]>
Commit | Line | Data |
---|---|---|
ad28e79a SW |
1 | /* Zebra Nexthop Group Code. |
2 | * Copyright (C) 2019 Cumulus Networks, Inc. | |
3 | * Donald Sharp | |
4 | * Stephen Worley | |
5 | * | |
6 | * This file is part of FRR. | |
7 | * | |
8 | * FRR is free software; you can redistribute it and/or modify it | |
9 | * under the terms of the GNU General Public License as published by the | |
10 | * Free Software Foundation; either version 2, or (at your option) any | |
11 | * later version. | |
12 | * | |
13 | * FRR is distributed in the hope that it will be useful, but | |
14 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 | * General Public License for more details. | |
17 | * | |
18 | * You should have received a copy of the GNU General Public License | |
19 | * along with FRR; see the file COPYING. If not, write to the Free | |
20 | * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA | |
21 | * 02111-1307, USA. | |
22 | */ | |
23 | #include <zebra.h> | |
24 | ||
25 | #include "lib/nexthop.h" | |
50d89650 | 26 | #include "lib/nexthop_group_private.h" |
ad28e79a | 27 | #include "lib/routemap.h" |
b43434ad | 28 | #include "lib/mpls.h" |
69171da2 | 29 | #include "lib/jhash.h" |
51d80884 | 30 | #include "lib/debug.h" |
31f937fb | 31 | #include "lib/lib_errors.h" |
ad28e79a SW |
32 | |
33 | #include "zebra/connected.h" | |
34 | #include "zebra/debug.h" | |
35 | #include "zebra/zebra_router.h" | |
5948f013 | 36 | #include "zebra/zebra_nhg_private.h" |
ad28e79a SW |
37 | #include "zebra/zebra_rnh.h" |
38 | #include "zebra/zebra_routemap.h" | |
31f937fb | 39 | #include "zebra/zebra_srte.h" |
51d80884 | 40 | #include "zebra/zserv.h" |
ad28e79a | 41 | #include "zebra/rt.h" |
d9f5b2f5 | 42 | #include "zebra_errors.h" |
0c8215cb | 43 | #include "zebra_dplane.h" |
fe593b78 | 44 | #include "zebra/interface.h" |
ee94437e | 45 | #include "zebra/zapi_msg.h" |
d9f5b2f5 | 46 | |
51d80884 | 47 | DEFINE_MTYPE_STATIC(ZEBRA, NHG, "Nexthop Group Entry"); |
a15d4c00 | 48 | DEFINE_MTYPE_STATIC(ZEBRA, NHG_CONNECTED, "Nexthop Group Connected"); |
e22e8001 | 49 | DEFINE_MTYPE_STATIC(ZEBRA, NHG_CTX, "Nexthop Group Context"); |
0c8215cb | 50 | |
5530d55d MS |
51 | /* Map backup nexthop indices between two nhes */ |
52 | struct backup_nh_map_s { | |
53 | int map_count; | |
54 | ||
55 | struct { | |
56 | uint8_t orig_idx; | |
57 | uint8_t new_idx; | |
58 | } map[MULTIPATH_NUM]; | |
59 | }; | |
60 | ||
38e40db1 SW |
61 | /* id counter to keep in sync with kernel */ |
62 | uint32_t id_counter; | |
63 | ||
aa458838 | 64 | /* Controlled through ui */ |
7c99d51b | 65 | static bool g_nexthops_enabled = true; |
73937edb | 66 | static bool proto_nexthops_only; |
aa458838 | 67 | static bool use_recursive_backups = true; |
7c99d51b | 68 | |
0885b1e3 | 69 | static struct nhg_hash_entry *depends_find(const struct nexthop *nh, afi_t afi, |
5588801e | 70 | int type, bool from_dplane); |
37c6708b | 71 | static void depends_add(struct nhg_connected_tree_head *head, |
5657e7e9 | 72 | struct nhg_hash_entry *depend); |
38e40db1 SW |
73 | static struct nhg_hash_entry * |
74 | depends_find_add(struct nhg_connected_tree_head *head, struct nexthop *nh, | |
5588801e | 75 | afi_t afi, int type, bool from_dplane); |
38e40db1 SW |
76 | static struct nhg_hash_entry * |
77 | depends_find_id_add(struct nhg_connected_tree_head *head, uint32_t id); | |
37c6708b | 78 | static void depends_decrement_free(struct nhg_connected_tree_head *head); |
0c8215cb | 79 | |
1d48702e MS |
80 | static struct nhg_backup_info * |
81 | nhg_backup_copy(const struct nhg_backup_info *orig); | |
82 | ||
ac5d1091 | 83 | /* Helper function for getting the next allocatable ID */ |
73937edb | 84 | static uint32_t nhg_get_next_id(void) |
ac5d1091 SW |
85 | { |
86 | while (1) { | |
87 | id_counter++; | |
88 | ||
89 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
90 | zlog_debug("%s: ID %u checking", __func__, id_counter); | |
91 | ||
92 | if (id_counter == ZEBRA_NHG_PROTO_LOWER) { | |
93 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
94 | zlog_debug("%s: ID counter wrapped", __func__); | |
95 | ||
96 | id_counter = 0; | |
97 | continue; | |
98 | } | |
99 | ||
100 | if (zebra_nhg_lookup_id(id_counter)) { | |
101 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
102 | zlog_debug("%s: ID already exists", __func__); | |
103 | ||
104 | continue; | |
105 | } | |
106 | ||
107 | break; | |
108 | } | |
109 | ||
110 | return id_counter; | |
111 | } | |
e22e8001 | 112 | |
5948f013 | 113 | static void nhg_connected_free(struct nhg_connected *dep) |
0c8215cb | 114 | { |
a15d4c00 | 115 | XFREE(MTYPE_NHG_CONNECTED, dep); |
0c8215cb SW |
116 | } |
117 | ||
5948f013 | 118 | static struct nhg_connected *nhg_connected_new(struct nhg_hash_entry *nhe) |
0c8215cb | 119 | { |
a15d4c00 | 120 | struct nhg_connected *new = NULL; |
0c8215cb | 121 | |
a15d4c00 | 122 | new = XCALLOC(MTYPE_NHG_CONNECTED, sizeof(struct nhg_connected)); |
0c8215cb SW |
123 | new->nhe = nhe; |
124 | ||
125 | return new; | |
126 | } | |
127 | ||
37c6708b | 128 | void nhg_connected_tree_free(struct nhg_connected_tree_head *head) |
0c8215cb | 129 | { |
a15d4c00 | 130 | struct nhg_connected *rb_node_dep = NULL; |
0c8215cb | 131 | |
37c6708b | 132 | if (!nhg_connected_tree_is_empty(head)) { |
fec211ad | 133 | frr_each_safe(nhg_connected_tree, head, rb_node_dep) { |
37c6708b | 134 | nhg_connected_tree_del(head, rb_node_dep); |
fe593b78 SW |
135 | nhg_connected_free(rb_node_dep); |
136 | } | |
0c8215cb | 137 | } |
0c8215cb SW |
138 | } |
139 | ||
37c6708b | 140 | bool nhg_connected_tree_is_empty(const struct nhg_connected_tree_head *head) |
0c8215cb | 141 | { |
fec211ad | 142 | return nhg_connected_tree_count(head) ? false : true; |
0c8215cb SW |
143 | } |
144 | ||
98cda54a | 145 | struct nhg_connected * |
37c6708b | 146 | nhg_connected_tree_root(struct nhg_connected_tree_head *head) |
98cda54a | 147 | { |
37c6708b | 148 | return nhg_connected_tree_first(head); |
98cda54a SW |
149 | } |
150 | ||
5bf15faa SW |
151 | struct nhg_hash_entry * |
152 | nhg_connected_tree_del_nhe(struct nhg_connected_tree_head *head, | |
153 | struct nhg_hash_entry *depend) | |
0c8215cb | 154 | { |
a15d4c00 | 155 | struct nhg_connected lookup = {}; |
085304dc | 156 | struct nhg_connected *remove = NULL; |
5bf15faa | 157 | struct nhg_hash_entry *removed_nhe; |
0c8215cb SW |
158 | |
159 | lookup.nhe = depend; | |
3119f6a1 | 160 | |
085304dc | 161 | /* Lookup to find the element, then remove it */ |
37c6708b | 162 | remove = nhg_connected_tree_find(head, &lookup); |
085304dc | 163 | if (remove) |
5bf15faa SW |
164 | /* Re-returning here just in case this API changes.. |
165 | * the _del list api's are a bit undefined at the moment. | |
166 | * | |
167 | * So hopefully returning here will make it fail if the api | |
168 | * changes to something different than currently expected. | |
169 | */ | |
170 | remove = nhg_connected_tree_del(head, remove); | |
171 | ||
172 | /* If the entry was sucessfully removed, free the 'connected` struct */ | |
173 | if (remove) { | |
174 | removed_nhe = remove->nhe; | |
085304dc | 175 | nhg_connected_free(remove); |
5bf15faa SW |
176 | return removed_nhe; |
177 | } | |
178 | ||
179 | return NULL; | |
3119f6a1 SW |
180 | } |
181 | ||
5bf15faa SW |
182 | /* Assuming UNIQUE RB tree. If this changes, assumptions here about |
183 | * insertion need to change. | |
184 | */ | |
185 | struct nhg_hash_entry * | |
186 | nhg_connected_tree_add_nhe(struct nhg_connected_tree_head *head, | |
187 | struct nhg_hash_entry *depend) | |
3119f6a1 | 188 | { |
a15d4c00 | 189 | struct nhg_connected *new = NULL; |
0c8215cb | 190 | |
a15d4c00 | 191 | new = nhg_connected_new(depend); |
0c8215cb | 192 | |
5bf15faa SW |
193 | /* On success, NULL will be returned from the |
194 | * RB code. | |
195 | */ | |
196 | if (new && (nhg_connected_tree_add(head, new) == NULL)) | |
197 | return NULL; | |
198 | ||
199 | /* If it wasn't successful, it must be a duplicate. We enforce the | |
200 | * unique property for the `nhg_connected` tree. | |
201 | */ | |
202 | nhg_connected_free(new); | |
203 | ||
204 | return depend; | |
3119f6a1 SW |
205 | } |
206 | ||
37c6708b SW |
207 | static void |
208 | nhg_connected_tree_decrement_ref(struct nhg_connected_tree_head *head) | |
32e29e79 SW |
209 | { |
210 | struct nhg_connected *rb_node_dep = NULL; | |
32e29e79 | 211 | |
fec211ad | 212 | frr_each_safe(nhg_connected_tree, head, rb_node_dep) { |
32e29e79 SW |
213 | zebra_nhg_decrement_ref(rb_node_dep->nhe); |
214 | } | |
215 | } | |
216 | ||
37c6708b SW |
217 | static void |
218 | nhg_connected_tree_increment_ref(struct nhg_connected_tree_head *head) | |
32e29e79 SW |
219 | { |
220 | struct nhg_connected *rb_node_dep = NULL; | |
221 | ||
fec211ad | 222 | frr_each(nhg_connected_tree, head, rb_node_dep) { |
32e29e79 SW |
223 | zebra_nhg_increment_ref(rb_node_dep->nhe); |
224 | } | |
225 | } | |
226 | ||
98cda54a SW |
227 | struct nhg_hash_entry *zebra_nhg_resolve(struct nhg_hash_entry *nhe) |
228 | { | |
229 | if (CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_RECURSIVE) | |
230 | && !zebra_nhg_depends_is_empty(nhe)) { | |
37c6708b | 231 | nhe = nhg_connected_tree_root(&nhe->nhg_depends)->nhe; |
98cda54a SW |
232 | return zebra_nhg_resolve(nhe); |
233 | } | |
234 | ||
235 | return nhe; | |
236 | } | |
237 | ||
fe593b78 | 238 | unsigned int zebra_nhg_depends_count(const struct nhg_hash_entry *nhe) |
a15d4c00 | 239 | { |
37c6708b | 240 | return nhg_connected_tree_count(&nhe->nhg_depends); |
a15d4c00 SW |
241 | } |
242 | ||
fe593b78 | 243 | bool zebra_nhg_depends_is_empty(const struct nhg_hash_entry *nhe) |
a15d4c00 | 244 | { |
37c6708b | 245 | return nhg_connected_tree_is_empty(&nhe->nhg_depends); |
a15d4c00 SW |
246 | } |
247 | ||
5948f013 SW |
248 | static void zebra_nhg_depends_del(struct nhg_hash_entry *from, |
249 | struct nhg_hash_entry *depend) | |
3119f6a1 | 250 | { |
37c6708b | 251 | nhg_connected_tree_del_nhe(&from->nhg_depends, depend); |
3119f6a1 SW |
252 | } |
253 | ||
5948f013 | 254 | static void zebra_nhg_depends_init(struct nhg_hash_entry *nhe) |
148a0103 | 255 | { |
37c6708b | 256 | nhg_connected_tree_init(&nhe->nhg_depends); |
148a0103 SW |
257 | } |
258 | ||
fe593b78 SW |
259 | unsigned int zebra_nhg_dependents_count(const struct nhg_hash_entry *nhe) |
260 | { | |
37c6708b | 261 | return nhg_connected_tree_count(&nhe->nhg_dependents); |
fe593b78 SW |
262 | } |
263 | ||
5948f013 | 264 | |
fe593b78 SW |
265 | bool zebra_nhg_dependents_is_empty(const struct nhg_hash_entry *nhe) |
266 | { | |
37c6708b | 267 | return nhg_connected_tree_is_empty(&nhe->nhg_dependents); |
fe593b78 SW |
268 | } |
269 | ||
5948f013 SW |
270 | static void zebra_nhg_dependents_del(struct nhg_hash_entry *from, |
271 | struct nhg_hash_entry *dependent) | |
fe593b78 | 272 | { |
37c6708b | 273 | nhg_connected_tree_del_nhe(&from->nhg_dependents, dependent); |
fe593b78 SW |
274 | } |
275 | ||
5948f013 SW |
276 | static void zebra_nhg_dependents_add(struct nhg_hash_entry *to, |
277 | struct nhg_hash_entry *dependent) | |
fe593b78 | 278 | { |
37c6708b | 279 | nhg_connected_tree_add_nhe(&to->nhg_dependents, dependent); |
fe593b78 SW |
280 | } |
281 | ||
5948f013 | 282 | static void zebra_nhg_dependents_init(struct nhg_hash_entry *nhe) |
fe593b78 | 283 | { |
37c6708b | 284 | nhg_connected_tree_init(&nhe->nhg_dependents); |
fe593b78 SW |
285 | } |
286 | ||
21615102 SW |
287 | /* Release this nhe from anything depending on it */ |
288 | static void zebra_nhg_dependents_release(struct nhg_hash_entry *nhe) | |
289 | { | |
80286aa5 | 290 | struct nhg_connected *rb_node_dep = NULL; |
21615102 | 291 | |
80286aa5 SW |
292 | frr_each_safe(nhg_connected_tree, &nhe->nhg_dependents, rb_node_dep) { |
293 | zebra_nhg_depends_del(rb_node_dep->nhe, nhe); | |
294 | /* recheck validity of the dependent */ | |
295 | zebra_nhg_check_valid(rb_node_dep->nhe); | |
21615102 SW |
296 | } |
297 | } | |
298 | ||
5948f013 SW |
299 | /* Release this nhe from anything that it depends on */ |
300 | static void zebra_nhg_depends_release(struct nhg_hash_entry *nhe) | |
301 | { | |
302 | if (!zebra_nhg_depends_is_empty(nhe)) { | |
303 | struct nhg_connected *rb_node_dep = NULL; | |
304 | ||
305 | frr_each_safe(nhg_connected_tree, &nhe->nhg_depends, | |
306 | rb_node_dep) { | |
307 | zebra_nhg_dependents_del(rb_node_dep->nhe, nhe); | |
308 | } | |
309 | } | |
310 | } | |
311 | ||
312 | ||
d9f5b2f5 SW |
313 | struct nhg_hash_entry *zebra_nhg_lookup_id(uint32_t id) |
314 | { | |
0c8215cb | 315 | struct nhg_hash_entry lookup = {}; |
d9f5b2f5 SW |
316 | |
317 | lookup.id = id; | |
318 | return hash_lookup(zrouter.nhgs_id, &lookup); | |
319 | } | |
320 | ||
5948f013 | 321 | static int zebra_nhg_insert_id(struct nhg_hash_entry *nhe) |
d9f5b2f5 SW |
322 | { |
323 | if (hash_lookup(zrouter.nhgs_id, nhe)) { | |
324 | flog_err( | |
325 | EC_ZEBRA_NHG_TABLE_INSERT_FAILED, | |
326 | "Failed inserting NHG id=%u into the ID hash table, entry already exists", | |
327 | nhe->id); | |
328 | return -1; | |
329 | } | |
330 | ||
331 | hash_get(zrouter.nhgs_id, nhe, hash_alloc_intern); | |
332 | ||
333 | return 0; | |
334 | } | |
ad28e79a | 335 | |
5948f013 SW |
336 | static void zebra_nhg_set_if(struct nhg_hash_entry *nhe, struct interface *ifp) |
337 | { | |
338 | nhe->ifp = ifp; | |
339 | if_nhg_dependents_add(ifp, nhe); | |
340 | } | |
341 | ||
38e40db1 SW |
342 | static void |
343 | zebra_nhg_connect_depends(struct nhg_hash_entry *nhe, | |
377e29f7 | 344 | struct nhg_connected_tree_head *nhg_depends) |
4e49c8b8 | 345 | { |
a15d4c00 SW |
346 | struct nhg_connected *rb_node_dep = NULL; |
347 | ||
38e40db1 SW |
348 | /* This has been allocated higher above in the stack. Could probably |
349 | * re-allocate and free the old stuff but just using the same memory | |
350 | * for now. Otherwise, their might be a time trade-off for repeated | |
351 | * alloc/frees as startup. | |
352 | */ | |
377e29f7 | 353 | nhe->nhg_depends = *nhg_depends; |
4e49c8b8 | 354 | |
a15d4c00 | 355 | /* Attach backpointer to anything that it depends on */ |
fe593b78 | 356 | zebra_nhg_dependents_init(nhe); |
a15d4c00 | 357 | if (!zebra_nhg_depends_is_empty(nhe)) { |
fec211ad | 358 | frr_each(nhg_connected_tree, &nhe->nhg_depends, rb_node_dep) { |
377e29f7 MS |
359 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
360 | zlog_debug("%s: nhe %p (%u), dep %p (%u)", | |
361 | __func__, nhe, nhe->id, | |
362 | rb_node_dep->nhe, | |
363 | rb_node_dep->nhe->id); | |
364 | ||
a15d4c00 SW |
365 | zebra_nhg_dependents_add(rb_node_dep->nhe, nhe); |
366 | } | |
367 | } | |
377e29f7 | 368 | } |
4e49c8b8 | 369 | |
377e29f7 MS |
370 | /* Init an nhe, for use in a hash lookup for example */ |
371 | void zebra_nhe_init(struct nhg_hash_entry *nhe, afi_t afi, | |
372 | const struct nexthop *nh) | |
373 | { | |
374 | memset(nhe, 0, sizeof(struct nhg_hash_entry)); | |
375 | nhe->vrf_id = VRF_DEFAULT; | |
376 | nhe->type = ZEBRA_ROUTE_NHG; | |
377 | nhe->afi = AFI_UNSPEC; | |
7b683a96 | 378 | |
377e29f7 MS |
379 | /* There are some special rules that apply to groups representing |
380 | * a single nexthop. | |
381 | */ | |
382 | if (nh && (nh->next == NULL)) { | |
383 | switch (nh->type) { | |
9b4ab909 MS |
384 | case NEXTHOP_TYPE_IFINDEX: |
385 | case NEXTHOP_TYPE_BLACKHOLE: | |
377e29f7 MS |
386 | /* |
387 | * This switch case handles setting the afi different | |
388 | * for ipv4/v6 routes. Ifindex/blackhole nexthop | |
389 | * objects cannot be ambiguous, they must be Address | |
390 | * Family specific. If we get here, we will either use | |
391 | * the AF of the route, or the one we got passed from | |
392 | * here from the kernel. | |
393 | */ | |
394 | nhe->afi = afi; | |
395 | break; | |
9b4ab909 MS |
396 | case NEXTHOP_TYPE_IPV4_IFINDEX: |
397 | case NEXTHOP_TYPE_IPV4: | |
377e29f7 MS |
398 | nhe->afi = AFI_IP; |
399 | break; | |
9b4ab909 MS |
400 | case NEXTHOP_TYPE_IPV6_IFINDEX: |
401 | case NEXTHOP_TYPE_IPV6: | |
377e29f7 MS |
402 | nhe->afi = AFI_IP6; |
403 | break; | |
404 | } | |
7b683a96 | 405 | } |
38e40db1 SW |
406 | } |
407 | ||
0eb97b86 | 408 | struct nhg_hash_entry *zebra_nhg_alloc(void) |
38e40db1 SW |
409 | { |
410 | struct nhg_hash_entry *nhe; | |
411 | ||
412 | nhe = XCALLOC(MTYPE_NHG, sizeof(struct nhg_hash_entry)); | |
413 | ||
0eb97b86 MS |
414 | return nhe; |
415 | } | |
416 | ||
f727646a MS |
417 | /* |
418 | * Allocate new nhe and make shallow copy of 'orig'; no | |
419 | * recursive info is copied. | |
420 | */ | |
421 | struct nhg_hash_entry *zebra_nhe_copy(const struct nhg_hash_entry *orig, | |
422 | uint32_t id) | |
0eb97b86 MS |
423 | { |
424 | struct nhg_hash_entry *nhe; | |
425 | ||
426 | nhe = zebra_nhg_alloc(); | |
427 | ||
38e40db1 SW |
428 | nhe->id = id; |
429 | ||
1d48702e | 430 | nexthop_group_copy(&(nhe->nhg), &(orig->nhg)); |
38e40db1 | 431 | |
1d48702e MS |
432 | nhe->vrf_id = orig->vrf_id; |
433 | nhe->afi = orig->afi; | |
434 | nhe->type = orig->type ? orig->type : ZEBRA_ROUTE_NHG; | |
38e40db1 SW |
435 | nhe->refcnt = 0; |
436 | nhe->dplane_ref = zebra_router_get_next_sequence(); | |
437 | ||
1d48702e MS |
438 | /* Copy backup info also, if present */ |
439 | if (orig->backup_info) | |
440 | nhe->backup_info = nhg_backup_copy(orig->backup_info); | |
441 | ||
38e40db1 SW |
442 | return nhe; |
443 | } | |
444 | ||
445 | /* Allocation via hash handler */ | |
446 | static void *zebra_nhg_hash_alloc(void *arg) | |
447 | { | |
448 | struct nhg_hash_entry *nhe = NULL; | |
449 | struct nhg_hash_entry *copy = arg; | |
7b683a96 | 450 | |
f727646a | 451 | nhe = zebra_nhe_copy(copy, copy->id); |
9ef49038 SW |
452 | |
453 | /* Mark duplicate nexthops in a group at creation time. */ | |
c415d895 | 454 | nexthop_group_mark_duplicates(&(nhe->nhg)); |
9ef49038 | 455 | |
377e29f7 | 456 | /* Add the ifp now if it's not a group or recursive and has ifindex */ |
84591282 | 457 | if (nhe->nhg.nexthop && nhe->nhg.nexthop->ifindex) { |
377e29f7 MS |
458 | struct interface *ifp = NULL; |
459 | ||
460 | ifp = if_lookup_by_index(nhe->nhg.nexthop->ifindex, | |
461 | nhe->nhg.nexthop->vrf_id); | |
462 | if (ifp) | |
463 | zebra_nhg_set_if(nhe, ifp); | |
30672034 SW |
464 | else { |
465 | if (IS_ZEBRA_DEBUG_NHG) | |
466 | zlog_debug( | |
467 | "Failed to lookup an interface with ifindex=%d in vrf=%u for NHE id=%u", | |
468 | nhe->nhg.nexthop->ifindex, | |
469 | nhe->nhg.nexthop->vrf_id, nhe->id); | |
470 | } | |
377e29f7 MS |
471 | } |
472 | ||
4e49c8b8 DS |
473 | return nhe; |
474 | } | |
475 | ||
4e49c8b8 DS |
476 | uint32_t zebra_nhg_hash_key(const void *arg) |
477 | { | |
478 | const struct nhg_hash_entry *nhe = arg; | |
0885b1e3 SW |
479 | uint32_t key = 0x5a351234; |
480 | uint32_t primary = 0; | |
481 | uint32_t backup = 0; | |
482 | ||
483 | primary = nexthop_group_hash(&(nhe->nhg)); | |
484 | if (nhe->backup_info) | |
485 | backup = nexthop_group_hash(&(nhe->backup_info->nhe->nhg)); | |
d9f5b2f5 | 486 | |
0885b1e3 SW |
487 | key = jhash_3words(primary, backup, nhe->type, key); |
488 | ||
489 | key = jhash_2words(nhe->vrf_id, nhe->afi, key); | |
d9f5b2f5 | 490 | |
d9f5b2f5 | 491 | return key; |
4e49c8b8 DS |
492 | } |
493 | ||
a95b8020 SW |
494 | uint32_t zebra_nhg_id_key(const void *arg) |
495 | { | |
496 | const struct nhg_hash_entry *nhe = arg; | |
497 | ||
498 | return nhe->id; | |
499 | } | |
500 | ||
1d48702e MS |
501 | /* Helper with common nhg/nhe nexthop comparison logic */ |
502 | static bool nhg_compare_nexthops(const struct nexthop *nh1, | |
503 | const struct nexthop *nh2) | |
504 | { | |
f924db49 | 505 | assert(nh1 != NULL && nh2 != NULL); |
1d48702e MS |
506 | |
507 | /* | |
508 | * We have to check the active flag of each individual one, | |
509 | * not just the overall active_num. This solves the special case | |
510 | * issue of a route with a nexthop group with one nexthop | |
511 | * resolving to itself and thus marking it inactive. If we | |
512 | * have two different routes each wanting to mark a different | |
513 | * nexthop inactive, they need to hash to two different groups. | |
514 | * | |
515 | * If we just hashed on num_active, they would hash the same | |
516 | * which is incorrect. | |
517 | * | |
518 | * ex) | |
519 | * 1.1.1.0/24 | |
520 | * -> 1.1.1.1 dummy1 (inactive) | |
521 | * -> 1.1.2.1 dummy2 | |
522 | * | |
523 | * 1.1.2.0/24 | |
524 | * -> 1.1.1.1 dummy1 | |
525 | * -> 1.1.2.1 dummy2 (inactive) | |
526 | * | |
527 | * Without checking each individual one, they would hash to | |
528 | * the same group and both have 1.1.1.1 dummy1 marked inactive. | |
529 | * | |
530 | */ | |
531 | if (CHECK_FLAG(nh1->flags, NEXTHOP_FLAG_ACTIVE) | |
532 | != CHECK_FLAG(nh2->flags, NEXTHOP_FLAG_ACTIVE)) | |
533 | return false; | |
534 | ||
535 | if (!nexthop_same(nh1, nh2)) | |
536 | return false; | |
537 | ||
538 | return true; | |
539 | } | |
540 | ||
4e49c8b8 DS |
541 | bool zebra_nhg_hash_equal(const void *arg1, const void *arg2) |
542 | { | |
543 | const struct nhg_hash_entry *nhe1 = arg1; | |
544 | const struct nhg_hash_entry *nhe2 = arg2; | |
148813c2 SW |
545 | struct nexthop *nexthop1; |
546 | struct nexthop *nexthop2; | |
4e49c8b8 | 547 | |
98cda54a SW |
548 | /* No matter what if they equal IDs, assume equal */ |
549 | if (nhe1->id && nhe2->id && (nhe1->id == nhe2->id)) | |
550 | return true; | |
551 | ||
0885b1e3 SW |
552 | if (nhe1->type != nhe2->type) |
553 | return false; | |
554 | ||
4e49c8b8 DS |
555 | if (nhe1->vrf_id != nhe2->vrf_id) |
556 | return false; | |
557 | ||
77b76fc9 SW |
558 | if (nhe1->afi != nhe2->afi) |
559 | return false; | |
560 | ||
1d48702e | 561 | /* Nexthops should be in-order, so we simply compare them in-place */ |
c415d895 | 562 | for (nexthop1 = nhe1->nhg.nexthop, nexthop2 = nhe2->nhg.nexthop; |
f924db49 | 563 | nexthop1 && nexthop2; |
148813c2 | 564 | nexthop1 = nexthop1->next, nexthop2 = nexthop2->next) { |
148813c2 | 565 | |
1d48702e | 566 | if (!nhg_compare_nexthops(nexthop1, nexthop2)) |
148813c2 | 567 | return false; |
1d48702e | 568 | } |
148813c2 | 569 | |
f924db49 MS |
570 | /* Check for unequal list lengths */ |
571 | if (nexthop1 || nexthop2) | |
572 | return false; | |
573 | ||
1d48702e MS |
574 | /* If there's no backup info, comparison is done. */ |
575 | if ((nhe1->backup_info == NULL) && (nhe2->backup_info == NULL)) | |
576 | return true; | |
577 | ||
578 | /* Compare backup info also - test the easy things first */ | |
579 | if (nhe1->backup_info && (nhe2->backup_info == NULL)) | |
580 | return false; | |
581 | if (nhe2->backup_info && (nhe1->backup_info == NULL)) | |
582 | return false; | |
583 | ||
584 | /* Compare number of backups before actually comparing any */ | |
585 | for (nexthop1 = nhe1->backup_info->nhe->nhg.nexthop, | |
586 | nexthop2 = nhe2->backup_info->nhe->nhg.nexthop; | |
587 | nexthop1 && nexthop2; | |
588 | nexthop1 = nexthop1->next, nexthop2 = nexthop2->next) { | |
589 | ; | |
590 | } | |
591 | ||
592 | /* Did we find the end of one list before the other? */ | |
593 | if (nexthop1 || nexthop2) | |
594 | return false; | |
595 | ||
596 | /* Have to compare the backup nexthops */ | |
597 | for (nexthop1 = nhe1->backup_info->nhe->nhg.nexthop, | |
598 | nexthop2 = nhe2->backup_info->nhe->nhg.nexthop; | |
f924db49 | 599 | nexthop1 && nexthop2; |
1d48702e | 600 | nexthop1 = nexthop1->next, nexthop2 = nexthop2->next) { |
148813c2 | 601 | |
1d48702e | 602 | if (!nhg_compare_nexthops(nexthop1, nexthop2)) |
148813c2 SW |
603 | return false; |
604 | } | |
e4ac313b | 605 | |
4e49c8b8 DS |
606 | return true; |
607 | } | |
608 | ||
d9f5b2f5 | 609 | bool zebra_nhg_hash_id_equal(const void *arg1, const void *arg2) |
4e49c8b8 | 610 | { |
d9f5b2f5 SW |
611 | const struct nhg_hash_entry *nhe1 = arg1; |
612 | const struct nhg_hash_entry *nhe2 = arg2; | |
4e49c8b8 | 613 | |
d9f5b2f5 SW |
614 | return nhe1->id == nhe2->id; |
615 | } | |
4e49c8b8 | 616 | |
1b366e63 SW |
617 | static int zebra_nhg_process_grp(struct nexthop_group *nhg, |
618 | struct nhg_connected_tree_head *depends, | |
619 | struct nh_grp *grp, uint8_t count) | |
e22e8001 | 620 | { |
37c6708b | 621 | nhg_connected_tree_init(depends); |
e22e8001 SW |
622 | |
623 | for (int i = 0; i < count; i++) { | |
624 | struct nhg_hash_entry *depend = NULL; | |
625 | /* We do not care about nexthop_grp.weight at | |
626 | * this time. But we should figure out | |
627 | * how to adapt this to our code in | |
628 | * the future. | |
629 | */ | |
38e40db1 | 630 | depend = depends_find_id_add(depends, grp[i].id); |
e22e8001 | 631 | |
38e40db1 | 632 | if (!depend) { |
e22e8001 SW |
633 | flog_err( |
634 | EC_ZEBRA_NHG_SYNC, | |
635 | "Received Nexthop Group from the kernel with a dependent Nexthop ID (%u) which we do not have in our table", | |
636 | grp[i].id); | |
1b366e63 | 637 | return -1; |
e22e8001 | 638 | } |
38e40db1 SW |
639 | |
640 | /* | |
641 | * If this is a nexthop with its own group | |
642 | * dependencies, add them as well. Not sure its | |
643 | * even possible to have a group within a group | |
644 | * in the kernel. | |
645 | */ | |
646 | ||
c415d895 | 647 | copy_nexthops(&nhg->nexthop, depend->nhg.nexthop, NULL); |
e22e8001 | 648 | } |
1b366e63 SW |
649 | |
650 | return 0; | |
e22e8001 SW |
651 | } |
652 | ||
6384cbcb | 653 | static void handle_recursive_depend(struct nhg_connected_tree_head *nhg_depends, |
0885b1e3 | 654 | struct nexthop *nh, afi_t afi, int type) |
6384cbcb SW |
655 | { |
656 | struct nhg_hash_entry *depend = NULL; | |
657 | struct nexthop_group resolved_ng = {}; | |
658 | ||
1d049aba | 659 | resolved_ng.nexthop = nh; |
6384cbcb | 660 | |
377e29f7 MS |
661 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
662 | zlog_debug("%s: head %p, nh %pNHv", | |
663 | __func__, nhg_depends, nh); | |
664 | ||
0885b1e3 | 665 | depend = zebra_nhg_rib_find(0, &resolved_ng, afi, type); |
a7e1b02d | 666 | |
377e29f7 MS |
667 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
668 | zlog_debug("%s: nh %pNHv => %p (%u)", | |
669 | __func__, nh, depend, | |
670 | depend ? depend->id : 0); | |
671 | ||
a7e1b02d SW |
672 | if (depend) |
673 | depends_add(nhg_depends, depend); | |
6384cbcb | 674 | } |
e22e8001 | 675 | |
377e29f7 MS |
676 | /* |
677 | * Lookup an nhe in the global hash, using data from another nhe. If 'lookup' | |
678 | * has an id value, that's used. Create a new global/shared nhe if not found. | |
679 | */ | |
680 | static bool zebra_nhe_find(struct nhg_hash_entry **nhe, /* return value */ | |
681 | struct nhg_hash_entry *lookup, | |
682 | struct nhg_connected_tree_head *nhg_depends, | |
5588801e | 683 | afi_t afi, bool from_dplane) |
377e29f7 MS |
684 | { |
685 | bool created = false; | |
686 | bool recursive = false; | |
0328a5bd | 687 | struct nhg_hash_entry *newnhe, *backup_nhe; |
377e29f7 MS |
688 | struct nexthop *nh = NULL; |
689 | ||
690 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
5588801e SW |
691 | zlog_debug( |
692 | "%s: id %u, lookup %p, vrf %d, type %d, depends %p%s", | |
693 | __func__, lookup->id, lookup, lookup->vrf_id, | |
694 | lookup->type, nhg_depends, | |
695 | (from_dplane ? " (from dplane)" : "")); | |
377e29f7 MS |
696 | |
697 | if (lookup->id) | |
698 | (*nhe) = zebra_nhg_lookup_id(lookup->id); | |
699 | else | |
700 | (*nhe) = hash_lookup(zrouter.nhgs, lookup); | |
701 | ||
702 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
703 | zlog_debug("%s: lookup => %p (%u)", | |
704 | __func__, (*nhe), | |
705 | (*nhe) ? (*nhe)->id : 0); | |
706 | ||
707 | /* If we found an existing object, we're done */ | |
708 | if (*nhe) | |
709 | goto done; | |
710 | ||
711 | /* We're going to create/insert a new nhe: | |
712 | * assign the next global id value if necessary. | |
713 | */ | |
714 | if (lookup->id == 0) | |
ac5d1091 | 715 | lookup->id = nhg_get_next_id(); |
0885b1e3 | 716 | |
475852b2 | 717 | if (!from_dplane && lookup->id < ZEBRA_NHG_PROTO_LOWER) { |
0885b1e3 SW |
718 | /* |
719 | * This is a zebra hashed/owned NHG. | |
720 | * | |
721 | * It goes in HASH and ID table. | |
722 | */ | |
723 | newnhe = hash_get(zrouter.nhgs, lookup, zebra_nhg_hash_alloc); | |
724 | zebra_nhg_insert_id(newnhe); | |
725 | } else { | |
726 | /* | |
475852b2 SW |
727 | * This is upperproto owned NHG or one we read in from dataplane |
728 | * and should not be hashed to. | |
0885b1e3 SW |
729 | * |
730 | * It goes in ID table. | |
731 | */ | |
732 | newnhe = | |
733 | hash_get(zrouter.nhgs_id, lookup, zebra_nhg_hash_alloc); | |
734 | } | |
735 | ||
377e29f7 MS |
736 | created = true; |
737 | ||
738 | /* Mail back the new object */ | |
739 | *nhe = newnhe; | |
740 | ||
741 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
742 | zlog_debug("%s: => created %p (%u)", __func__, newnhe, | |
743 | newnhe->id); | |
744 | ||
745 | /* Only hash/lookup the depends if the first lookup | |
746 | * fails to find something. This should hopefully save a | |
747 | * lot of cycles for larger ecmp sizes. | |
748 | */ | |
749 | if (nhg_depends) { | |
750 | /* If you don't want to hash on each nexthop in the | |
751 | * nexthop group struct you can pass the depends | |
752 | * directly. Kernel-side we do this since it just looks | |
753 | * them up via IDs. | |
754 | */ | |
755 | zebra_nhg_connect_depends(newnhe, nhg_depends); | |
756 | goto done; | |
757 | } | |
758 | ||
759 | /* Prepare dependency relationships if this is not a | |
760 | * singleton nexthop. There are two cases: a single | |
761 | * recursive nexthop, where we need a relationship to the | |
762 | * resolving nexthop; or a group of nexthops, where we need | |
763 | * relationships with the corresponding singletons. | |
764 | */ | |
84591282 | 765 | zebra_nhg_depends_init(newnhe); |
377e29f7 MS |
766 | |
767 | nh = newnhe->nhg.nexthop; | |
768 | ||
769 | if (CHECK_FLAG(nh->flags, NEXTHOP_FLAG_ACTIVE)) | |
770 | SET_FLAG(newnhe->flags, NEXTHOP_GROUP_VALID); | |
771 | ||
54c89c93 | 772 | if (nh->next == NULL && newnhe->id < ZEBRA_NHG_PROTO_LOWER) { |
377e29f7 MS |
773 | if (CHECK_FLAG(nh->flags, NEXTHOP_FLAG_RECURSIVE)) { |
774 | /* Single recursive nexthop */ | |
775 | handle_recursive_depend(&newnhe->nhg_depends, | |
0885b1e3 SW |
776 | nh->resolved, afi, |
777 | newnhe->type); | |
377e29f7 MS |
778 | recursive = true; |
779 | } | |
780 | } else { | |
dd1e105f | 781 | /* Proto-owned are groups by default */ |
377e29f7 MS |
782 | /* List of nexthops */ |
783 | for (nh = newnhe->nhg.nexthop; nh; nh = nh->next) { | |
784 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
785 | zlog_debug("%s: depends NH %pNHv %s", | |
786 | __func__, nh, | |
787 | CHECK_FLAG(nh->flags, | |
788 | NEXTHOP_FLAG_RECURSIVE) ? | |
789 | "(R)" : ""); | |
790 | ||
0885b1e3 | 791 | depends_find_add(&newnhe->nhg_depends, nh, afi, |
5588801e | 792 | newnhe->type, from_dplane); |
377e29f7 MS |
793 | } |
794 | } | |
795 | ||
0328a5bd | 796 | if (recursive) |
84591282 SW |
797 | SET_FLAG(newnhe->flags, NEXTHOP_GROUP_RECURSIVE); |
798 | ||
799 | /* Attach dependent backpointers to singletons */ | |
800 | zebra_nhg_connect_depends(newnhe, &newnhe->nhg_depends); | |
801 | ||
802 | /** | |
803 | * Backup Nexthops | |
804 | */ | |
0328a5bd | 805 | |
377e29f7 MS |
806 | if (zebra_nhg_get_backup_nhg(newnhe) == NULL || |
807 | zebra_nhg_get_backup_nhg(newnhe)->nexthop == NULL) | |
0328a5bd MS |
808 | goto done; |
809 | ||
810 | /* If there are backup nexthops, add them to the backup | |
811 | * depends tree. The rules here are a little different. | |
812 | */ | |
813 | recursive = false; | |
814 | backup_nhe = newnhe->backup_info->nhe; | |
377e29f7 | 815 | |
0328a5bd | 816 | nh = backup_nhe->nhg.nexthop; |
377e29f7 MS |
817 | |
818 | /* Singleton recursive NH */ | |
819 | if (nh->next == NULL && | |
820 | CHECK_FLAG(nh->flags, NEXTHOP_FLAG_RECURSIVE)) { | |
821 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
822 | zlog_debug("%s: backup depend NH %pNHv (R)", | |
823 | __func__, nh); | |
824 | ||
825 | /* Single recursive nexthop */ | |
0885b1e3 SW |
826 | handle_recursive_depend(&backup_nhe->nhg_depends, nh->resolved, |
827 | afi, backup_nhe->type); | |
377e29f7 MS |
828 | recursive = true; |
829 | } else { | |
830 | /* One or more backup NHs */ | |
831 | for (; nh; nh = nh->next) { | |
832 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
833 | zlog_debug("%s: backup depend NH %pNHv %s", | |
834 | __func__, nh, | |
835 | CHECK_FLAG(nh->flags, | |
836 | NEXTHOP_FLAG_RECURSIVE) ? | |
837 | "(R)" : ""); | |
838 | ||
0885b1e3 | 839 | depends_find_add(&backup_nhe->nhg_depends, nh, afi, |
5588801e | 840 | backup_nhe->type, from_dplane); |
377e29f7 MS |
841 | } |
842 | } | |
843 | ||
377e29f7 | 844 | if (recursive) |
0328a5bd | 845 | SET_FLAG(backup_nhe->flags, NEXTHOP_GROUP_RECURSIVE); |
377e29f7 MS |
846 | |
847 | done: | |
45691de9 SW |
848 | /* Reset time since last update */ |
849 | (*nhe)->uptime = monotime(NULL); | |
377e29f7 MS |
850 | |
851 | return created; | |
852 | } | |
853 | ||
854 | /* | |
855 | * Lookup or create an nhe, based on an nhg or an nhe id. | |
856 | */ | |
4505578b SW |
857 | static bool zebra_nhg_find(struct nhg_hash_entry **nhe, uint32_t id, |
858 | struct nexthop_group *nhg, | |
37c6708b | 859 | struct nhg_connected_tree_head *nhg_depends, |
5588801e SW |
860 | vrf_id_t vrf_id, afi_t afi, int type, |
861 | bool from_dplane) | |
a95b8020 | 862 | { |
0c8215cb | 863 | struct nhg_hash_entry lookup = {}; |
4505578b SW |
864 | bool created = false; |
865 | ||
1d48702e MS |
866 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
867 | zlog_debug("%s: id %u, nhg %p, vrf %d, type %d, depends %p", | |
868 | __func__, id, nhg, vrf_id, type, | |
869 | nhg_depends); | |
870 | ||
377e29f7 MS |
871 | /* Use a temporary nhe and call into the superset/common code */ |
872 | lookup.id = id; | |
9a1588c4 | 873 | lookup.type = type ? type : ZEBRA_ROUTE_NHG; |
c415d895 | 874 | lookup.nhg = *nhg; |
e22e8001 | 875 | |
88cafda7 | 876 | lookup.vrf_id = vrf_id; |
475852b2 | 877 | if (nhg_depends || lookup.nhg.nexthop->next) { |
6384cbcb SW |
878 | /* Groups can have all vrfs and AF's in them */ |
879 | lookup.afi = AFI_UNSPEC; | |
6384cbcb | 880 | } else { |
c415d895 | 881 | switch (lookup.nhg.nexthop->type) { |
4d21c7c0 SW |
882 | case (NEXTHOP_TYPE_IFINDEX): |
883 | case (NEXTHOP_TYPE_BLACKHOLE): | |
884 | /* | |
885 | * This switch case handles setting the afi different | |
886 | * for ipv4/v6 routes. Ifindex/blackhole nexthop | |
887 | * objects cannot be ambiguous, they must be Address | |
888 | * Family specific. If we get here, we will either use | |
889 | * the AF of the route, or the one we got passed from | |
890 | * here from the kernel. | |
891 | */ | |
892 | lookup.afi = afi; | |
893 | break; | |
894 | case (NEXTHOP_TYPE_IPV4_IFINDEX): | |
895 | case (NEXTHOP_TYPE_IPV4): | |
896 | lookup.afi = AFI_IP; | |
897 | break; | |
898 | case (NEXTHOP_TYPE_IPV6_IFINDEX): | |
899 | case (NEXTHOP_TYPE_IPV6): | |
900 | lookup.afi = AFI_IP6; | |
901 | break; | |
902 | } | |
6384cbcb | 903 | } |
a95b8020 | 904 | |
5588801e | 905 | created = zebra_nhe_find(nhe, &lookup, nhg_depends, afi, from_dplane); |
d9f5b2f5 | 906 | |
4505578b | 907 | return created; |
a95b8020 SW |
908 | } |
909 | ||
e22e8001 | 910 | /* Find/create a single nexthop */ |
5588801e SW |
911 | static struct nhg_hash_entry *zebra_nhg_find_nexthop(uint32_t id, |
912 | struct nexthop *nh, | |
913 | afi_t afi, int type, | |
914 | bool from_dplane) | |
3057df51 | 915 | { |
6384cbcb | 916 | struct nhg_hash_entry *nhe = NULL; |
e22e8001 | 917 | struct nexthop_group nhg = {}; |
88cafda7 | 918 | vrf_id_t vrf_id = !vrf_is_backend_netns() ? VRF_DEFAULT : nh->vrf_id; |
e22e8001 | 919 | |
0eb97b86 | 920 | nexthop_group_add_sorted(&nhg, nh); |
e22e8001 | 921 | |
5588801e | 922 | zebra_nhg_find(&nhe, id, &nhg, NULL, vrf_id, afi, type, from_dplane); |
8a507796 | 923 | |
377e29f7 MS |
924 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
925 | zlog_debug("%s: nh %pNHv => %p (%u)", | |
926 | __func__, nh, nhe, nhe ? nhe->id : 0); | |
927 | ||
6384cbcb | 928 | return nhe; |
e22e8001 SW |
929 | } |
930 | ||
10200d40 SW |
931 | static uint32_t nhg_ctx_get_id(const struct nhg_ctx *ctx) |
932 | { | |
933 | return ctx->id; | |
934 | } | |
935 | ||
1b366e63 | 936 | static void nhg_ctx_set_status(struct nhg_ctx *ctx, enum nhg_ctx_status status) |
e22e8001 SW |
937 | { |
938 | ctx->status = status; | |
939 | } | |
940 | ||
1b366e63 | 941 | static enum nhg_ctx_status nhg_ctx_get_status(const struct nhg_ctx *ctx) |
e22e8001 SW |
942 | { |
943 | return ctx->status; | |
944 | } | |
945 | ||
946 | static void nhg_ctx_set_op(struct nhg_ctx *ctx, enum nhg_ctx_op_e op) | |
947 | { | |
948 | ctx->op = op; | |
949 | } | |
950 | ||
951 | static enum nhg_ctx_op_e nhg_ctx_get_op(const struct nhg_ctx *ctx) | |
952 | { | |
953 | return ctx->op; | |
954 | } | |
955 | ||
10200d40 SW |
956 | static vrf_id_t nhg_ctx_get_vrf_id(const struct nhg_ctx *ctx) |
957 | { | |
958 | return ctx->vrf_id; | |
959 | } | |
960 | ||
961 | static int nhg_ctx_get_type(const struct nhg_ctx *ctx) | |
962 | { | |
963 | return ctx->type; | |
964 | } | |
965 | ||
966 | static int nhg_ctx_get_afi(const struct nhg_ctx *ctx) | |
967 | { | |
968 | return ctx->afi; | |
969 | } | |
970 | ||
971 | static struct nexthop *nhg_ctx_get_nh(struct nhg_ctx *ctx) | |
972 | { | |
973 | return &ctx->u.nh; | |
974 | } | |
975 | ||
976 | static uint8_t nhg_ctx_get_count(const struct nhg_ctx *ctx) | |
977 | { | |
978 | return ctx->count; | |
979 | } | |
980 | ||
981 | static struct nh_grp *nhg_ctx_get_grp(struct nhg_ctx *ctx) | |
982 | { | |
983 | return ctx->u.grp; | |
984 | } | |
985 | ||
99e7ab12 | 986 | static struct nhg_ctx *nhg_ctx_new(void) |
7c6d5f25 | 987 | { |
99e7ab12 | 988 | struct nhg_ctx *new; |
7c6d5f25 SW |
989 | |
990 | new = XCALLOC(MTYPE_NHG_CTX, sizeof(struct nhg_ctx)); | |
991 | ||
992 | return new; | |
993 | } | |
994 | ||
04bec7b2 | 995 | void nhg_ctx_free(struct nhg_ctx **ctx) |
7c6d5f25 SW |
996 | { |
997 | struct nexthop *nh; | |
998 | ||
999 | if (ctx == NULL) | |
1000 | return; | |
1001 | ||
1002 | assert((*ctx) != NULL); | |
1003 | ||
1004 | if (nhg_ctx_get_count(*ctx)) | |
1005 | goto done; | |
1006 | ||
1007 | nh = nhg_ctx_get_nh(*ctx); | |
1008 | ||
1009 | nexthop_del_labels(nh); | |
eab0f8f0 HS |
1010 | nexthop_del_srv6_seg6local(nh); |
1011 | nexthop_del_srv6_seg6(nh); | |
7c6d5f25 SW |
1012 | |
1013 | done: | |
1014 | XFREE(MTYPE_NHG_CTX, *ctx); | |
7c6d5f25 SW |
1015 | } |
1016 | ||
81505946 SW |
1017 | static struct nhg_ctx *nhg_ctx_init(uint32_t id, struct nexthop *nh, |
1018 | struct nh_grp *grp, vrf_id_t vrf_id, | |
1019 | afi_t afi, int type, uint8_t count) | |
1020 | { | |
1021 | struct nhg_ctx *ctx = NULL; | |
1022 | ||
1023 | ctx = nhg_ctx_new(); | |
1024 | ||
1025 | ctx->id = id; | |
1026 | ctx->vrf_id = vrf_id; | |
1027 | ctx->afi = afi; | |
1028 | ctx->type = type; | |
1029 | ctx->count = count; | |
1030 | ||
1031 | if (count) | |
1032 | /* Copy over the array */ | |
1033 | memcpy(&ctx->u.grp, grp, count * sizeof(struct nh_grp)); | |
1034 | else if (nh) | |
1035 | ctx->u.nh = *nh; | |
1036 | ||
1037 | return ctx; | |
1038 | } | |
1039 | ||
80286aa5 SW |
1040 | static void zebra_nhg_set_valid(struct nhg_hash_entry *nhe) |
1041 | { | |
1042 | struct nhg_connected *rb_node_dep; | |
1043 | ||
1044 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_VALID); | |
1045 | ||
1046 | frr_each(nhg_connected_tree, &nhe->nhg_dependents, rb_node_dep) | |
1047 | zebra_nhg_set_valid(rb_node_dep->nhe); | |
1048 | } | |
1049 | ||
1050 | static void zebra_nhg_set_invalid(struct nhg_hash_entry *nhe) | |
1051 | { | |
1052 | struct nhg_connected *rb_node_dep; | |
1053 | ||
1054 | UNSET_FLAG(nhe->flags, NEXTHOP_GROUP_VALID); | |
1055 | ||
1056 | /* Update validity of nexthops depending on it */ | |
1057 | frr_each(nhg_connected_tree, &nhe->nhg_dependents, rb_node_dep) | |
1058 | zebra_nhg_check_valid(rb_node_dep->nhe); | |
1059 | } | |
1060 | ||
1061 | void zebra_nhg_check_valid(struct nhg_hash_entry *nhe) | |
1062 | { | |
1063 | struct nhg_connected *rb_node_dep = NULL; | |
1064 | bool valid = false; | |
1065 | ||
1066 | /* If anthing else in the group is valid, the group is valid */ | |
1067 | frr_each(nhg_connected_tree, &nhe->nhg_depends, rb_node_dep) { | |
1068 | if (CHECK_FLAG(rb_node_dep->nhe->flags, NEXTHOP_GROUP_VALID)) { | |
1069 | valid = true; | |
1070 | goto done; | |
1071 | } | |
1072 | } | |
1073 | ||
1074 | done: | |
1075 | if (valid) | |
1076 | zebra_nhg_set_valid(nhe); | |
1077 | else | |
1078 | zebra_nhg_set_invalid(nhe); | |
1079 | } | |
1080 | ||
dd1e105f | 1081 | static void zebra_nhg_release_all_deps(struct nhg_hash_entry *nhe) |
9a1588c4 SW |
1082 | { |
1083 | /* Remove it from any lists it may be on */ | |
1084 | zebra_nhg_depends_release(nhe); | |
1085 | zebra_nhg_dependents_release(nhe); | |
1086 | if (nhe->ifp) | |
1087 | if_nhg_dependents_del(nhe->ifp, nhe); | |
dd1e105f SW |
1088 | } |
1089 | ||
1090 | static void zebra_nhg_release(struct nhg_hash_entry *nhe) | |
1091 | { | |
1092 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
1093 | zlog_debug("%s: nhe %p (%u)", __func__, nhe, nhe->id); | |
1094 | ||
1095 | zebra_nhg_release_all_deps(nhe); | |
9a1588c4 SW |
1096 | |
1097 | /* | |
0885b1e3 | 1098 | * If its not zebra owned, we didn't store it here and have to be |
9a1588c4 SW |
1099 | * sure we don't clear one thats actually being used. |
1100 | */ | |
3bccc0f5 | 1101 | if (nhe->id < ZEBRA_NHG_PROTO_LOWER) |
9a1588c4 | 1102 | hash_release(zrouter.nhgs, nhe); |
9a1588c4 | 1103 | |
9a1588c4 SW |
1104 | hash_release(zrouter.nhgs_id, nhe); |
1105 | } | |
1106 | ||
9a1588c4 SW |
1107 | static void zebra_nhg_handle_uninstall(struct nhg_hash_entry *nhe) |
1108 | { | |
177e711d | 1109 | zebra_nhg_release(nhe); |
9a1588c4 SW |
1110 | zebra_nhg_free(nhe); |
1111 | } | |
1112 | ||
80286aa5 SW |
1113 | static void zebra_nhg_handle_install(struct nhg_hash_entry *nhe) |
1114 | { | |
1115 | /* Update validity of groups depending on it */ | |
1116 | struct nhg_connected *rb_node_dep; | |
1117 | ||
1118 | frr_each_safe(nhg_connected_tree, &nhe->nhg_dependents, rb_node_dep) | |
1119 | zebra_nhg_set_valid(rb_node_dep->nhe); | |
1120 | } | |
1121 | ||
9a1588c4 SW |
1122 | /* |
1123 | * The kernel/other program has changed the state of a nexthop object we are | |
1124 | * using. | |
1125 | */ | |
1126 | static void zebra_nhg_handle_kernel_state_change(struct nhg_hash_entry *nhe, | |
1127 | bool is_delete) | |
1128 | { | |
1129 | if (nhe->refcnt) { | |
1130 | flog_err( | |
1131 | EC_ZEBRA_NHG_SYNC, | |
1132 | "Kernel %s a nexthop group with ID (%u) that we are still using for a route, sending it back down", | |
1133 | (is_delete ? "deleted" : "updated"), nhe->id); | |
1134 | ||
1135 | UNSET_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED); | |
1136 | zebra_nhg_install_kernel(nhe); | |
177e711d | 1137 | } else |
9a1588c4 | 1138 | zebra_nhg_handle_uninstall(nhe); |
9a1588c4 SW |
1139 | } |
1140 | ||
e22e8001 SW |
1141 | static int nhg_ctx_process_new(struct nhg_ctx *ctx) |
1142 | { | |
1143 | struct nexthop_group *nhg = NULL; | |
37c6708b | 1144 | struct nhg_connected_tree_head nhg_depends = {}; |
9a1588c4 | 1145 | struct nhg_hash_entry *lookup = NULL; |
3057df51 SW |
1146 | struct nhg_hash_entry *nhe = NULL; |
1147 | ||
10200d40 SW |
1148 | uint32_t id = nhg_ctx_get_id(ctx); |
1149 | uint8_t count = nhg_ctx_get_count(ctx); | |
1150 | vrf_id_t vrf_id = nhg_ctx_get_vrf_id(ctx); | |
1151 | int type = nhg_ctx_get_type(ctx); | |
1152 | afi_t afi = nhg_ctx_get_afi(ctx); | |
1153 | ||
1154 | lookup = zebra_nhg_lookup_id(id); | |
9a1588c4 | 1155 | |
377e29f7 MS |
1156 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
1157 | zlog_debug("%s: id %u, count %d, lookup => %p", | |
1158 | __func__, id, count, lookup); | |
1159 | ||
9a1588c4 SW |
1160 | if (lookup) { |
1161 | /* This is already present in our table, hence an update | |
1162 | * that we did not initate. | |
1163 | */ | |
1164 | zebra_nhg_handle_kernel_state_change(lookup, false); | |
1165 | return 0; | |
1166 | } | |
1167 | ||
10200d40 | 1168 | if (nhg_ctx_get_count(ctx)) { |
e22e8001 | 1169 | nhg = nexthop_group_new(); |
1b366e63 SW |
1170 | if (zebra_nhg_process_grp(nhg, &nhg_depends, |
1171 | nhg_ctx_get_grp(ctx), count)) { | |
1172 | depends_decrement_free(&nhg_depends); | |
d3a35138 | 1173 | nexthop_group_delete(&nhg); |
fec211ad | 1174 | return -ENOENT; |
1b366e63 SW |
1175 | } |
1176 | ||
0885b1e3 | 1177 | if (!zebra_nhg_find(&nhe, id, nhg, &nhg_depends, vrf_id, afi, |
5588801e | 1178 | type, true)) |
38e40db1 | 1179 | depends_decrement_free(&nhg_depends); |
4505578b | 1180 | |
e22e8001 | 1181 | /* These got copied over in zebra_nhg_alloc() */ |
d3a35138 | 1182 | nexthop_group_delete(&nhg); |
38e40db1 | 1183 | } else |
5588801e SW |
1184 | nhe = zebra_nhg_find_nexthop(id, nhg_ctx_get_nh(ctx), afi, type, |
1185 | true); | |
e22e8001 | 1186 | |
5b27c09d | 1187 | if (!nhe) { |
e22e8001 SW |
1188 | flog_err( |
1189 | EC_ZEBRA_TABLE_LOOKUP_FAILED, | |
1190 | "Zebra failed to find or create a nexthop hash entry for ID (%u)", | |
10200d40 | 1191 | id); |
e22e8001 SW |
1192 | return -1; |
1193 | } | |
1194 | ||
5b27c09d SW |
1195 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
1196 | zlog_debug("%s: nhe %p (%u) is new", __func__, nhe, nhe->id); | |
1197 | ||
dc65cd99 SW |
1198 | /* |
1199 | * If daemon nhg from the kernel, add a refcnt here to indicate the | |
1200 | * daemon owns it. | |
1201 | */ | |
1202 | if (PROTO_OWNED(nhe)) | |
1203 | zebra_nhg_increment_ref(nhe); | |
1204 | ||
5b27c09d SW |
1205 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_VALID); |
1206 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED); | |
1207 | ||
e22e8001 SW |
1208 | return 0; |
1209 | } | |
1210 | ||
9a1588c4 SW |
1211 | static int nhg_ctx_process_del(struct nhg_ctx *ctx) |
1212 | { | |
1213 | struct nhg_hash_entry *nhe = NULL; | |
10200d40 | 1214 | uint32_t id = nhg_ctx_get_id(ctx); |
9a1588c4 | 1215 | |
10200d40 | 1216 | nhe = zebra_nhg_lookup_id(id); |
9a1588c4 SW |
1217 | |
1218 | if (!nhe) { | |
1219 | flog_warn( | |
1220 | EC_ZEBRA_BAD_NHG_MESSAGE, | |
1221 | "Kernel delete message received for nexthop group ID (%u) that we do not have in our ID table", | |
10200d40 | 1222 | id); |
81505946 | 1223 | return -1; |
9a1588c4 SW |
1224 | } |
1225 | ||
1226 | zebra_nhg_handle_kernel_state_change(nhe, true); | |
1227 | ||
1228 | return 0; | |
1229 | } | |
1230 | ||
7c6d5f25 | 1231 | static void nhg_ctx_fini(struct nhg_ctx **ctx) |
e22e8001 SW |
1232 | { |
1233 | /* | |
1234 | * Just freeing for now, maybe do something more in the future | |
1235 | * based on flag. | |
1236 | */ | |
1237 | ||
7134ba70 | 1238 | nhg_ctx_free(ctx); |
e22e8001 SW |
1239 | } |
1240 | ||
1b366e63 SW |
1241 | static int queue_add(struct nhg_ctx *ctx) |
1242 | { | |
1243 | /* If its queued or already processed do nothing */ | |
1244 | if (nhg_ctx_get_status(ctx) == NHG_CTX_QUEUED) | |
1245 | return 0; | |
1246 | ||
04bec7b2 | 1247 | if (rib_queue_nhg_ctx_add(ctx)) { |
1b366e63 SW |
1248 | nhg_ctx_set_status(ctx, NHG_CTX_FAILURE); |
1249 | return -1; | |
1250 | } | |
1251 | ||
1252 | nhg_ctx_set_status(ctx, NHG_CTX_QUEUED); | |
1253 | ||
1254 | return 0; | |
1255 | } | |
1256 | ||
e22e8001 SW |
1257 | int nhg_ctx_process(struct nhg_ctx *ctx) |
1258 | { | |
1259 | int ret = 0; | |
1260 | ||
1261 | switch (nhg_ctx_get_op(ctx)) { | |
1262 | case NHG_CTX_OP_NEW: | |
1263 | ret = nhg_ctx_process_new(ctx); | |
fec211ad | 1264 | if (nhg_ctx_get_count(ctx) && ret == -ENOENT |
1b366e63 | 1265 | && nhg_ctx_get_status(ctx) != NHG_CTX_REQUEUED) { |
e1292378 SW |
1266 | /** |
1267 | * We have entered a situation where we are | |
1268 | * processing a group from the kernel | |
1269 | * that has a contained nexthop which | |
1270 | * we have not yet processed. | |
1b366e63 | 1271 | * |
e1292378 SW |
1272 | * Re-enqueue this ctx to be handled exactly one |
1273 | * more time (indicated by the flag). | |
1274 | * | |
1275 | * By the time we get back to it, we | |
1276 | * should have processed its depends. | |
1b366e63 SW |
1277 | */ |
1278 | nhg_ctx_set_status(ctx, NHG_CTX_NONE); | |
1279 | if (queue_add(ctx) == 0) { | |
1280 | nhg_ctx_set_status(ctx, NHG_CTX_REQUEUED); | |
1281 | return 0; | |
1282 | } | |
1283 | } | |
e22e8001 SW |
1284 | break; |
1285 | case NHG_CTX_OP_DEL: | |
9a1588c4 | 1286 | ret = nhg_ctx_process_del(ctx); |
e22e8001 SW |
1287 | case NHG_CTX_OP_NONE: |
1288 | break; | |
1289 | } | |
1290 | ||
1291 | nhg_ctx_set_status(ctx, (ret ? NHG_CTX_FAILURE : NHG_CTX_SUCCESS)); | |
1292 | ||
7c6d5f25 | 1293 | nhg_ctx_fini(&ctx); |
e22e8001 SW |
1294 | |
1295 | return ret; | |
1296 | } | |
3057df51 | 1297 | |
e22e8001 SW |
1298 | /* Kernel-side, you either get a single new nexthop or a array of ID's */ |
1299 | int zebra_nhg_kernel_find(uint32_t id, struct nexthop *nh, struct nh_grp *grp, | |
38e40db1 SW |
1300 | uint8_t count, vrf_id_t vrf_id, afi_t afi, int type, |
1301 | int startup) | |
e22e8001 | 1302 | { |
e22e8001 SW |
1303 | struct nhg_ctx *ctx = NULL; |
1304 | ||
377e29f7 MS |
1305 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
1306 | zlog_debug("%s: nh %pNHv, id %u, count %d", | |
1307 | __func__, nh, id, (int)count); | |
1308 | ||
54c89c93 | 1309 | if (id > id_counter && id < ZEBRA_NHG_PROTO_LOWER) |
38e40db1 SW |
1310 | /* Increase our counter so we don't try to create |
1311 | * an ID that already exists | |
1312 | */ | |
1313 | id_counter = id; | |
1314 | ||
81505946 | 1315 | ctx = nhg_ctx_init(id, nh, grp, vrf_id, afi, type, count); |
e22e8001 SW |
1316 | nhg_ctx_set_op(ctx, NHG_CTX_OP_NEW); |
1317 | ||
38e40db1 SW |
1318 | /* Under statup conditions, we need to handle them immediately |
1319 | * like we do for routes. Otherwise, we are going to get a route | |
1320 | * with a nhe_id that we have not handled. | |
1321 | */ | |
1322 | if (startup) | |
1323 | return nhg_ctx_process(ctx); | |
1324 | ||
e22e8001 | 1325 | if (queue_add(ctx)) { |
7c6d5f25 | 1326 | nhg_ctx_fini(&ctx); |
e22e8001 SW |
1327 | return -1; |
1328 | } | |
1329 | ||
1330 | return 0; | |
1331 | } | |
1332 | ||
9a1588c4 | 1333 | /* Kernel-side, received delete message */ |
88cafda7 | 1334 | int zebra_nhg_kernel_del(uint32_t id, vrf_id_t vrf_id) |
9a1588c4 SW |
1335 | { |
1336 | struct nhg_ctx *ctx = NULL; | |
1337 | ||
88cafda7 | 1338 | ctx = nhg_ctx_init(id, NULL, NULL, vrf_id, 0, 0, 0); |
9a1588c4 SW |
1339 | |
1340 | nhg_ctx_set_op(ctx, NHG_CTX_OP_DEL); | |
1341 | ||
1342 | if (queue_add(ctx)) { | |
7c6d5f25 | 1343 | nhg_ctx_fini(&ctx); |
9a1588c4 SW |
1344 | return -1; |
1345 | } | |
1346 | ||
1347 | return 0; | |
1348 | } | |
1349 | ||
5657e7e9 | 1350 | /* Some dependency helper functions */ |
0fff714e | 1351 | static struct nhg_hash_entry *depends_find_recursive(const struct nexthop *nh, |
0885b1e3 | 1352 | afi_t afi, int type) |
98cda54a | 1353 | { |
0fff714e SW |
1354 | struct nhg_hash_entry *nhe; |
1355 | struct nexthop *lookup = NULL; | |
98cda54a | 1356 | |
77bf9504 | 1357 | lookup = nexthop_dup(nh, NULL); |
0fff714e | 1358 | |
5588801e | 1359 | nhe = zebra_nhg_find_nexthop(0, lookup, afi, type, false); |
0fff714e SW |
1360 | |
1361 | nexthops_free(lookup); | |
1362 | ||
1363 | return nhe; | |
1364 | } | |
1365 | ||
1366 | static struct nhg_hash_entry *depends_find_singleton(const struct nexthop *nh, | |
5588801e SW |
1367 | afi_t afi, int type, |
1368 | bool from_dplane) | |
0fff714e SW |
1369 | { |
1370 | struct nhg_hash_entry *nhe; | |
1371 | struct nexthop lookup = {}; | |
606fa9e5 | 1372 | |
cb86eba3 MS |
1373 | /* Capture a snapshot of this single nh; it might be part of a list, |
1374 | * so we need to make a standalone copy. | |
1375 | */ | |
77bf9504 | 1376 | nexthop_copy_no_recurse(&lookup, nh, NULL); |
8a507796 | 1377 | |
5588801e | 1378 | nhe = zebra_nhg_find_nexthop(0, &lookup, afi, type, from_dplane); |
8a507796 | 1379 | |
cb86eba3 MS |
1380 | /* The copy may have allocated labels; free them if necessary. */ |
1381 | nexthop_del_labels(&lookup); | |
eab0f8f0 HS |
1382 | nexthop_del_srv6_seg6local(&lookup); |
1383 | nexthop_del_srv6_seg6(&lookup); | |
4505578b | 1384 | |
377e29f7 MS |
1385 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
1386 | zlog_debug("%s: nh %pNHv => %p (%u)", | |
1387 | __func__, nh, nhe, nhe ? nhe->id : 0); | |
1388 | ||
0fff714e SW |
1389 | return nhe; |
1390 | } | |
1391 | ||
0885b1e3 | 1392 | static struct nhg_hash_entry *depends_find(const struct nexthop *nh, afi_t afi, |
5588801e | 1393 | int type, bool from_dplane) |
0fff714e SW |
1394 | { |
1395 | struct nhg_hash_entry *nhe = NULL; | |
1396 | ||
1397 | if (!nh) | |
1398 | goto done; | |
1399 | ||
1400 | /* We are separating these functions out to increase handling speed | |
1401 | * in the non-recursive case (by not alloc/freeing) | |
1402 | */ | |
bed74d17 | 1403 | if (CHECK_FLAG(nh->flags, NEXTHOP_FLAG_RECURSIVE)) |
0885b1e3 | 1404 | nhe = depends_find_recursive(nh, afi, type); |
bed74d17 | 1405 | else |
5588801e | 1406 | nhe = depends_find_singleton(nh, afi, type, from_dplane); |
377e29f7 | 1407 | |
bed74d17 DS |
1408 | |
1409 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) { | |
1410 | zlog_debug("%s: nh %pNHv %s => %p (%u)", __func__, nh, | |
1411 | CHECK_FLAG(nh->flags, NEXTHOP_FLAG_RECURSIVE) ? "(R)" | |
1412 | : "", | |
377e29f7 | 1413 | nhe, nhe ? nhe->id : 0); |
bed74d17 | 1414 | } |
0fff714e | 1415 | |
606fa9e5 | 1416 | done: |
4505578b | 1417 | return nhe; |
98cda54a SW |
1418 | } |
1419 | ||
37c6708b | 1420 | static void depends_add(struct nhg_connected_tree_head *head, |
5657e7e9 SW |
1421 | struct nhg_hash_entry *depend) |
1422 | { | |
377e29f7 MS |
1423 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
1424 | zlog_debug("%s: head %p nh %pNHv", | |
1425 | __func__, head, depend->nhg.nexthop); | |
1426 | ||
5bf15faa SW |
1427 | /* If NULL is returned, it was successfully added and |
1428 | * needs to have its refcnt incremented. | |
1429 | * | |
1430 | * Else the NHE is already present in the tree and doesn't | |
1431 | * need to increment the refcnt. | |
1432 | */ | |
1433 | if (nhg_connected_tree_add_nhe(head, depend) == NULL) | |
1434 | zebra_nhg_increment_ref(depend); | |
5657e7e9 SW |
1435 | } |
1436 | ||
38e40db1 SW |
1437 | static struct nhg_hash_entry * |
1438 | depends_find_add(struct nhg_connected_tree_head *head, struct nexthop *nh, | |
5588801e | 1439 | afi_t afi, int type, bool from_dplane) |
5657e7e9 SW |
1440 | { |
1441 | struct nhg_hash_entry *depend = NULL; | |
1442 | ||
5588801e | 1443 | depend = depends_find(nh, afi, type, from_dplane); |
1b366e63 | 1444 | |
1d48702e MS |
1445 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
1446 | zlog_debug("%s: nh %pNHv => %p", | |
1447 | __func__, nh, depend); | |
1448 | ||
1b366e63 SW |
1449 | if (depend) |
1450 | depends_add(head, depend); | |
38e40db1 SW |
1451 | |
1452 | return depend; | |
1453 | } | |
1454 | ||
1455 | static struct nhg_hash_entry * | |
1456 | depends_find_id_add(struct nhg_connected_tree_head *head, uint32_t id) | |
1457 | { | |
1458 | struct nhg_hash_entry *depend = NULL; | |
1459 | ||
1460 | depend = zebra_nhg_lookup_id(id); | |
1b366e63 SW |
1461 | |
1462 | if (depend) | |
1463 | depends_add(head, depend); | |
38e40db1 SW |
1464 | |
1465 | return depend; | |
5657e7e9 SW |
1466 | } |
1467 | ||
37c6708b | 1468 | static void depends_decrement_free(struct nhg_connected_tree_head *head) |
5657e7e9 | 1469 | { |
37c6708b SW |
1470 | nhg_connected_tree_decrement_ref(head); |
1471 | nhg_connected_tree_free(head); | |
5657e7e9 SW |
1472 | } |
1473 | ||
377e29f7 | 1474 | /* Find an nhe based on a list of nexthops */ |
0885b1e3 SW |
1475 | struct nhg_hash_entry *zebra_nhg_rib_find(uint32_t id, |
1476 | struct nexthop_group *nhg, | |
1477 | afi_t rt_afi, int type) | |
e22e8001 SW |
1478 | { |
1479 | struct nhg_hash_entry *nhe = NULL; | |
88cafda7 DS |
1480 | vrf_id_t vrf_id; |
1481 | ||
1482 | /* | |
1483 | * CLANG SA is complaining that nexthop may be NULL | |
1484 | * Make it happy but this is ridonc | |
1485 | */ | |
1486 | assert(nhg->nexthop); | |
1487 | vrf_id = !vrf_is_backend_netns() ? VRF_DEFAULT : nhg->nexthop->vrf_id; | |
98cda54a | 1488 | |
5588801e | 1489 | zebra_nhg_find(&nhe, id, nhg, NULL, vrf_id, rt_afi, type, false); |
4505578b | 1490 | |
377e29f7 MS |
1491 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
1492 | zlog_debug("%s: => nhe %p (%u)", | |
1493 | __func__, nhe, nhe ? nhe->id : 0); | |
1494 | ||
1495 | return nhe; | |
1496 | } | |
1497 | ||
1498 | /* Find an nhe based on a route's nhe */ | |
1499 | struct nhg_hash_entry * | |
1500 | zebra_nhg_rib_find_nhe(struct nhg_hash_entry *rt_nhe, afi_t rt_afi) | |
1501 | { | |
1502 | struct nhg_hash_entry *nhe = NULL; | |
1503 | ||
1504 | if (!(rt_nhe && rt_nhe->nhg.nexthop)) { | |
1505 | flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED, | |
1506 | "No nexthop passed to %s", __func__); | |
1507 | return NULL; | |
1508 | } | |
1509 | ||
1510 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
f924db49 | 1511 | zlog_debug("%s: rt_nhe %p (%u)", __func__, rt_nhe, rt_nhe->id); |
377e29f7 | 1512 | |
5588801e | 1513 | zebra_nhe_find(&nhe, rt_nhe, NULL, rt_afi, false); |
377e29f7 MS |
1514 | |
1515 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
1516 | zlog_debug("%s: => nhe %p (%u)", | |
1517 | __func__, nhe, nhe ? nhe->id : 0); | |
1518 | ||
3057df51 SW |
1519 | return nhe; |
1520 | } | |
1521 | ||
1d48702e MS |
1522 | /* |
1523 | * Allocate backup nexthop info object. Typically these are embedded in | |
1524 | * nhg_hash_entry objects. | |
1525 | */ | |
1526 | struct nhg_backup_info *zebra_nhg_backup_alloc(void) | |
1527 | { | |
1528 | struct nhg_backup_info *p; | |
1529 | ||
1530 | p = XCALLOC(MTYPE_NHG, sizeof(struct nhg_backup_info)); | |
1531 | ||
1532 | p->nhe = zebra_nhg_alloc(); | |
1533 | ||
1534 | /* Identify the embedded group used to hold the list of backups */ | |
1535 | SET_FLAG(p->nhe->flags, NEXTHOP_GROUP_BACKUP); | |
1536 | ||
1537 | return p; | |
1538 | } | |
1539 | ||
1540 | /* | |
1541 | * Free backup nexthop info object, deal with any embedded allocations | |
1542 | */ | |
1543 | void zebra_nhg_backup_free(struct nhg_backup_info **p) | |
1544 | { | |
1545 | if (p && *p) { | |
1546 | if ((*p)->nhe) | |
1547 | zebra_nhg_free((*p)->nhe); | |
1548 | ||
1549 | XFREE(MTYPE_NHG, (*p)); | |
1550 | } | |
1551 | } | |
1552 | ||
1d48702e MS |
1553 | /* Accessor for backup nexthop group */ |
1554 | struct nexthop_group *zebra_nhg_get_backup_nhg(struct nhg_hash_entry *nhe) | |
1555 | { | |
1556 | struct nexthop_group *p = NULL; | |
1557 | ||
1558 | if (nhe) { | |
1559 | if (nhe->backup_info && nhe->backup_info->nhe) | |
1560 | p = &(nhe->backup_info->nhe->nhg); | |
1561 | } | |
1562 | ||
1563 | return p; | |
1564 | } | |
1565 | ||
1566 | /* | |
1567 | * Helper to return a copy of a backup_info - note that this is a shallow | |
1568 | * copy, meant to be used when creating a new nhe from info passed in with | |
1569 | * a route e.g. | |
1570 | */ | |
1571 | static struct nhg_backup_info * | |
1572 | nhg_backup_copy(const struct nhg_backup_info *orig) | |
1573 | { | |
1574 | struct nhg_backup_info *b; | |
1575 | ||
1576 | b = zebra_nhg_backup_alloc(); | |
1577 | ||
1578 | /* Copy list of nexthops */ | |
1579 | nexthop_group_copy(&(b->nhe->nhg), &(orig->nhe->nhg)); | |
1580 | ||
1581 | return b; | |
1582 | } | |
1583 | ||
5948f013 | 1584 | static void zebra_nhg_free_members(struct nhg_hash_entry *nhe) |
b599cd2a | 1585 | { |
c415d895 MS |
1586 | nexthops_free(nhe->nhg.nexthop); |
1587 | ||
1d48702e MS |
1588 | zebra_nhg_backup_free(&nhe->backup_info); |
1589 | ||
58396544 | 1590 | /* Decrement to remove connection ref */ |
37c6708b SW |
1591 | nhg_connected_tree_decrement_ref(&nhe->nhg_depends); |
1592 | nhg_connected_tree_free(&nhe->nhg_depends); | |
1593 | nhg_connected_tree_free(&nhe->nhg_dependents); | |
b599cd2a SW |
1594 | } |
1595 | ||
0eb97b86 | 1596 | void zebra_nhg_free(struct nhg_hash_entry *nhe) |
a95b8020 | 1597 | { |
377e29f7 MS |
1598 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) { |
1599 | /* Group or singleton? */ | |
1600 | if (nhe->nhg.nexthop && nhe->nhg.nexthop->next) | |
1601 | zlog_debug("%s: nhe %p (%u), refcnt %d", | |
f924db49 | 1602 | __func__, nhe, nhe->id, nhe->refcnt); |
377e29f7 MS |
1603 | else |
1604 | zlog_debug("%s: nhe %p (%u), refcnt %d, NH %pNHv", | |
f924db49 | 1605 | __func__, nhe, nhe->id, nhe->refcnt, |
377e29f7 MS |
1606 | nhe->nhg.nexthop); |
1607 | } | |
1608 | ||
38e40db1 SW |
1609 | if (nhe->refcnt) |
1610 | zlog_debug("nhe_id=%u hash refcnt=%d", nhe->id, nhe->refcnt); | |
1611 | ||
8e401b25 | 1612 | zebra_nhg_free_members(nhe); |
51d80884 SW |
1613 | |
1614 | XFREE(MTYPE_NHG, nhe); | |
a95b8020 SW |
1615 | } |
1616 | ||
0eb97b86 MS |
1617 | void zebra_nhg_hash_free(void *p) |
1618 | { | |
84591282 | 1619 | zebra_nhg_release_all_deps((struct nhg_hash_entry *)p); |
0eb97b86 MS |
1620 | zebra_nhg_free((struct nhg_hash_entry *)p); |
1621 | } | |
1622 | ||
d9f5b2f5 SW |
1623 | void zebra_nhg_decrement_ref(struct nhg_hash_entry *nhe) |
1624 | { | |
377e29f7 MS |
1625 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
1626 | zlog_debug("%s: nhe %p (%u) %d => %d", | |
1627 | __func__, nhe, nhe->id, nhe->refcnt, | |
1628 | nhe->refcnt - 1); | |
1629 | ||
e22e8001 SW |
1630 | nhe->refcnt--; |
1631 | ||
32e29e79 | 1632 | if (!zebra_nhg_depends_is_empty(nhe)) |
37c6708b | 1633 | nhg_connected_tree_decrement_ref(&nhe->nhg_depends); |
f54ef6a5 | 1634 | |
38e40db1 | 1635 | if (ZEBRA_NHG_CREATED(nhe) && nhe->refcnt <= 0) |
cb50cbc9 | 1636 | zebra_nhg_uninstall_kernel(nhe); |
7fd392cc SW |
1637 | } |
1638 | ||
7fd392cc SW |
1639 | void zebra_nhg_increment_ref(struct nhg_hash_entry *nhe) |
1640 | { | |
377e29f7 MS |
1641 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
1642 | zlog_debug("%s: nhe %p (%u) %d => %d", | |
1643 | __func__, nhe, nhe->id, nhe->refcnt, | |
1644 | nhe->refcnt + 1); | |
1645 | ||
e22e8001 SW |
1646 | nhe->refcnt++; |
1647 | ||
32e29e79 | 1648 | if (!zebra_nhg_depends_is_empty(nhe)) |
37c6708b | 1649 | nhg_connected_tree_increment_ref(&nhe->nhg_depends); |
e22e8001 | 1650 | } |
d9f5b2f5 | 1651 | |
5530d55d MS |
1652 | static struct nexthop *nexthop_set_resolved(afi_t afi, |
1653 | const struct nexthop *newhop, | |
1654 | struct nexthop *nexthop, | |
1655 | struct zebra_sr_policy *policy) | |
ad28e79a SW |
1656 | { |
1657 | struct nexthop *resolved_hop; | |
b43434ad SW |
1658 | uint8_t num_labels = 0; |
1659 | mpls_label_t labels[MPLS_MAX_LABELS]; | |
1660 | enum lsp_types_t label_type = ZEBRA_LSP_NONE; | |
1661 | int i = 0; | |
ad28e79a SW |
1662 | |
1663 | resolved_hop = nexthop_new(); | |
1664 | SET_FLAG(resolved_hop->flags, NEXTHOP_FLAG_ACTIVE); | |
1665 | ||
1666 | resolved_hop->vrf_id = nexthop->vrf_id; | |
1667 | switch (newhop->type) { | |
1668 | case NEXTHOP_TYPE_IPV4: | |
1669 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
1670 | /* If the resolving route specifies a gateway, use it */ | |
1671 | resolved_hop->type = newhop->type; | |
1672 | resolved_hop->gate.ipv4 = newhop->gate.ipv4; | |
1673 | ||
1674 | if (newhop->ifindex) { | |
1675 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
1676 | resolved_hop->ifindex = newhop->ifindex; | |
1677 | } | |
1678 | break; | |
1679 | case NEXTHOP_TYPE_IPV6: | |
1680 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
1681 | resolved_hop->type = newhop->type; | |
1682 | resolved_hop->gate.ipv6 = newhop->gate.ipv6; | |
1683 | ||
1684 | if (newhop->ifindex) { | |
1685 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
1686 | resolved_hop->ifindex = newhop->ifindex; | |
1687 | } | |
1688 | break; | |
1689 | case NEXTHOP_TYPE_IFINDEX: | |
1690 | /* If the resolving route is an interface route, | |
1691 | * it means the gateway we are looking up is connected | |
1692 | * to that interface. (The actual network is _not_ onlink). | |
1693 | * Therefore, the resolved route should have the original | |
1694 | * gateway as nexthop as it is directly connected. | |
1695 | * | |
1696 | * On Linux, we have to set the onlink netlink flag because | |
1697 | * otherwise, the kernel won't accept the route. | |
1698 | */ | |
1699 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
1700 | if (afi == AFI_IP) { | |
1701 | resolved_hop->type = NEXTHOP_TYPE_IPV4_IFINDEX; | |
1702 | resolved_hop->gate.ipv4 = nexthop->gate.ipv4; | |
1703 | } else if (afi == AFI_IP6) { | |
1704 | resolved_hop->type = NEXTHOP_TYPE_IPV6_IFINDEX; | |
1705 | resolved_hop->gate.ipv6 = nexthop->gate.ipv6; | |
1706 | } | |
1707 | resolved_hop->ifindex = newhop->ifindex; | |
1708 | break; | |
1709 | case NEXTHOP_TYPE_BLACKHOLE: | |
1710 | resolved_hop->type = NEXTHOP_TYPE_BLACKHOLE; | |
2dc359a6 | 1711 | resolved_hop->bh_type = newhop->bh_type; |
ad28e79a SW |
1712 | break; |
1713 | } | |
1714 | ||
1715 | if (newhop->flags & NEXTHOP_FLAG_ONLINK) | |
1716 | resolved_hop->flags |= NEXTHOP_FLAG_ONLINK; | |
1717 | ||
b43434ad | 1718 | /* Copy labels of the resolved route and the parent resolving to it */ |
31f937fb SM |
1719 | if (policy) { |
1720 | int i = 0; | |
1721 | ||
1722 | /* | |
1723 | * Don't push the first SID if the corresponding action in the | |
1724 | * LFIB is POP. | |
1725 | */ | |
1726 | if (!newhop->nh_label || !newhop->nh_label->num_labels | |
1727 | || newhop->nh_label->label[0] == MPLS_LABEL_IMPLICIT_NULL) | |
1728 | i = 1; | |
1729 | ||
1730 | for (; i < policy->segment_list.label_num; i++) | |
1731 | labels[num_labels++] = policy->segment_list.labels[i]; | |
1732 | label_type = policy->segment_list.type; | |
1733 | } else if (newhop->nh_label) { | |
6bc5d977 MS |
1734 | for (i = 0; i < newhop->nh_label->num_labels; i++) { |
1735 | /* Be a bit picky about overrunning the local array */ | |
1736 | if (num_labels >= MPLS_MAX_LABELS) { | |
1737 | if (IS_ZEBRA_DEBUG_NHG || IS_ZEBRA_DEBUG_RIB) | |
1738 | zlog_debug("%s: too many labels in newhop %pNHv", | |
1739 | __func__, newhop); | |
1740 | break; | |
1741 | } | |
b43434ad | 1742 | labels[num_labels++] = newhop->nh_label->label[i]; |
6bc5d977 MS |
1743 | } |
1744 | /* Use the "outer" type */ | |
b43434ad SW |
1745 | label_type = newhop->nh_label_type; |
1746 | } | |
1747 | ||
1748 | if (nexthop->nh_label) { | |
6bc5d977 MS |
1749 | for (i = 0; i < nexthop->nh_label->num_labels; i++) { |
1750 | /* Be a bit picky about overrunning the local array */ | |
1751 | if (num_labels >= MPLS_MAX_LABELS) { | |
1752 | if (IS_ZEBRA_DEBUG_NHG || IS_ZEBRA_DEBUG_RIB) | |
1753 | zlog_debug("%s: too many labels in nexthop %pNHv", | |
1754 | __func__, nexthop); | |
1755 | break; | |
1756 | } | |
b43434ad | 1757 | labels[num_labels++] = nexthop->nh_label->label[i]; |
6bc5d977 | 1758 | } |
b43434ad | 1759 | |
6bc5d977 MS |
1760 | /* If the parent has labels, use its type if |
1761 | * we don't already have one. | |
1762 | */ | |
1763 | if (label_type == ZEBRA_LSP_NONE) | |
1764 | label_type = nexthop->nh_label_type; | |
b43434ad SW |
1765 | } |
1766 | ||
1767 | if (num_labels) | |
1768 | nexthop_add_labels(resolved_hop, label_type, num_labels, | |
1769 | labels); | |
ad28e79a SW |
1770 | |
1771 | resolved_hop->rparent = nexthop; | |
50d89650 | 1772 | _nexthop_add(&nexthop->resolved, resolved_hop); |
5530d55d MS |
1773 | |
1774 | return resolved_hop; | |
ad28e79a SW |
1775 | } |
1776 | ||
6913cb1b SW |
1777 | /* Checks if nexthop we are trying to resolve to is valid */ |
1778 | static bool nexthop_valid_resolve(const struct nexthop *nexthop, | |
1779 | const struct nexthop *resolved) | |
1780 | { | |
1781 | /* Can't resolve to a recursive nexthop */ | |
1782 | if (CHECK_FLAG(resolved->flags, NEXTHOP_FLAG_RECURSIVE)) | |
1783 | return false; | |
1784 | ||
9d43854d MS |
1785 | /* Must be ACTIVE */ |
1786 | if (!CHECK_FLAG(resolved->flags, NEXTHOP_FLAG_ACTIVE)) | |
1787 | return false; | |
1788 | ||
3d30f6de SW |
1789 | /* Must not be duplicate */ |
1790 | if (CHECK_FLAG(resolved->flags, NEXTHOP_FLAG_DUPLICATE)) | |
1791 | return false; | |
1792 | ||
6913cb1b SW |
1793 | switch (nexthop->type) { |
1794 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
1795 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
1796 | /* If the nexthop we are resolving to does not match the | |
1797 | * ifindex for the nexthop the route wanted, its not valid. | |
1798 | */ | |
1799 | if (nexthop->ifindex != resolved->ifindex) | |
1800 | return false; | |
1801 | break; | |
1802 | case NEXTHOP_TYPE_IPV4: | |
1803 | case NEXTHOP_TYPE_IPV6: | |
1804 | case NEXTHOP_TYPE_IFINDEX: | |
1805 | case NEXTHOP_TYPE_BLACKHOLE: | |
1806 | break; | |
1807 | } | |
1808 | ||
1809 | return true; | |
1810 | } | |
1811 | ||
ad28e79a | 1812 | /* |
5530d55d MS |
1813 | * When resolving a recursive nexthop, capture backup nexthop(s) also |
1814 | * so they can be conveyed through the dataplane to the FIB. We'll look | |
1815 | * at the backups in the resolving nh 'nexthop' and its nhe, and copy them | |
1816 | * into the route's resolved nh 'resolved' and its nhe 'nhe'. | |
1817 | */ | |
1818 | static int resolve_backup_nexthops(const struct nexthop *nexthop, | |
1819 | const struct nhg_hash_entry *nhe, | |
1820 | struct nexthop *resolved, | |
1821 | struct nhg_hash_entry *resolve_nhe, | |
1822 | struct backup_nh_map_s *map) | |
1823 | { | |
1824 | int i, j, idx; | |
1825 | const struct nexthop *bnh; | |
1826 | struct nexthop *nh, *newnh; | |
a082cd9a MS |
1827 | mpls_label_t labels[MPLS_MAX_LABELS]; |
1828 | uint8_t num_labels; | |
5530d55d MS |
1829 | |
1830 | assert(nexthop->backup_num <= NEXTHOP_MAX_BACKUPS); | |
1831 | ||
5530d55d MS |
1832 | /* Locate backups from the original nexthop's backup index and nhe */ |
1833 | for (i = 0; i < nexthop->backup_num; i++) { | |
1834 | idx = nexthop->backup_idx[i]; | |
1835 | ||
1836 | /* Do we already know about this particular backup? */ | |
1837 | for (j = 0; j < map->map_count; j++) { | |
1838 | if (map->map[j].orig_idx == idx) | |
1839 | break; | |
1840 | } | |
1841 | ||
1842 | if (j < map->map_count) { | |
1843 | resolved->backup_idx[resolved->backup_num] = | |
1844 | map->map[j].new_idx; | |
1845 | resolved->backup_num++; | |
1846 | ||
c56c16eb MS |
1847 | SET_FLAG(resolved->flags, NEXTHOP_FLAG_HAS_BACKUP); |
1848 | ||
5530d55d MS |
1849 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1850 | zlog_debug("%s: found map idx orig %d, new %d", | |
1851 | __func__, map->map[j].orig_idx, | |
1852 | map->map[j].new_idx); | |
1853 | ||
1854 | continue; | |
1855 | } | |
1856 | ||
1857 | /* We can't handle any new map entries at this point. */ | |
1858 | if (map->map_count == MULTIPATH_NUM) | |
1859 | break; | |
1860 | ||
1861 | /* Need to create/copy a new backup */ | |
1862 | bnh = nhe->backup_info->nhe->nhg.nexthop; | |
1863 | for (j = 0; j < idx; j++) { | |
1864 | if (bnh == NULL) | |
1865 | break; | |
1866 | bnh = bnh->next; | |
1867 | } | |
1868 | ||
1869 | /* Whoops - bad index in the nexthop? */ | |
1870 | if (bnh == NULL) | |
1871 | continue; | |
1872 | ||
c56c16eb MS |
1873 | if (resolve_nhe->backup_info == NULL) |
1874 | resolve_nhe->backup_info = zebra_nhg_backup_alloc(); | |
1875 | ||
5530d55d MS |
1876 | /* Update backup info in the resolving nexthop and its nhe */ |
1877 | newnh = nexthop_dup_no_recurse(bnh, NULL); | |
1878 | ||
a082cd9a MS |
1879 | /* We may need some special handling for mpls labels: the new |
1880 | * backup needs to carry the recursive nexthop's labels, | |
1881 | * if any: they may be vrf labels e.g. | |
1882 | * The original/inner labels are in the stack of 'resolve_nhe', | |
1883 | * if that is longer than the stack in 'nexthop'. | |
1884 | */ | |
1885 | if (newnh->nh_label && resolved->nh_label && | |
1886 | nexthop->nh_label) { | |
1887 | if (resolved->nh_label->num_labels > | |
1888 | nexthop->nh_label->num_labels) { | |
1889 | /* Prepare new label stack */ | |
1890 | num_labels = 0; | |
1891 | for (j = 0; j < newnh->nh_label->num_labels; | |
1892 | j++) { | |
1893 | labels[j] = newnh->nh_label->label[j]; | |
1894 | num_labels++; | |
1895 | } | |
1896 | ||
1897 | /* Include inner labels */ | |
1898 | for (j = nexthop->nh_label->num_labels; | |
1899 | j < resolved->nh_label->num_labels; | |
1900 | j++) { | |
1901 | labels[num_labels] = | |
1902 | resolved->nh_label->label[j]; | |
1903 | num_labels++; | |
1904 | } | |
1905 | ||
1906 | /* Replace existing label stack in the backup */ | |
1907 | nexthop_del_labels(newnh); | |
1908 | nexthop_add_labels(newnh, bnh->nh_label_type, | |
1909 | num_labels, labels); | |
1910 | } | |
1911 | } | |
1912 | ||
5530d55d MS |
1913 | /* Need to compute the new backup index in the new |
1914 | * backup list, and add to map struct. | |
1915 | */ | |
1916 | j = 0; | |
1917 | nh = resolve_nhe->backup_info->nhe->nhg.nexthop; | |
1918 | if (nh) { | |
1919 | while (nh->next) { | |
1920 | nh = nh->next; | |
1921 | j++; | |
1922 | } | |
1923 | ||
1924 | nh->next = newnh; | |
c56c16eb | 1925 | j++; |
5530d55d MS |
1926 | |
1927 | } else /* First one */ | |
1928 | resolve_nhe->backup_info->nhe->nhg.nexthop = newnh; | |
1929 | ||
1930 | /* Capture index */ | |
1931 | resolved->backup_idx[resolved->backup_num] = j; | |
1932 | resolved->backup_num++; | |
1933 | ||
c56c16eb MS |
1934 | SET_FLAG(resolved->flags, NEXTHOP_FLAG_HAS_BACKUP); |
1935 | ||
5530d55d MS |
1936 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
1937 | zlog_debug("%s: added idx orig %d, new %d", | |
1938 | __func__, idx, j); | |
1939 | ||
1940 | /* Update map/cache */ | |
1941 | map->map[map->map_count].orig_idx = idx; | |
1942 | map->map[map->map_count].new_idx = j; | |
1943 | map->map_count++; | |
1944 | } | |
1945 | ||
1946 | return 0; | |
1947 | } | |
1948 | ||
1949 | /* | |
1950 | * Given a nexthop we need to properly recursively resolve, | |
1951 | * do a table lookup to find and match if at all possible. | |
1952 | * Set the nexthop->ifindex and resolution info as appropriate. | |
ad28e79a | 1953 | */ |
5530d55d | 1954 | static int nexthop_active(struct nexthop *nexthop, struct nhg_hash_entry *nhe, |
9b4ab909 MS |
1955 | const struct prefix *top, int type, uint32_t flags, |
1956 | uint32_t *pmtu) | |
ad28e79a SW |
1957 | { |
1958 | struct prefix p; | |
1959 | struct route_table *table; | |
1960 | struct route_node *rn; | |
1961 | struct route_entry *match = NULL; | |
1962 | int resolved; | |
31f937fb | 1963 | zebra_nhlfe_t *nhlfe; |
ad28e79a SW |
1964 | struct nexthop *newhop; |
1965 | struct interface *ifp; | |
1966 | rib_dest_t *dest; | |
5a0bdc78 | 1967 | struct zebra_vrf *zvrf; |
31f937fb SM |
1968 | struct in_addr local_ipv4; |
1969 | struct in_addr *ipv4; | |
5530d55d | 1970 | afi_t afi = AFI_IP; |
ad28e79a | 1971 | |
9b4ab909 | 1972 | /* Reset some nexthop attributes that we'll recompute if necessary */ |
ad28e79a | 1973 | if ((nexthop->type == NEXTHOP_TYPE_IPV4) |
9b4ab909 | 1974 | || (nexthop->type == NEXTHOP_TYPE_IPV6)) |
ad28e79a SW |
1975 | nexthop->ifindex = 0; |
1976 | ||
1977 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE); | |
1978 | nexthops_free(nexthop->resolved); | |
1979 | nexthop->resolved = NULL; | |
377e29f7 | 1980 | |
ad28e79a | 1981 | /* |
5530d55d | 1982 | * Set afi based on nexthop type. |
9b4ab909 MS |
1983 | * Some nexthop types get special handling, possibly skipping |
1984 | * the normal processing. | |
ad28e79a | 1985 | */ |
9b4ab909 MS |
1986 | switch (nexthop->type) { |
1987 | case NEXTHOP_TYPE_IFINDEX: | |
5530d55d | 1988 | |
9b4ab909 MS |
1989 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); |
1990 | /* | |
1991 | * If the interface exists and its operative or its a kernel | |
1992 | * route and interface is up, its active. We trust kernel routes | |
1993 | * to be good. | |
1994 | */ | |
1995 | if (ifp | |
1996 | && (if_is_operative(ifp) | |
1997 | || (if_is_up(ifp) | |
1998 | && (type == ZEBRA_ROUTE_KERNEL | |
1999 | || type == ZEBRA_ROUTE_SYSTEM)))) | |
2000 | return 1; | |
2001 | else | |
2002 | return 0; | |
2003 | break; | |
2004 | ||
2005 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
5530d55d MS |
2006 | afi = AFI_IP6; |
2007 | ||
9b4ab909 MS |
2008 | if (IN6_IS_ADDR_LINKLOCAL(&nexthop->gate.ipv6)) { |
2009 | ifp = if_lookup_by_index(nexthop->ifindex, | |
2010 | nexthop->vrf_id); | |
2011 | if (ifp && if_is_operative(ifp)) | |
2012 | return 1; | |
2013 | else | |
2014 | return 0; | |
2015 | } | |
2016 | break; | |
2017 | ||
9b4ab909 MS |
2018 | case NEXTHOP_TYPE_IPV4: |
2019 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
5530d55d MS |
2020 | afi = AFI_IP; |
2021 | break; | |
9b4ab909 | 2022 | case NEXTHOP_TYPE_IPV6: |
5530d55d | 2023 | afi = AFI_IP6; |
9b4ab909 | 2024 | break; |
5530d55d MS |
2025 | |
2026 | case NEXTHOP_TYPE_BLACKHOLE: | |
2027 | return 1; | |
9b4ab909 MS |
2028 | } |
2029 | ||
ad28e79a | 2030 | /* |
12b4d77b | 2031 | * If the nexthop has been marked as 'onlink' we just need to make |
2032 | * sure the nexthop's interface is known and is operational. | |
ad28e79a SW |
2033 | */ |
2034 | if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK)) { | |
2035 | ifp = if_lookup_by_index(nexthop->ifindex, nexthop->vrf_id); | |
2036 | if (!ifp) { | |
a24d04f4 | 2037 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
12b4d77b | 2038 | zlog_debug("nexthop %pNHv marked onlink but nhif %u doesn't exist", |
2039 | nexthop, nexthop->ifindex); | |
ad28e79a SW |
2040 | return 0; |
2041 | } | |
12b4d77b | 2042 | if (!if_is_operative(ifp)) { |
a24d04f4 | 2043 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
12b4d77b | 2044 | zlog_debug("nexthop %pNHv marked onlink but nhif %s is not operational", |
2045 | nexthop, ifp->name); | |
2d3c57e6 | 2046 | return 0; |
ad28e79a | 2047 | } |
12b4d77b | 2048 | return 1; |
ad28e79a SW |
2049 | } |
2050 | ||
13ccce6e | 2051 | if (top |
12256b84 | 2052 | && ((top->family == AF_INET && top->prefixlen == IPV4_MAX_BITLEN |
13ccce6e DA |
2053 | && nexthop->gate.ipv4.s_addr == top->u.prefix4.s_addr) |
2054 | || (top->family == AF_INET6 && top->prefixlen == IPV6_MAX_BITLEN | |
8643c2e5 DA |
2055 | && memcmp(&nexthop->gate.ipv6, &top->u.prefix6, |
2056 | IPV6_MAX_BYTELEN) | |
13ccce6e | 2057 | == 0))) { |
4dcc2276 DS |
2058 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2059 | zlog_debug( | |
d6951e5e | 2060 | " :%s: Attempting to install a max prefixlength route through itself", |
5e81f5dd | 2061 | __func__); |
4dcc2276 DS |
2062 | return 0; |
2063 | } | |
2064 | ||
92d6f769 K |
2065 | /* Validation for ipv4 mapped ipv6 nexthop. */ |
2066 | if (IS_MAPPED_IPV6(&nexthop->gate.ipv6)) { | |
2067 | afi = AFI_IP; | |
31f937fb SM |
2068 | ipv4 = &local_ipv4; |
2069 | ipv4_mapped_ipv6_to_ipv4(&nexthop->gate.ipv6, ipv4); | |
2070 | } else { | |
2071 | ipv4 = &nexthop->gate.ipv4; | |
2072 | } | |
2073 | ||
9b4ab909 MS |
2074 | /* Processing for nexthops with SR 'color' attribute, using |
2075 | * the corresponding SR policy object. | |
2076 | */ | |
31f937fb SM |
2077 | if (nexthop->srte_color) { |
2078 | struct ipaddr endpoint = {0}; | |
2079 | struct zebra_sr_policy *policy; | |
2080 | ||
2081 | switch (afi) { | |
2082 | case AFI_IP: | |
2083 | endpoint.ipa_type = IPADDR_V4; | |
2084 | endpoint.ipaddr_v4 = *ipv4; | |
2085 | break; | |
2086 | case AFI_IP6: | |
2087 | endpoint.ipa_type = IPADDR_V6; | |
2088 | endpoint.ipaddr_v6 = nexthop->gate.ipv6; | |
2089 | break; | |
2090 | default: | |
2091 | flog_err(EC_LIB_DEVELOPMENT, | |
2092 | "%s: unknown address-family: %u", __func__, | |
2093 | afi); | |
2094 | exit(1); | |
2095 | } | |
2096 | ||
2097 | policy = zebra_sr_policy_find(nexthop->srte_color, &endpoint); | |
2098 | if (policy && policy->status == ZEBRA_SR_POLICY_UP) { | |
2099 | resolved = 0; | |
2100 | frr_each_safe (nhlfe_list, &policy->lsp->nhlfe_list, | |
2101 | nhlfe) { | |
2102 | if (!CHECK_FLAG(nhlfe->flags, | |
2103 | NHLFE_FLAG_SELECTED) | |
2104 | || CHECK_FLAG(nhlfe->flags, | |
2105 | NHLFE_FLAG_DELETED)) | |
2106 | continue; | |
2107 | SET_FLAG(nexthop->flags, | |
2108 | NEXTHOP_FLAG_RECURSIVE); | |
2109 | nexthop_set_resolved(afi, nhlfe->nexthop, | |
2110 | nexthop, policy); | |
2111 | resolved = 1; | |
2112 | } | |
2113 | if (resolved) | |
2114 | return 1; | |
2115 | } | |
92d6f769 K |
2116 | } |
2117 | ||
ad28e79a SW |
2118 | /* Make lookup prefix. */ |
2119 | memset(&p, 0, sizeof(struct prefix)); | |
2120 | switch (afi) { | |
2121 | case AFI_IP: | |
2122 | p.family = AF_INET; | |
936fbaef | 2123 | p.prefixlen = IPV4_MAX_BITLEN; |
31f937fb | 2124 | p.u.prefix4 = *ipv4; |
ad28e79a SW |
2125 | break; |
2126 | case AFI_IP6: | |
2127 | p.family = AF_INET6; | |
f4d81e55 | 2128 | p.prefixlen = IPV6_MAX_BITLEN; |
ad28e79a SW |
2129 | p.u.prefix6 = nexthop->gate.ipv6; |
2130 | break; | |
2131 | default: | |
2132 | assert(afi != AFI_IP && afi != AFI_IP6); | |
2133 | break; | |
2134 | } | |
2135 | /* Lookup table. */ | |
2136 | table = zebra_vrf_table(afi, SAFI_UNICAST, nexthop->vrf_id); | |
5a0bdc78 PG |
2137 | /* get zvrf */ |
2138 | zvrf = zebra_vrf_lookup_by_id(nexthop->vrf_id); | |
2139 | if (!table || !zvrf) { | |
ad28e79a | 2140 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
d6951e5e | 2141 | zlog_debug(" %s: Table not found", __func__); |
ad28e79a SW |
2142 | return 0; |
2143 | } | |
2144 | ||
2145 | rn = route_node_match(table, (struct prefix *)&p); | |
2146 | while (rn) { | |
2147 | route_unlock_node(rn); | |
2148 | ||
2149 | /* Lookup should halt if we've matched against ourselves ('top', | |
2150 | * if specified) - i.e., we cannot have a nexthop NH1 is | |
2151 | * resolved by a route NH1. The exception is if the route is a | |
2152 | * host route. | |
2153 | */ | |
9b4ab909 | 2154 | if (prefix_same(&rn->p, top)) |
12256b84 DA |
2155 | if (((afi == AFI_IP) |
2156 | && (rn->p.prefixlen != IPV4_MAX_BITLEN)) | |
13ccce6e DA |
2157 | || ((afi == AFI_IP6) |
2158 | && (rn->p.prefixlen != IPV6_MAX_BITLEN))) { | |
ad28e79a SW |
2159 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2160 | zlog_debug( | |
d6951e5e | 2161 | " %s: Matched against ourself and prefix length is not max bit length", |
5e81f5dd | 2162 | __func__); |
ad28e79a SW |
2163 | return 0; |
2164 | } | |
2165 | ||
2166 | /* Pick up selected route. */ | |
2167 | /* However, do not resolve over default route unless explicitly | |
2d3c57e6 SW |
2168 | * allowed. |
2169 | */ | |
ad28e79a | 2170 | if (is_default_prefix(&rn->p) |
5a0bdc78 | 2171 | && !rnh_resolve_via_default(zvrf, p.family)) { |
ad28e79a SW |
2172 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
2173 | zlog_debug( | |
d6951e5e | 2174 | " :%s: Resolved against default route", |
5e81f5dd | 2175 | __func__); |
ad28e79a SW |
2176 | return 0; |
2177 | } | |
2178 | ||
2179 | dest = rib_dest_from_rnode(rn); | |
2180 | if (dest && dest->selected_fib | |
2181 | && !CHECK_FLAG(dest->selected_fib->status, | |
2182 | ROUTE_ENTRY_REMOVED) | |
2183 | && dest->selected_fib->type != ZEBRA_ROUTE_TABLE) | |
2184 | match = dest->selected_fib; | |
2185 | ||
2186 | /* If there is no selected route or matched route is EGP, go up | |
2d3c57e6 SW |
2187 | * tree. |
2188 | */ | |
ad28e79a SW |
2189 | if (!match) { |
2190 | do { | |
2191 | rn = rn->parent; | |
2192 | } while (rn && rn->info == NULL); | |
2193 | if (rn) | |
2194 | route_lock_node(rn); | |
2195 | ||
2196 | continue; | |
2197 | } | |
2198 | ||
2199 | if (match->type == ZEBRA_ROUTE_CONNECT) { | |
2200 | /* Directly point connected route. */ | |
c415d895 | 2201 | newhop = match->nhe->nhg.nexthop; |
ad28e79a SW |
2202 | if (newhop) { |
2203 | if (nexthop->type == NEXTHOP_TYPE_IPV4 | |
2204 | || nexthop->type == NEXTHOP_TYPE_IPV6) | |
2205 | nexthop->ifindex = newhop->ifindex; | |
c479909b | 2206 | else if (nexthop->ifindex != newhop->ifindex) { |
a24d04f4 | 2207 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
81194fee DS |
2208 | zlog_debug( |
2209 | "%s: %pNHv given ifindex does not match nexthops ifindex found found: %pNHv", | |
2210 | __func__, nexthop, | |
2211 | newhop); | |
c479909b SW |
2212 | /* |
2213 | * NEXTHOP_TYPE_*_IFINDEX but ifindex | |
2214 | * doesn't match what we found. | |
2215 | */ | |
2216 | return 0; | |
2217 | } | |
ad28e79a | 2218 | } |
377e29f7 MS |
2219 | |
2220 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
2221 | zlog_debug("%s: CONNECT match %p (%u), newhop %pNHv", | |
2222 | __func__, match, | |
2223 | match->nhe->id, newhop); | |
2224 | ||
ad28e79a | 2225 | return 1; |
9b4ab909 | 2226 | } else if (CHECK_FLAG(flags, ZEBRA_FLAG_ALLOW_RECURSION)) { |
f2646720 | 2227 | struct nexthop_group *nhg; |
5530d55d MS |
2228 | struct nexthop *resolver; |
2229 | struct backup_nh_map_s map = {}; | |
f2646720 | 2230 | |
ad28e79a | 2231 | resolved = 0; |
f2646720 | 2232 | |
92ad0c55 MS |
2233 | /* Only useful if installed */ |
2234 | if (!CHECK_FLAG(match->status, ROUTE_ENTRY_INSTALLED)) { | |
a24d04f4 | 2235 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) |
92ad0c55 MS |
2236 | zlog_debug("%s: match %p (%u) not installed", |
2237 | __func__, match, | |
2238 | match->nhe->id); | |
2239 | ||
2240 | goto done_with_match; | |
2241 | } | |
2242 | ||
3c0e1622 MS |
2243 | /* Examine installed nexthops; note that there |
2244 | * may not be any installed primary nexthops if | |
2245 | * only backups are installed. | |
2246 | */ | |
2247 | nhg = rib_get_fib_nhg(match); | |
f2646720 | 2248 | for (ALL_NEXTHOPS_PTR(nhg, newhop)) { |
6913cb1b | 2249 | if (!nexthop_valid_resolve(nexthop, newhop)) |
ad28e79a SW |
2250 | continue; |
2251 | ||
377e29f7 MS |
2252 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
2253 | zlog_debug("%s: RECURSIVE match %p (%u), newhop %pNHv", | |
2254 | __func__, match, | |
2255 | match->nhe->id, newhop); | |
2256 | ||
ad28e79a SW |
2257 | SET_FLAG(nexthop->flags, |
2258 | NEXTHOP_FLAG_RECURSIVE); | |
5530d55d MS |
2259 | resolver = nexthop_set_resolved(afi, newhop, |
2260 | nexthop, NULL); | |
ad28e79a | 2261 | resolved = 1; |
5530d55d MS |
2262 | |
2263 | /* If there are backup nexthops, capture | |
2264 | * that info with the resolving nexthop. | |
2265 | */ | |
2266 | if (resolver && newhop->backup_num > 0) { | |
2267 | resolve_backup_nexthops(newhop, | |
2268 | match->nhe, | |
2269 | resolver, nhe, | |
2270 | &map); | |
2271 | } | |
ad28e79a | 2272 | } |
f924db49 | 2273 | |
9959f1da MS |
2274 | /* Examine installed backup nexthops, if any. There |
2275 | * are only installed backups *if* there is a | |
aa458838 MS |
2276 | * dedicated fib list. The UI can also control use |
2277 | * of backups for resolution. | |
9959f1da MS |
2278 | */ |
2279 | nhg = rib_get_fib_backup_nhg(match); | |
aa458838 MS |
2280 | if (!use_recursive_backups || |
2281 | nhg == NULL || nhg->nexthop == NULL) | |
f2646720 MS |
2282 | goto done_with_match; |
2283 | ||
2284 | for (ALL_NEXTHOPS_PTR(nhg, newhop)) { | |
f2646720 MS |
2285 | if (!nexthop_valid_resolve(nexthop, newhop)) |
2286 | continue; | |
2287 | ||
2288 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
2289 | zlog_debug("%s: RECURSIVE match backup %p (%u), newhop %pNHv", | |
2290 | __func__, match, | |
2291 | match->nhe->id, newhop); | |
2292 | ||
2293 | SET_FLAG(nexthop->flags, | |
2294 | NEXTHOP_FLAG_RECURSIVE); | |
31f937fb SM |
2295 | nexthop_set_resolved(afi, newhop, nexthop, |
2296 | NULL); | |
f2646720 MS |
2297 | resolved = 1; |
2298 | } | |
9b4ab909 | 2299 | |
f2646720 | 2300 | done_with_match: |
9b4ab909 MS |
2301 | /* Capture resolving mtu */ |
2302 | if (resolved) { | |
2303 | if (pmtu) | |
2304 | *pmtu = match->mtu; | |
2305 | ||
2306 | } else if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
d6951e5e DL |
2307 | zlog_debug( |
2308 | " %s: Recursion failed to find", | |
2309 | __func__); | |
f924db49 | 2310 | |
ad28e79a | 2311 | return resolved; |
ad28e79a SW |
2312 | } else { |
2313 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) { | |
2314 | zlog_debug( | |
d6951e5e | 2315 | " %s: Route Type %s has not turned on recursion", |
9b4ab909 MS |
2316 | __func__, zebra_route_string(type)); |
2317 | if (type == ZEBRA_ROUTE_BGP | |
2318 | && !CHECK_FLAG(flags, ZEBRA_FLAG_IBGP)) | |
ad28e79a | 2319 | zlog_debug( |
d6951e5e | 2320 | " EBGP: see \"disable-ebgp-connected-route-check\" or \"disable-connected-check\""); |
ad28e79a SW |
2321 | } |
2322 | return 0; | |
2323 | } | |
2324 | } | |
2325 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
d6951e5e DL |
2326 | zlog_debug(" %s: Nexthop did not lookup in table", |
2327 | __func__); | |
ad28e79a SW |
2328 | return 0; |
2329 | } | |
2330 | ||
2331 | /* This function verifies reachability of one given nexthop, which can be | |
2332 | * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored | |
2333 | * in nexthop->flags field. The nexthop->ifindex will be updated | |
5530d55d MS |
2334 | * appropriately as well. |
2335 | * | |
2336 | * An existing route map can turn an otherwise active nexthop into inactive, | |
2337 | * but not vice versa. | |
ad28e79a SW |
2338 | * |
2339 | * The return value is the final value of 'ACTIVE' flag. | |
2340 | */ | |
2341 | static unsigned nexthop_active_check(struct route_node *rn, | |
2342 | struct route_entry *re, | |
5530d55d MS |
2343 | struct nexthop *nexthop, |
2344 | struct nhg_hash_entry *nhe) | |
ad28e79a | 2345 | { |
b68885f9 | 2346 | route_map_result_t ret = RMAP_PERMITMATCH; |
f5b7e50f | 2347 | afi_t family; |
ad28e79a SW |
2348 | const struct prefix *p, *src_p; |
2349 | struct zebra_vrf *zvrf; | |
9b4ab909 | 2350 | uint32_t mtu = 0; |
ad28e79a SW |
2351 | |
2352 | srcdest_rnode_prefixes(rn, &p, &src_p); | |
2353 | ||
2354 | if (rn->p.family == AF_INET) | |
2355 | family = AFI_IP; | |
2356 | else if (rn->p.family == AF_INET6) | |
2357 | family = AFI_IP6; | |
2358 | else | |
2359 | family = 0; | |
ee94437e | 2360 | |
9b4ab909 MS |
2361 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
2362 | zlog_debug("%s: re %p, nexthop %pNHv", __func__, re, nexthop); | |
2363 | ||
2364 | /* | |
2365 | * If the kernel has sent us a NEW route, then | |
2366 | * by golly gee whiz it's a good route. | |
2367 | * | |
2368 | * If its an already INSTALLED route we have already handled, then the | |
2369 | * kernel route's nexthop might have became unreachable | |
2370 | * and we have to handle that. | |
2371 | */ | |
2372 | if (!CHECK_FLAG(re->status, ROUTE_ENTRY_INSTALLED) && | |
2373 | (re->type == ZEBRA_ROUTE_KERNEL || | |
2374 | re->type == ZEBRA_ROUTE_SYSTEM)) { | |
2375 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2376 | goto skip_check; | |
2377 | } | |
2378 | ||
ad28e79a SW |
2379 | switch (nexthop->type) { |
2380 | case NEXTHOP_TYPE_IFINDEX: | |
5530d55d | 2381 | if (nexthop_active(nexthop, nhe, &rn->p, re->type, |
9b4ab909 | 2382 | re->flags, &mtu)) |
ad28e79a SW |
2383 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
2384 | else | |
2385 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2386 | break; | |
2387 | case NEXTHOP_TYPE_IPV4: | |
2388 | case NEXTHOP_TYPE_IPV4_IFINDEX: | |
2389 | family = AFI_IP; | |
5530d55d | 2390 | if (nexthop_active(nexthop, nhe, &rn->p, re->type, |
9b4ab909 | 2391 | re->flags, &mtu)) |
ad28e79a SW |
2392 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
2393 | else | |
2394 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2395 | break; | |
2396 | case NEXTHOP_TYPE_IPV6: | |
2397 | family = AFI_IP6; | |
5530d55d | 2398 | if (nexthop_active(nexthop, nhe, &rn->p, re->type, |
9b4ab909 | 2399 | re->flags, &mtu)) |
ad28e79a SW |
2400 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
2401 | else | |
2402 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2403 | break; | |
2404 | case NEXTHOP_TYPE_IPV6_IFINDEX: | |
2405 | /* RFC 5549, v4 prefix with v6 NH */ | |
2406 | if (rn->p.family != AF_INET) | |
2407 | family = AFI_IP6; | |
9b4ab909 | 2408 | |
5530d55d | 2409 | if (nexthop_active(nexthop, nhe, &rn->p, re->type, |
9b4ab909 MS |
2410 | re->flags, &mtu)) |
2411 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2412 | else | |
2413 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
ad28e79a SW |
2414 | break; |
2415 | case NEXTHOP_TYPE_BLACKHOLE: | |
2416 | SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2417 | break; | |
2418 | default: | |
2419 | break; | |
2420 | } | |
377e29f7 | 2421 | |
9b4ab909 MS |
2422 | skip_check: |
2423 | ||
ad28e79a SW |
2424 | if (!CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)) { |
2425 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
377e29f7 MS |
2426 | zlog_debug(" %s: Unable to find active nexthop", |
2427 | __func__); | |
ad28e79a SW |
2428 | return 0; |
2429 | } | |
2430 | ||
9b4ab909 MS |
2431 | /* Capture recursive nexthop mtu. |
2432 | * TODO -- the code used to just reset the re's value to zero | |
2433 | * for each nexthop, and then jam any resolving route's mtu value in, | |
2434 | * whether or not that was zero, or lt/gt any existing value? The | |
2435 | * way this is used appears to be as a floor value, so let's try | |
2436 | * using it that way here. | |
2437 | */ | |
2438 | if (mtu > 0) { | |
2439 | if (re->nexthop_mtu == 0 || re->nexthop_mtu > mtu) | |
2440 | re->nexthop_mtu = mtu; | |
2441 | } | |
2442 | ||
ad28e79a SW |
2443 | /* XXX: What exactly do those checks do? Do we support |
2444 | * e.g. IPv4 routes with IPv6 nexthops or vice versa? | |
2445 | */ | |
2446 | if (RIB_SYSTEM_ROUTE(re) || (family == AFI_IP && p->family != AF_INET) | |
2447 | || (family == AFI_IP6 && p->family != AF_INET6)) | |
2448 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2449 | ||
2450 | /* The original code didn't determine the family correctly | |
2451 | * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi | |
2452 | * from the rib_table_info in those cases. | |
2453 | * Possibly it may be better to use only the rib_table_info | |
2454 | * in every case. | |
2455 | */ | |
9b4ab909 | 2456 | if (family == 0) { |
630d5962 | 2457 | struct rib_table_info *info; |
ad28e79a SW |
2458 | |
2459 | info = srcdest_rnode_table_info(rn); | |
2460 | family = info->afi; | |
2461 | } | |
2462 | ||
2463 | memset(&nexthop->rmap_src.ipv6, 0, sizeof(union g_addr)); | |
2464 | ||
2465 | zvrf = zebra_vrf_lookup_by_id(nexthop->vrf_id); | |
2466 | if (!zvrf) { | |
2467 | if (IS_ZEBRA_DEBUG_RIB_DETAILED) | |
d6951e5e | 2468 | zlog_debug(" %s: zvrf is NULL", __func__); |
ad28e79a SW |
2469 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); |
2470 | } | |
2471 | ||
2472 | /* It'll get set if required inside */ | |
2473 | ret = zebra_route_map_check(family, re->type, re->instance, p, nexthop, | |
2474 | zvrf, re->tag); | |
2475 | if (ret == RMAP_DENYMATCH) { | |
2476 | if (IS_ZEBRA_DEBUG_RIB) { | |
ad28e79a | 2477 | zlog_debug( |
7e010c4b DS |
2478 | "%u:%pRN: Filtering out with NH out %s due to route map", |
2479 | re->vrf_id, rn, | |
ad28e79a SW |
2480 | ifindex2ifname(nexthop->ifindex, |
2481 | nexthop->vrf_id)); | |
2482 | } | |
2483 | UNSET_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2484 | } | |
2485 | return CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2486 | } | |
2487 | ||
086e4e02 SW |
2488 | /* Helper function called after resolution to walk nhg rb trees |
2489 | * and toggle the NEXTHOP_GROUP_VALID flag if the nexthop | |
2490 | * is active on singleton NHEs. | |
2491 | */ | |
2492 | static bool zebra_nhg_set_valid_if_active(struct nhg_hash_entry *nhe) | |
2493 | { | |
2494 | struct nhg_connected *rb_node_dep = NULL; | |
2495 | bool valid = false; | |
2496 | ||
2497 | if (!zebra_nhg_depends_is_empty(nhe)) { | |
2498 | /* Is at least one depend valid? */ | |
2499 | frr_each(nhg_connected_tree, &nhe->nhg_depends, rb_node_dep) { | |
2500 | if (zebra_nhg_set_valid_if_active(rb_node_dep->nhe)) | |
2501 | valid = true; | |
2502 | } | |
2503 | ||
2504 | goto done; | |
2505 | } | |
2506 | ||
2507 | /* should be fully resolved singleton at this point */ | |
2508 | if (CHECK_FLAG(nhe->nhg.nexthop->flags, NEXTHOP_FLAG_ACTIVE)) | |
2509 | valid = true; | |
2510 | ||
2511 | done: | |
2512 | if (valid) | |
2513 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_VALID); | |
2514 | ||
2515 | return valid; | |
2516 | } | |
2517 | ||
ad28e79a | 2518 | /* |
5530d55d | 2519 | * Process a list of nexthops, given an nhe, determining |
377e29f7 | 2520 | * whether each one is ACTIVE/installable at this time. |
ad28e79a | 2521 | */ |
377e29f7 MS |
2522 | static uint32_t nexthop_list_active_update(struct route_node *rn, |
2523 | struct route_entry *re, | |
5530d55d MS |
2524 | struct nhg_hash_entry *nhe, |
2525 | bool is_backup) | |
ad28e79a | 2526 | { |
ad28e79a SW |
2527 | union g_addr prev_src; |
2528 | unsigned int prev_active, new_active; | |
2529 | ifindex_t prev_index; | |
377e29f7 | 2530 | uint32_t counter = 0; |
f2646720 | 2531 | struct nexthop *nexthop; |
5530d55d | 2532 | struct nexthop_group *nhg = &nhe->nhg; |
f2646720 MS |
2533 | |
2534 | nexthop = nhg->nexthop; | |
e22e8001 | 2535 | |
9b4ab909 MS |
2536 | /* Init recursive nh mtu */ |
2537 | re->nexthop_mtu = 0; | |
2538 | ||
377e29f7 MS |
2539 | /* Process nexthops one-by-one */ |
2540 | for ( ; nexthop; nexthop = nexthop->next) { | |
ad28e79a | 2541 | |
ad28e79a | 2542 | /* No protocol daemon provides src and so we're skipping |
377e29f7 MS |
2543 | * tracking it |
2544 | */ | |
ad28e79a SW |
2545 | prev_src = nexthop->rmap_src; |
2546 | prev_active = CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE); | |
2547 | prev_index = nexthop->ifindex; | |
5530d55d MS |
2548 | |
2549 | /* Include the containing nhe for primary nexthops: if there's | |
2550 | * recursive resolution, we capture the backup info also. | |
2551 | */ | |
2552 | new_active = | |
2553 | nexthop_active_check(rn, re, nexthop, | |
2554 | (is_backup ? NULL : nhe)); | |
2555 | ||
ad28e79a SW |
2556 | /* |
2557 | * We need to respect the multipath_num here | |
2558 | * as that what we should be able to install from | |
377e29f7 | 2559 | * a multipath perspective should not be a data plane |
ad28e79a SW |
2560 | * decision point. |
2561 | */ | |
377e29f7 | 2562 | if (new_active && counter >= zrouter.multipath_num) { |
4c55b5ff SW |
2563 | struct nexthop *nh; |
2564 | ||
2565 | /* Set it and its resolved nexthop as inactive. */ | |
2566 | for (nh = nexthop; nh; nh = nh->resolved) | |
2567 | UNSET_FLAG(nh->flags, NEXTHOP_FLAG_ACTIVE); | |
2568 | ||
ad28e79a SW |
2569 | new_active = 0; |
2570 | } | |
9a0d4dd3 | 2571 | |
df31a989 | 2572 | if (new_active) |
377e29f7 | 2573 | counter++; |
9a0d4dd3 | 2574 | |
5530d55d | 2575 | /* Check for changes to the nexthop - set ROUTE_ENTRY_CHANGED */ |
ad28e79a SW |
2576 | if (prev_active != new_active || prev_index != nexthop->ifindex |
2577 | || ((nexthop->type >= NEXTHOP_TYPE_IFINDEX | |
2578 | && nexthop->type < NEXTHOP_TYPE_IPV6) | |
2579 | && prev_src.ipv4.s_addr | |
2580 | != nexthop->rmap_src.ipv4.s_addr) | |
2581 | || ((nexthop->type >= NEXTHOP_TYPE_IPV6 | |
2582 | && nexthop->type < NEXTHOP_TYPE_BLACKHOLE) | |
2583 | && !(IPV6_ADDR_SAME(&prev_src.ipv6, | |
2584 | &nexthop->rmap_src.ipv6))) | |
8a507796 | 2585 | || CHECK_FLAG(re->status, ROUTE_ENTRY_LABELS_CHANGED)) |
ad28e79a | 2586 | SET_FLAG(re->status, ROUTE_ENTRY_CHANGED); |
ad28e79a SW |
2587 | } |
2588 | ||
377e29f7 MS |
2589 | return counter; |
2590 | } | |
2591 | ||
2c41ef8c SW |
2592 | |
2593 | static uint32_t proto_nhg_nexthop_active_update(struct nexthop_group *nhg) | |
2594 | { | |
2595 | struct nexthop *nh; | |
2596 | uint32_t curr_active = 0; | |
2597 | ||
2598 | /* Assume all active for now */ | |
2599 | ||
2600 | for (nh = nhg->nexthop; nh; nh = nh->next) { | |
2601 | SET_FLAG(nh->flags, NEXTHOP_FLAG_ACTIVE); | |
2602 | curr_active++; | |
2603 | } | |
2604 | ||
2605 | return curr_active; | |
2606 | } | |
2607 | ||
377e29f7 MS |
2608 | /* |
2609 | * Iterate over all nexthops of the given RIB entry and refresh their | |
2610 | * ACTIVE flag. If any nexthop is found to toggle the ACTIVE flag, | |
2611 | * the whole re structure is flagged with ROUTE_ENTRY_CHANGED. | |
2612 | * | |
2613 | * Return value is the new number of active nexthops. | |
2614 | */ | |
2615 | int nexthop_active_update(struct route_node *rn, struct route_entry *re) | |
2616 | { | |
2617 | struct nhg_hash_entry *curr_nhe; | |
2618 | uint32_t curr_active = 0, backup_active = 0; | |
2619 | ||
65f137fe | 2620 | if (PROTO_OWNED(re->nhe)) |
2c41ef8c SW |
2621 | return proto_nhg_nexthop_active_update(&re->nhe->nhg); |
2622 | ||
377e29f7 MS |
2623 | afi_t rt_afi = family2afi(rn->p.family); |
2624 | ||
2625 | UNSET_FLAG(re->status, ROUTE_ENTRY_CHANGED); | |
2626 | ||
2627 | /* Make a local copy of the existing nhe, so we don't work on/modify | |
2628 | * the shared nhe. | |
2629 | */ | |
f727646a | 2630 | curr_nhe = zebra_nhe_copy(re->nhe, re->nhe->id); |
377e29f7 MS |
2631 | |
2632 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
2633 | zlog_debug("%s: re %p nhe %p (%u), curr_nhe %p", | |
2634 | __func__, re, re->nhe, re->nhe->id, | |
2635 | curr_nhe); | |
2636 | ||
2637 | /* Clear the existing id, if any: this will avoid any confusion | |
2638 | * if the id exists, and will also force the creation | |
2639 | * of a new nhe reflecting the changes we may make in this local copy. | |
2640 | */ | |
2641 | curr_nhe->id = 0; | |
2642 | ||
2643 | /* Process nexthops */ | |
5530d55d | 2644 | curr_active = nexthop_list_active_update(rn, re, curr_nhe, false); |
377e29f7 MS |
2645 | |
2646 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
2647 | zlog_debug("%s: re %p curr_active %u", __func__, re, | |
2648 | curr_active); | |
2649 | ||
2650 | /* If there are no backup nexthops, we are done */ | |
2651 | if (zebra_nhg_get_backup_nhg(curr_nhe) == NULL) | |
2652 | goto backups_done; | |
2653 | ||
2654 | backup_active = nexthop_list_active_update( | |
5530d55d | 2655 | rn, re, curr_nhe->backup_info->nhe, true /*is_backup*/); |
377e29f7 MS |
2656 | |
2657 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
2658 | zlog_debug("%s: re %p backup_active %u", __func__, re, | |
2659 | backup_active); | |
2660 | ||
2661 | backups_done: | |
2662 | ||
2663 | /* | |
2664 | * Ref or create an nhe that matches the current state of the | |
2665 | * nexthop(s). | |
2666 | */ | |
8a507796 | 2667 | if (CHECK_FLAG(re->status, ROUTE_ENTRY_CHANGED)) { |
98cda54a | 2668 | struct nhg_hash_entry *new_nhe = NULL; |
98cda54a | 2669 | |
377e29f7 MS |
2670 | new_nhe = zebra_nhg_rib_find_nhe(curr_nhe, rt_afi); |
2671 | ||
2672 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
2673 | zlog_debug("%s: re %p CHANGED: nhe %p (%u) => new_nhe %p (%u)", | |
2674 | __func__, re, re->nhe, | |
2675 | re->nhe->id, new_nhe, new_nhe->id); | |
98cda54a | 2676 | |
5463ce26 | 2677 | route_entry_update_nhe(re, new_nhe); |
e22e8001 SW |
2678 | } |
2679 | ||
377e29f7 | 2680 | |
086e4e02 SW |
2681 | /* Walk the NHE depends tree and toggle NEXTHOP_GROUP_VALID |
2682 | * flag where appropriate. | |
2683 | */ | |
715e5c70 | 2684 | if (curr_active) |
086e4e02 | 2685 | zebra_nhg_set_valid_if_active(re->nhe); |
98cda54a SW |
2686 | |
2687 | /* | |
377e29f7 MS |
2688 | * Do not need the old / copied nhe anymore since it |
2689 | * was either copied over into a new nhe or not | |
98cda54a SW |
2690 | * used at all. |
2691 | */ | |
377e29f7 | 2692 | zebra_nhg_free(curr_nhe); |
9a0d4dd3 | 2693 | return curr_active; |
ad28e79a | 2694 | } |
5be96a2d | 2695 | |
497ff579 SW |
2696 | /* Recursively construct a grp array of fully resolved IDs. |
2697 | * | |
2698 | * This function allows us to account for groups within groups, | |
2699 | * by converting them into a flat array of IDs. | |
2700 | * | |
2701 | * nh_grp is modified at every level of recursion to append | |
2702 | * to it the next unique, fully resolved ID from the entire tree. | |
2703 | * | |
2704 | * | |
2705 | * Note: | |
2706 | * I'm pretty sure we only allow ONE level of group within group currently. | |
2707 | * But making this recursive just in case that ever changes. | |
2708 | */ | |
2709 | static uint8_t zebra_nhg_nhe2grp_internal(struct nh_grp *grp, | |
2710 | uint8_t curr_index, | |
2711 | struct nhg_hash_entry *nhe, | |
2712 | int max_num) | |
98cda54a SW |
2713 | { |
2714 | struct nhg_connected *rb_node_dep = NULL; | |
2715 | struct nhg_hash_entry *depend = NULL; | |
497ff579 | 2716 | uint8_t i = curr_index; |
98cda54a | 2717 | |
fec211ad | 2718 | frr_each(nhg_connected_tree, &nhe->nhg_depends, rb_node_dep) { |
8dbc800f SW |
2719 | bool duplicate = false; |
2720 | ||
497ff579 SW |
2721 | if (i >= max_num) |
2722 | goto done; | |
2723 | ||
98cda54a SW |
2724 | depend = rb_node_dep->nhe; |
2725 | ||
2726 | /* | |
2727 | * If its recursive, use its resolved nhe in the group | |
2728 | */ | |
2729 | if (CHECK_FLAG(depend->flags, NEXTHOP_GROUP_RECURSIVE)) { | |
2730 | depend = zebra_nhg_resolve(depend); | |
2731 | if (!depend) { | |
2732 | flog_err( | |
2733 | EC_ZEBRA_NHG_FIB_UPDATE, | |
df31a989 SW |
2734 | "Failed to recursively resolve Nexthop Hash Entry in the group id=%u", |
2735 | nhe->id); | |
98cda54a SW |
2736 | continue; |
2737 | } | |
2738 | } | |
2739 | ||
497ff579 SW |
2740 | if (!zebra_nhg_depends_is_empty(depend)) { |
2741 | /* This is a group within a group */ | |
2742 | i = zebra_nhg_nhe2grp_internal(grp, i, depend, max_num); | |
2743 | } else { | |
086e4e02 SW |
2744 | if (!CHECK_FLAG(depend->flags, NEXTHOP_GROUP_VALID)) { |
2745 | if (IS_ZEBRA_DEBUG_RIB_DETAILED | |
2746 | || IS_ZEBRA_DEBUG_NHG) | |
2747 | zlog_debug( | |
2748 | "%s: Nexthop ID (%u) not valid, not appending to dataplane install group", | |
2749 | __func__, depend->id); | |
2750 | continue; | |
2751 | } | |
2752 | ||
1866b3af SW |
2753 | /* If the nexthop not installed/queued for install don't |
2754 | * put in the ID array. | |
2755 | */ | |
2756 | if (!(CHECK_FLAG(depend->flags, NEXTHOP_GROUP_INSTALLED) | |
2757 | || CHECK_FLAG(depend->flags, | |
2758 | NEXTHOP_GROUP_QUEUED))) { | |
2759 | if (IS_ZEBRA_DEBUG_RIB_DETAILED | |
2760 | || IS_ZEBRA_DEBUG_NHG) | |
2761 | zlog_debug( | |
2762 | "%s: Nexthop ID (%u) not installed or queued for install, not appending to dataplane install group", | |
2763 | __func__, depend->id); | |
2764 | continue; | |
2765 | } | |
2766 | ||
b1c3f7ef | 2767 | /* Check for duplicate IDs, ignore if found. */ |
497ff579 | 2768 | for (int j = 0; j < i; j++) { |
d43122b5 | 2769 | if (depend->id == grp[j].id) { |
497ff579 | 2770 | duplicate = true; |
d43122b5 SW |
2771 | break; |
2772 | } | |
497ff579 | 2773 | } |
8dbc800f | 2774 | |
b1c3f7ef SW |
2775 | if (duplicate) { |
2776 | if (IS_ZEBRA_DEBUG_RIB_DETAILED | |
2777 | || IS_ZEBRA_DEBUG_NHG) | |
2778 | zlog_debug( | |
2779 | "%s: Nexthop ID (%u) is duplicate, not appending to dataplane install group", | |
2780 | __func__, depend->id); | |
2781 | continue; | |
497ff579 | 2782 | } |
b1c3f7ef SW |
2783 | |
2784 | grp[i].id = depend->id; | |
2785 | grp[i].weight = depend->nhg.nexthop->weight; | |
2786 | i++; | |
8dbc800f | 2787 | } |
98cda54a | 2788 | } |
8dbc800f | 2789 | |
0328a5bd MS |
2790 | if (nhe->backup_info == NULL || nhe->backup_info->nhe == NULL) |
2791 | goto done; | |
2792 | ||
2793 | /* TODO -- For now, we are not trying to use or install any | |
2794 | * backup info in this nexthop-id path: we aren't prepared | |
2795 | * to use the backups here yet. We're just debugging what we find. | |
2796 | */ | |
2797 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
2798 | zlog_debug("%s: skipping backup nhe", __func__); | |
2799 | ||
8dbc800f | 2800 | done: |
98cda54a SW |
2801 | return i; |
2802 | } | |
2803 | ||
497ff579 SW |
2804 | /* Convert a nhe into a group array */ |
2805 | uint8_t zebra_nhg_nhe2grp(struct nh_grp *grp, struct nhg_hash_entry *nhe, | |
2806 | int max_num) | |
2807 | { | |
2808 | /* Call into the recursive function */ | |
2809 | return zebra_nhg_nhe2grp_internal(grp, 0, nhe, max_num); | |
2810 | } | |
2811 | ||
5be96a2d SW |
2812 | void zebra_nhg_install_kernel(struct nhg_hash_entry *nhe) |
2813 | { | |
f429bd1b SW |
2814 | struct nhg_connected *rb_node_dep = NULL; |
2815 | ||
2816 | /* Resolve it first */ | |
2817 | nhe = zebra_nhg_resolve(nhe); | |
2818 | ||
2819 | /* Make sure all depends are installed/queued */ | |
fec211ad | 2820 | frr_each(nhg_connected_tree, &nhe->nhg_depends, rb_node_dep) { |
f429bd1b SW |
2821 | zebra_nhg_install_kernel(rb_node_dep->nhe); |
2822 | } | |
2823 | ||
086e4e02 | 2824 | if (CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_VALID) |
e3b9c0f2 | 2825 | && !CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED) |
e22e8001 | 2826 | && !CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_QUEUED)) { |
724583ed | 2827 | /* Change its type to us since we are installing it */ |
0885b1e3 SW |
2828 | if (!ZEBRA_NHG_CREATED(nhe)) |
2829 | nhe->type = ZEBRA_ROUTE_NHG; | |
724583ed | 2830 | |
147bad16 | 2831 | int ret = dplane_nexthop_add(nhe); |
2d3c57e6 | 2832 | |
147bad16 SW |
2833 | switch (ret) { |
2834 | case ZEBRA_DPLANE_REQUEST_QUEUED: | |
2835 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_QUEUED); | |
2836 | break; | |
2837 | case ZEBRA_DPLANE_REQUEST_FAILURE: | |
2838 | flog_err( | |
2839 | EC_ZEBRA_DP_INSTALL_FAIL, | |
2840 | "Failed to install Nexthop ID (%u) into the kernel", | |
2841 | nhe->id); | |
2842 | break; | |
2843 | case ZEBRA_DPLANE_REQUEST_SUCCESS: | |
2844 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED); | |
80286aa5 | 2845 | zebra_nhg_handle_install(nhe); |
147bad16 SW |
2846 | break; |
2847 | } | |
2848 | } | |
2849 | } | |
2850 | ||
147bad16 SW |
2851 | void zebra_nhg_uninstall_kernel(struct nhg_hash_entry *nhe) |
2852 | { | |
2853 | if (CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED)) { | |
2854 | int ret = dplane_nexthop_delete(nhe); | |
2d3c57e6 | 2855 | |
147bad16 SW |
2856 | switch (ret) { |
2857 | case ZEBRA_DPLANE_REQUEST_QUEUED: | |
2858 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_QUEUED); | |
2859 | break; | |
2860 | case ZEBRA_DPLANE_REQUEST_FAILURE: | |
2861 | flog_err( | |
2862 | EC_ZEBRA_DP_DELETE_FAIL, | |
2863 | "Failed to uninstall Nexthop ID (%u) from the kernel", | |
2864 | nhe->id); | |
2865 | break; | |
2866 | case ZEBRA_DPLANE_REQUEST_SUCCESS: | |
2867 | UNSET_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED); | |
2868 | break; | |
2869 | } | |
177e711d SW |
2870 | } |
2871 | ||
2872 | zebra_nhg_handle_uninstall(nhe); | |
147bad16 SW |
2873 | } |
2874 | ||
5f3c9e52 SW |
2875 | void zebra_nhg_dplane_result(struct zebra_dplane_ctx *ctx) |
2876 | { | |
2877 | enum dplane_op_e op; | |
2878 | enum zebra_dplane_result status; | |
2879 | uint32_t id = 0; | |
2880 | struct nhg_hash_entry *nhe = NULL; | |
2881 | ||
2882 | op = dplane_ctx_get_op(ctx); | |
2883 | status = dplane_ctx_get_status(ctx); | |
2884 | ||
0c8215cb | 2885 | id = dplane_ctx_get_nhe_id(ctx); |
e22e8001 | 2886 | |
377e29f7 | 2887 | if (IS_ZEBRA_DEBUG_DPLANE_DETAIL || IS_ZEBRA_DEBUG_NHG_DETAIL) |
177e711d SW |
2888 | zlog_debug( |
2889 | "Nexthop dplane ctx %p, op %s, nexthop ID (%u), result %s", | |
2890 | ctx, dplane_op2str(op), id, dplane_res2str(status)); | |
5f3c9e52 | 2891 | |
177e711d SW |
2892 | switch (op) { |
2893 | case DPLANE_OP_NH_DELETE: | |
2894 | if (status != ZEBRA_DPLANE_REQUEST_SUCCESS) | |
2895 | flog_err( | |
2896 | EC_ZEBRA_DP_DELETE_FAIL, | |
2897 | "Failed to uninstall Nexthop ID (%u) from the kernel", | |
2898 | id); | |
ee94437e | 2899 | |
177e711d SW |
2900 | /* We already free'd the data, nothing to do */ |
2901 | break; | |
2902 | case DPLANE_OP_NH_INSTALL: | |
2903 | case DPLANE_OP_NH_UPDATE: | |
2904 | nhe = zebra_nhg_lookup_id(id); | |
2905 | ||
2906 | if (!nhe) { | |
30672034 SW |
2907 | if (IS_ZEBRA_DEBUG_NHG) |
2908 | zlog_debug( | |
2909 | "%s operation preformed on Nexthop ID (%u) in the kernel, that we no longer have in our table", | |
2910 | dplane_op2str(op), id); | |
2911 | ||
5f3c9e52 SW |
2912 | break; |
2913 | } | |
177e711d SW |
2914 | |
2915 | UNSET_FLAG(nhe->flags, NEXTHOP_GROUP_QUEUED); | |
2916 | if (status == ZEBRA_DPLANE_REQUEST_SUCCESS) { | |
2917 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_VALID); | |
2918 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED); | |
80286aa5 | 2919 | zebra_nhg_handle_install(nhe); |
ee94437e MS |
2920 | |
2921 | /* If daemon nhg, send it an update */ | |
65f137fe | 2922 | if (PROTO_OWNED(nhe)) |
ee94437e MS |
2923 | zsend_nhg_notify(nhe->type, nhe->zapi_instance, |
2924 | nhe->zapi_session, nhe->id, | |
2925 | ZAPI_NHG_INSTALLED); | |
2926 | } else { | |
2927 | /* If daemon nhg, send it an update */ | |
65f137fe | 2928 | if (PROTO_OWNED(nhe)) |
ee94437e MS |
2929 | zsend_nhg_notify(nhe->type, nhe->zapi_instance, |
2930 | nhe->zapi_session, nhe->id, | |
2931 | ZAPI_NHG_FAIL_INSTALL); | |
2932 | ||
177e711d SW |
2933 | flog_err( |
2934 | EC_ZEBRA_DP_INSTALL_FAIL, | |
2935 | "Failed to install Nexthop ID (%u) into the kernel", | |
2936 | nhe->id); | |
ee94437e | 2937 | } |
177e711d | 2938 | break; |
ee94437e | 2939 | |
177e711d SW |
2940 | case DPLANE_OP_ROUTE_INSTALL: |
2941 | case DPLANE_OP_ROUTE_UPDATE: | |
2942 | case DPLANE_OP_ROUTE_DELETE: | |
2943 | case DPLANE_OP_ROUTE_NOTIFY: | |
2944 | case DPLANE_OP_LSP_INSTALL: | |
2945 | case DPLANE_OP_LSP_UPDATE: | |
2946 | case DPLANE_OP_LSP_DELETE: | |
2947 | case DPLANE_OP_LSP_NOTIFY: | |
2948 | case DPLANE_OP_PW_INSTALL: | |
2949 | case DPLANE_OP_PW_UNINSTALL: | |
2950 | case DPLANE_OP_SYS_ROUTE_ADD: | |
2951 | case DPLANE_OP_SYS_ROUTE_DELETE: | |
2952 | case DPLANE_OP_ADDR_INSTALL: | |
2953 | case DPLANE_OP_ADDR_UNINSTALL: | |
2954 | case DPLANE_OP_MAC_INSTALL: | |
2955 | case DPLANE_OP_MAC_DELETE: | |
2956 | case DPLANE_OP_NEIGH_INSTALL: | |
2957 | case DPLANE_OP_NEIGH_UPDATE: | |
2958 | case DPLANE_OP_NEIGH_DELETE: | |
0a27a2fe PG |
2959 | case DPLANE_OP_NEIGH_IP_INSTALL: |
2960 | case DPLANE_OP_NEIGH_IP_DELETE: | |
177e711d SW |
2961 | case DPLANE_OP_VTEP_ADD: |
2962 | case DPLANE_OP_VTEP_DELETE: | |
60d8d43b JU |
2963 | case DPLANE_OP_RULE_ADD: |
2964 | case DPLANE_OP_RULE_DELETE: | |
2965 | case DPLANE_OP_RULE_UPDATE: | |
d68e74b4 | 2966 | case DPLANE_OP_NEIGH_DISCOVER: |
c60522f7 | 2967 | case DPLANE_OP_BR_PORT_UPDATE: |
177e711d | 2968 | case DPLANE_OP_NONE: |
5162e000 PG |
2969 | case DPLANE_OP_IPTABLE_ADD: |
2970 | case DPLANE_OP_IPTABLE_DELETE: | |
ef524230 PG |
2971 | case DPLANE_OP_IPSET_ADD: |
2972 | case DPLANE_OP_IPSET_DELETE: | |
2973 | case DPLANE_OP_IPSET_ENTRY_ADD: | |
2974 | case DPLANE_OP_IPSET_ENTRY_DELETE: | |
e18747a9 | 2975 | case DPLANE_OP_NEIGH_TABLE_UPDATE: |
62b4b7e4 | 2976 | case DPLANE_OP_GRE_SET: |
177e711d SW |
2977 | break; |
2978 | } | |
71593b3f SW |
2979 | |
2980 | dplane_ctx_fini(&ctx); | |
5be96a2d SW |
2981 | } |
2982 | ||
38e40db1 SW |
2983 | static void zebra_nhg_sweep_entry(struct hash_bucket *bucket, void *arg) |
2984 | { | |
2985 | struct nhg_hash_entry *nhe = NULL; | |
2986 | ||
2987 | nhe = (struct nhg_hash_entry *)bucket->data; | |
2988 | ||
dc65cd99 SW |
2989 | /* |
2990 | * same logic as with routes. | |
2991 | * | |
2992 | * If older than startup time, we know we read them in from the | |
2993 | * kernel and have not gotten and update for them since startup | |
2994 | * from an upper level proto. | |
2995 | */ | |
2996 | if (zrouter.startup_time < nhe->uptime) | |
2997 | return; | |
2998 | ||
2999 | /* | |
3000 | * If it's proto-owned and not being used by a route, remove it since | |
3001 | * we haven't gotten an update about it from the proto since startup. | |
3002 | * This means that either the config for it was removed or the daemon | |
3003 | * didn't get started. This handles graceful restart & retain scenario. | |
3004 | */ | |
3005 | if (PROTO_OWNED(nhe) && nhe->refcnt == 1) { | |
3006 | zebra_nhg_decrement_ref(nhe); | |
3007 | return; | |
3008 | } | |
3009 | ||
3010 | /* | |
3011 | * If its being ref'd by routes, just let it be uninstalled via a route | |
3012 | * removal. | |
3013 | */ | |
38e40db1 SW |
3014 | if (ZEBRA_NHG_CREATED(nhe) && nhe->refcnt <= 0) |
3015 | zebra_nhg_uninstall_kernel(nhe); | |
3016 | } | |
3017 | ||
3018 | void zebra_nhg_sweep_table(struct hash *hash) | |
3019 | { | |
3020 | hash_iterate(hash, zebra_nhg_sweep_entry, NULL); | |
3021 | } | |
7c99d51b | 3022 | |
b1b07ef5 DS |
3023 | static void zebra_nhg_mark_keep_entry(struct hash_bucket *bucket, void *arg) |
3024 | { | |
3025 | struct nhg_hash_entry *nhe = bucket->data; | |
3026 | ||
3027 | UNSET_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED); | |
3028 | } | |
3029 | ||
3030 | /* | |
3031 | * When we are shutting down and we have retain mode enabled | |
3032 | * in zebra the process is to mark each vrf that it's | |
3033 | * routes should not be deleted. The problem with that | |
3034 | * is that shutdown actually free's up memory which | |
3035 | * causes the nexthop group's ref counts to go to zero | |
3036 | * we need a way to subtly tell the system to not remove | |
3037 | * the nexthop groups from the kernel at the same time. | |
3038 | * The easiest just looks like that we should not mark | |
3039 | * the nhg's as installed any more and when the ref count | |
3040 | * goes to zero we'll attempt to delete and do nothing | |
3041 | */ | |
3042 | void zebra_nhg_mark_keep(void) | |
3043 | { | |
3044 | hash_iterate(zrouter.nhgs_id, zebra_nhg_mark_keep_entry, NULL); | |
3045 | } | |
3046 | ||
7c99d51b MS |
3047 | /* Global control to disable use of kernel nexthops, if available. We can't |
3048 | * force the kernel to support nexthop ids, of course, but we can disable | |
3049 | * zebra's use of them, for testing e.g. By default, if the kernel supports | |
3050 | * nexthop ids, zebra uses them. | |
3051 | */ | |
3052 | void zebra_nhg_enable_kernel_nexthops(bool set) | |
3053 | { | |
3054 | g_nexthops_enabled = set; | |
3055 | } | |
3056 | ||
3057 | bool zebra_nhg_kernel_nexthops_enabled(void) | |
3058 | { | |
3059 | return g_nexthops_enabled; | |
3060 | } | |
0885b1e3 | 3061 | |
aa458838 MS |
3062 | /* Global control for use of activated backups for recursive resolution. */ |
3063 | void zebra_nhg_set_recursive_use_backups(bool set) | |
3064 | { | |
3065 | use_recursive_backups = set; | |
3066 | } | |
3067 | ||
3068 | bool zebra_nhg_recursive_use_backups(void) | |
3069 | { | |
3070 | return use_recursive_backups; | |
3071 | } | |
3072 | ||
6c67f41f SW |
3073 | /* |
3074 | * Global control to only use kernel nexthops for protocol created NHGs. | |
3075 | * There are some use cases where you may not want zebra to implicitly | |
3076 | * create kernel nexthops for all routes and only create them for NHGs | |
3077 | * passed down by upper level protos. | |
3078 | * | |
3079 | * Default is off. | |
3080 | */ | |
3081 | void zebra_nhg_set_proto_nexthops_only(bool set) | |
3082 | { | |
3083 | proto_nexthops_only = set; | |
3084 | } | |
3085 | ||
3086 | bool zebra_nhg_proto_nexthops_only(void) | |
3087 | { | |
3088 | return proto_nexthops_only; | |
3089 | } | |
3090 | ||
0885b1e3 SW |
3091 | /* Add NHE from upper level proto */ |
3092 | struct nhg_hash_entry *zebra_nhg_proto_add(uint32_t id, int type, | |
ee94437e | 3093 | uint16_t instance, uint32_t session, |
0885b1e3 SW |
3094 | struct nexthop_group *nhg, afi_t afi) |
3095 | { | |
3096 | struct nhg_hash_entry lookup; | |
dd1e105f | 3097 | struct nhg_hash_entry *new, *old; |
0885b1e3 | 3098 | struct nhg_connected *rb_node_dep = NULL; |
1f655680 SW |
3099 | struct nexthop *newhop; |
3100 | bool replace = false; | |
3101 | ||
3102 | if (!nhg->nexthop) { | |
3103 | if (IS_ZEBRA_DEBUG_NHG) | |
3104 | zlog_debug("%s: id %u, no nexthops passed to add", | |
3105 | __func__, id); | |
3106 | return NULL; | |
3107 | } | |
3108 | ||
3109 | ||
3110 | /* Set nexthop list as active, since they wont go through rib | |
3111 | * processing. | |
3112 | * | |
3113 | * Assuming valid/onlink for now. | |
3114 | * | |
3115 | * Once resolution is figured out, we won't need this! | |
3116 | */ | |
70347b7a | 3117 | for (ALL_NEXTHOPS_PTR(nhg, newhop)) { |
21735352 SW |
3118 | if (CHECK_FLAG(newhop->flags, NEXTHOP_FLAG_HAS_BACKUP)) { |
3119 | if (IS_ZEBRA_DEBUG_NHG) | |
3120 | zlog_debug( | |
3121 | "%s: id %u, backup nexthops not supported", | |
3122 | __func__, id); | |
3123 | return NULL; | |
3124 | } | |
3125 | ||
70f3cda6 SW |
3126 | if (newhop->type == NEXTHOP_TYPE_BLACKHOLE) { |
3127 | if (IS_ZEBRA_DEBUG_NHG) | |
3128 | zlog_debug( | |
3129 | "%s: id %u, blackhole nexthop not supported", | |
3130 | __func__, id); | |
3131 | return NULL; | |
3132 | } | |
3133 | ||
3134 | if (newhop->type == NEXTHOP_TYPE_IFINDEX) { | |
3135 | if (IS_ZEBRA_DEBUG_NHG) | |
3136 | zlog_debug( | |
3137 | "%s: id %u, nexthop without gateway not supported", | |
3138 | __func__, id); | |
3139 | return NULL; | |
3140 | } | |
3141 | ||
70347b7a SW |
3142 | if (!newhop->ifindex) { |
3143 | if (IS_ZEBRA_DEBUG_NHG) | |
3144 | zlog_debug( | |
70f3cda6 | 3145 | "%s: id %u, nexthop without ifindex is not supported", |
70347b7a SW |
3146 | __func__, id); |
3147 | return NULL; | |
3148 | } | |
1f655680 | 3149 | SET_FLAG(newhop->flags, NEXTHOP_FLAG_ACTIVE); |
70347b7a | 3150 | } |
0885b1e3 SW |
3151 | |
3152 | zebra_nhe_init(&lookup, afi, nhg->nexthop); | |
3153 | lookup.nhg.nexthop = nhg->nexthop; | |
3154 | lookup.id = id; | |
3155 | lookup.type = type; | |
3156 | ||
dd1e105f SW |
3157 | old = zebra_nhg_lookup_id(id); |
3158 | ||
3159 | if (old) { | |
3160 | /* | |
3161 | * This is a replace, just release NHE from ID for now, The | |
84591282 SW |
3162 | * depends/dependents may still be used in the replacement so |
3163 | * we don't touch them other than to remove their refs to their | |
3164 | * old parent. | |
dd1e105f | 3165 | */ |
1f655680 | 3166 | replace = true; |
dd1e105f | 3167 | hash_release(zrouter.nhgs_id, old); |
84591282 SW |
3168 | |
3169 | /* Free all the things */ | |
3170 | zebra_nhg_release_all_deps(old); | |
dd1e105f SW |
3171 | } |
3172 | ||
0885b1e3 SW |
3173 | new = zebra_nhg_rib_find_nhe(&lookup, afi); |
3174 | ||
1f655680 SW |
3175 | zebra_nhg_increment_ref(new); |
3176 | ||
ee94437e MS |
3177 | /* Capture zapi client info */ |
3178 | new->zapi_instance = instance; | |
3179 | new->zapi_session = session; | |
3180 | ||
1f655680 SW |
3181 | zebra_nhg_set_valid_if_active(new); |
3182 | ||
3183 | zebra_nhg_install_kernel(new); | |
3184 | ||
dd1e105f | 3185 | if (old) { |
8155e8c5 SW |
3186 | /* |
3187 | * Check to handle recving DEL while routes still in use then | |
3188 | * a replace. | |
3189 | * | |
3190 | * In this case we would have decremented the refcnt already | |
3191 | * but set the FLAG here. Go ahead and increment once to fix | |
3192 | * the misordering we have been sent. | |
3193 | */ | |
3194 | if (CHECK_FLAG(old->flags, NEXTHOP_GROUP_PROTO_RELEASED)) | |
3195 | zebra_nhg_increment_ref(old); | |
3196 | ||
1f655680 | 3197 | rib_handle_nhg_replace(old, new); |
dd1e105f | 3198 | |
1f655680 SW |
3199 | /* if this != 1 at this point, we have a bug */ |
3200 | assert(old->refcnt == 1); | |
0885b1e3 | 3201 | |
1f655680 SW |
3202 | /* We have to decrement its singletons |
3203 | * because some might not exist in NEW. | |
3204 | */ | |
3205 | if (!zebra_nhg_depends_is_empty(old)) { | |
3206 | frr_each (nhg_connected_tree, &old->nhg_depends, | |
3207 | rb_node_dep) | |
3208 | zebra_nhg_decrement_ref(rb_node_dep->nhe); | |
3209 | } | |
3210 | ||
1f655680 SW |
3211 | /* Dont call the dec API, we dont want to uninstall the ID */ |
3212 | old->refcnt = 0; | |
3213 | zebra_nhg_free(old); | |
3214 | old = NULL; | |
0885b1e3 SW |
3215 | } |
3216 | ||
3217 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
dd1e105f | 3218 | zlog_debug("%s: %s nhe %p (%u), vrf %d, type %s", __func__, |
1f655680 | 3219 | (replace ? "replaced" : "added"), new, new->id, |
dd1e105f | 3220 | new->vrf_id, zebra_route_string(new->type)); |
0885b1e3 SW |
3221 | |
3222 | return new; | |
3223 | } | |
3224 | ||
1f655680 | 3225 | /* Delete NHE from upper level proto, caller must decrement ref */ |
aaa42e05 | 3226 | struct nhg_hash_entry *zebra_nhg_proto_del(uint32_t id, int type) |
0885b1e3 SW |
3227 | { |
3228 | struct nhg_hash_entry *nhe; | |
3229 | ||
3230 | nhe = zebra_nhg_lookup_id(id); | |
3231 | ||
3232 | if (!nhe) { | |
8155e8c5 | 3233 | if (IS_ZEBRA_DEBUG_NHG) |
0885b1e3 SW |
3234 | zlog_debug("%s: id %u, lookup failed", __func__, id); |
3235 | ||
3236 | return NULL; | |
3237 | } | |
3238 | ||
aaa42e05 SW |
3239 | if (type != nhe->type) { |
3240 | if (IS_ZEBRA_DEBUG_NHG) | |
3241 | zlog_debug( | |
3242 | "%s: id %u, type %s mismatch, sent by %s, ignoring", | |
3243 | __func__, id, zebra_route_string(nhe->type), | |
3244 | zebra_route_string(type)); | |
3245 | return NULL; | |
3246 | } | |
3247 | ||
8155e8c5 SW |
3248 | if (CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_PROTO_RELEASED)) { |
3249 | if (IS_ZEBRA_DEBUG_NHG) | |
3250 | zlog_debug("%s: id %u, already released", __func__, id); | |
3251 | ||
3252 | return NULL; | |
3253 | } | |
3254 | ||
3255 | SET_FLAG(nhe->flags, NEXTHOP_GROUP_PROTO_RELEASED); | |
3256 | ||
1f655680 | 3257 | if (nhe->refcnt > 1) { |
0885b1e3 | 3258 | if (IS_ZEBRA_DEBUG_NHG) |
1f655680 SW |
3259 | zlog_debug( |
3260 | "%s: id %u, still being used by routes refcnt %u", | |
3261 | __func__, nhe->id, nhe->refcnt); | |
9c6c48bc | 3262 | return nhe; |
0885b1e3 SW |
3263 | } |
3264 | ||
3265 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) | |
3266 | zlog_debug("%s: deleted nhe %p (%u), vrf %d, type %s", __func__, | |
3267 | nhe, nhe->id, nhe->vrf_id, | |
3268 | zebra_route_string(nhe->type)); | |
3269 | ||
3270 | return nhe; | |
3271 | } | |
3272 | ||
24db1a7b SW |
3273 | struct nhg_score_proto_iter { |
3274 | int type; | |
8f830b8c | 3275 | struct list *found; |
24db1a7b SW |
3276 | }; |
3277 | ||
3278 | static void zebra_nhg_score_proto_entry(struct hash_bucket *bucket, void *arg) | |
3279 | { | |
3280 | struct nhg_hash_entry *nhe; | |
3281 | struct nhg_score_proto_iter *iter; | |
3282 | ||
3283 | nhe = (struct nhg_hash_entry *)bucket->data; | |
3284 | iter = arg; | |
3285 | ||
3286 | /* Needs to match type and outside zebra ID space */ | |
65f137fe | 3287 | if (nhe->type == iter->type && PROTO_OWNED(nhe)) { |
24db1a7b SW |
3288 | if (IS_ZEBRA_DEBUG_NHG_DETAIL) |
3289 | zlog_debug( | |
3290 | "%s: found nhe %p (%u), vrf %d, type %s after client disconnect", | |
3291 | __func__, nhe, nhe->id, nhe->vrf_id, | |
3292 | zebra_route_string(nhe->type)); | |
3293 | ||
8f830b8c SW |
3294 | /* Add to removal list */ |
3295 | listnode_add(iter->found, nhe); | |
24db1a7b SW |
3296 | } |
3297 | } | |
3298 | ||
3299 | /* Remove specific by proto NHGs */ | |
3300 | unsigned long zebra_nhg_score_proto(int type) | |
3301 | { | |
8f830b8c | 3302 | struct nhg_hash_entry *nhe; |
24db1a7b | 3303 | struct nhg_score_proto_iter iter = {}; |
8f830b8c SW |
3304 | struct listnode *ln; |
3305 | unsigned long count; | |
24db1a7b SW |
3306 | |
3307 | iter.type = type; | |
8f830b8c | 3308 | iter.found = list_new(); |
24db1a7b | 3309 | |
8f830b8c | 3310 | /* Find matching entries to remove */ |
24db1a7b SW |
3311 | hash_iterate(zrouter.nhgs_id, zebra_nhg_score_proto_entry, &iter); |
3312 | ||
8f830b8c SW |
3313 | /* Now remove them */ |
3314 | for (ALL_LIST_ELEMENTS_RO(iter.found, ln, nhe)) { | |
3315 | /* | |
3316 | * This should be the last ref if we remove client routes too, | |
3317 | * and thus should remove and free them. | |
3318 | */ | |
3319 | zebra_nhg_decrement_ref(nhe); | |
3320 | } | |
3321 | ||
3322 | count = iter.found->count; | |
3323 | list_delete(&iter.found); | |
3324 | ||
3325 | return count; | |
24db1a7b | 3326 | } |