e380 MS |
3482 | /* Check for nexthop tracking processing after finishing with results */ |
3483 | do_nht_processing(); | |
3484 | ||
5709131c | 3485 | return 0; |
7cdb1a84 MS |
3486 | } |
3487 | ||
3488 | /* | |
3489 | * Results are returned from the dataplane subsystem, in the context of | |
1bcea841 | 3490 | * the dataplane pthread. We enqueue the results here for processing by |
7cdb1a84 MS |
3491 | * the main thread later. |
3492 | */ | |
4c206c8f | 3493 | static int rib_dplane_results(struct dplane_ctx_q *ctxlist) |
7cdb1a84 MS |
3494 | { |
3495 | /* Take lock controlling queue of results */ | |
3496 | pthread_mutex_lock(&dplane_mutex); | |
3497 | { | |
4c206c8f MS |
3498 | /* Enqueue context blocks */ |
3499 | dplane_ctx_list_append(&rib_dplane_q, ctxlist); | |
7cdb1a84 MS |
3500 | } |
3501 | pthread_mutex_unlock(&dplane_mutex); | |
3502 | ||
4c206c8f | 3503 | /* Ensure event is signalled to zebra main pthread */ |
3801e764 | 3504 | thread_add_event(zrouter.master, rib_process_dplane_results, NULL, 0, |
7cdb1a84 MS |
3505 | &t_dplane); |
3506 | ||
5709131c | 3507 | return 0; |
7cdb1a84 MS |
3508 | } |
3509 | ||
eaa2716d SW |
3510 | /* |
3511 | * Ensure there are no empty slots in the route_info array. | |
3512 | * Every route type in zebra should be present there. | |
3513 | */ | |
3514 | static void check_route_info(void) | |
3515 | { | |
3516 | int len = array_size(route_info); | |
3517 | ||
3518 | /* | |
3519 | * ZEBRA_ROUTE_SYSTEM is special cased since | |
3520 | * its key is 0 anyway. | |
3521 | * | |
3522 | * ZEBRA_ROUTE_ALL is also ignored. | |
3523 | */ | |
3524 | for (int i = 0; i < len; i++) { | |
3525 | if (i == ZEBRA_ROUTE_SYSTEM || i == ZEBRA_ROUTE_ALL) | |
3526 | continue; | |
3527 | assert(route_info[i].key); | |
8dc7a759 | 3528 | assert(route_info[i].meta_q_map < MQ_SIZE); |
eaa2716d SW |
3529 | } |
3530 | } | |
3531 | ||
718e3744 | 3532 | /* Routing information base initialize. */ |
d62a17ae | 3533 | void rib_init(void) |
718e3744 | 3534 | { |
eaa2716d SW |
3535 | check_route_info(); |
3536 | ||
2561d12e | 3537 | rib_queue_init(); |
7cdb1a84 MS |
3538 | |
3539 | /* Init dataplane, and register for results */ | |
3540 | pthread_mutex_init(&dplane_mutex, NULL); | |
3541 | TAILQ_INIT(&rib_dplane_q); | |
4c206c8f | 3542 | zebra_dplane_init(rib_dplane_results); |
718e3744 | 3543 | } |
0915bb0c AS |
3544 | |
3545 | /* | |
3546 | * vrf_id_get_next | |
3547 | * | |
3548 | * Get the first vrf id that is greater than the given vrf id if any. | |
3549 | * | |
3550 | * Returns TRUE if a vrf id was found, FALSE otherwise. | |
3551 | */ | |
d62a17ae | 3552 | static inline int vrf_id_get_next(vrf_id_t vrf_id, vrf_id_t *next_id_p) |
0915bb0c | 3553 | { |
d62a17ae | 3554 | struct vrf *vrf; |
b72ede27 | 3555 | |
d62a17ae | 3556 | vrf = vrf_lookup_by_id(vrf_id); |
3557 | if (vrf) { | |
3558 | vrf = RB_NEXT(vrf_id_head, vrf); | |
3559 | if (vrf) { | |
3560 | *next_id_p = vrf->vrf_id; | |
3561 | return 1; | |
3562 | } | |
3563 | } | |
0915bb0c | 3564 | |
d62a17ae | 3565 | return 0; |
0915bb0c AS |
3566 | } |
3567 | ||
3568 | /* | |
3569 | * rib_tables_iter_next | |
3570 | * | |
3571 | * Returns the next table in the iteration. | |
3572 | */ | |
d62a17ae | 3573 | struct route_table *rib_tables_iter_next(rib_tables_iter_t *iter) |
3574 | { | |
3575 | struct route_table *table; | |
3576 | ||
3577 | /* | |
3578 | * Array that helps us go over all AFI/SAFI combinations via one | |
3579 | * index. | |
3580 | */ | |
3581 | static struct { | |
3582 | afi_t afi; | |
3583 | safi_t safi; | |
3584 | } afi_safis[] = { | |
3585 | {AFI_IP, SAFI_UNICAST}, {AFI_IP, SAFI_MULTICAST}, | |
3586 | {AFI_IP, SAFI_LABELED_UNICAST}, {AFI_IP6, SAFI_UNICAST}, | |
3587 | {AFI_IP6, SAFI_MULTICAST}, {AFI_IP6, SAFI_LABELED_UNICAST}, | |
3588 | }; | |
3589 | ||
3590 | table = NULL; | |
3591 | ||
3592 | switch (iter->state) { | |
3593 | ||
3594 | case RIB_TABLES_ITER_S_INIT: | |
3595 | iter->vrf_id = VRF_DEFAULT; | |
3596 | iter->afi_safi_ix = -1; | |
3597 | ||
3598 | /* Fall through */ | |
3599 | ||
3600 | case RIB_TABLES_ITER_S_ITERATING: | |
3601 | iter->afi_safi_ix++; | |
3602 | while (1) { | |
3603 | ||
3604 | while (iter->afi_safi_ix | |
7e3a1ec7 | 3605 | < (int)array_size(afi_safis)) { |
d62a17ae | 3606 | table = zebra_vrf_table( |
3607 | afi_safis[iter->afi_safi_ix].afi, | |
3608 | afi_safis[iter->afi_safi_ix].safi, | |
3609 | iter->vrf_id); | |
3610 | if (table) | |
3611 | break; | |
3612 | ||
3613 | iter->afi_safi_ix++; | |
3614 | } | |
3615 | ||
3616 | /* | |
3617 | * Found another table in this vrf. | |
3618 | */ | |
3619 | if (table) | |
3620 | break; | |
3621 | ||
3622 | /* | |
3623 | * Done with all tables in the current vrf, go to the | |
3624 | * next | |
3625 | * one. | |
3626 | */ | |
3627 | if (!vrf_id_get_next(iter->vrf_id, &iter->vrf_id)) | |
3628 | break; | |
3629 | ||
3630 | iter->afi_safi_ix = 0; | |
3631 | } | |
0915bb0c | 3632 | |
0915bb0c AS |
3633 | break; |
3634 | ||
d62a17ae | 3635 | case RIB_TABLES_ITER_S_DONE: |
3636 | return NULL; | |
0915bb0c AS |
3637 | } |
3638 | ||
d62a17ae | 3639 | if (table) |
3640 | iter->state = RIB_TABLES_ITER_S_ITERATING; | |
3641 | else | |
3642 | iter->state = RIB_TABLES_ITER_S_DONE; | |
0915bb0c | 3643 | |
d62a17ae | 3644 | return table; |
0915bb0c | 3645 | } |