3 * Copyright 2009-2016, LabN Consulting, L.L.C.
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version 2
9 * of the License, or (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License along
17 * with this program; see the file COPYING; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
23 * Purpose: maintain per-nve ribs and generate change lists
28 #include "lib/zebra.h"
29 #include "lib/prefix.h"
30 #include "lib/table.h"
32 #include "lib/memory.h"
34 #include "lib/skiplist.h"
35 #include "lib/workqueue.h"
37 #include "bgpd/bgpd.h"
38 #include "bgpd/bgp_route.h"
39 #include "bgpd/bgp_ecommunity.h"
40 #include "bgpd/bgp_mplsvpn.h"
41 #include "bgpd/bgp_vnc_types.h"
43 #include "bgpd/rfapi/rfapi.h"
44 #include "bgpd/rfapi/bgp_rfapi_cfg.h"
45 #include "bgpd/rfapi/rfapi_import.h"
46 #include "bgpd/rfapi/rfapi_private.h"
47 #include "bgpd/rfapi/rfapi_vty.h"
48 #include "bgpd/rfapi/vnc_import_bgp.h"
49 #include "bgpd/rfapi/rfapi_rib.h"
50 #include "bgpd/rfapi/rfapi_monitor.h"
51 #include "bgpd/rfapi/rfapi_encap_tlv.h"
52 #include "bgpd/rfapi/vnc_debug.h"
54 #define DEBUG_PROCESS_PENDING_NODE 0
55 #define DEBUG_PENDING_DELETE_ROUTE 0
57 #define DEBUG_RIB_SL_RD 0
61 static void rfapiRibShowRibSl(void *stream
, struct prefix
*pfx
,
68 * Model of the set of routes currently in the NVE's RIB.
70 * node->info ptr to "struct skiplist".
71 * MUST be NULL if there are no routes.
72 * key = ptr to struct prefix {vn}
73 * val = ptr to struct rfapi_info
75 * skiplist.cmp = vnc_prefix_cmp
77 * node->aggregate ptr to "struct skiplist".
78 * key = ptr to struct prefix {vn}
79 * val = ptr to struct rfapi_info
80 * skiplist.del = rfapi_info_free
81 * skiplist.cmp = vnc_prefix_cmp
83 * This skiplist at "aggregate"
84 * contains the routes recently
90 * Sparse list of prefixes that need to be updated. Each node
91 * will have the complete set of routes for the prefix.
93 * node->info ptr to "struct list" (lib/linklist.h)
95 * List of routes sorted lowest cost first.
96 * This list is how the new complete set
97 * of routes should look.
98 * Set if there are updates to the prefix;
99 * MUST be NULL if there are no updates.
101 * .data = ptr to struct rfapi_info
102 * list.cmp = NULL (sorted manually)
103 * list.del = rfapi_info_free
105 * Special case: if node->info is 1, it means
106 * "delete all routes at this prefix".
108 * node->aggregate ptr to struct skiplist
109 * key = ptr to struct prefix {vn} (part of ri)
110 * val = struct rfapi_info
111 * skiplist.cmp = vnc_prefix_cmp
112 * skiplist.del = NULL
114 * ptlist is rewritten anew each time
115 * rfapiRibUpdatePendingNode() is called
117 * THE ptlist VALUES ARE REFERENCES TO THE
118 * rfapi_info STRUCTS IN THE node->info LIST.
122 * iterate over RIB to count responses, compare with running counters
124 void rfapiRibCheckCounts(
125 int checkstats
, /* validate rfd & global counts */
126 unsigned int offset
) /* number of ri's held separately */
128 struct rfapi_descriptor
*rfd
;
129 struct listnode
*node
;
131 struct bgp
*bgp
= bgp_get_default();
133 uint32_t t_pfx_active
= 0;
134 uint32_t t_pfx_deleted
= 0;
136 uint32_t t_ri_active
= 0;
137 uint32_t t_ri_deleted
= 0;
138 uint32_t t_ri_pend
= 0;
140 unsigned int alloc_count
;
145 for (ALL_LIST_ELEMENTS_RO(&bgp
->rfapi
->descriptors
, node
, rfd
)) {
148 uint32_t pfx_active
= 0;
149 uint32_t pfx_deleted
= 0;
151 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
153 struct route_node
*rn
;
155 for (rn
= route_top(rfd
->rib
[afi
]); rn
;
156 rn
= route_next(rn
)) {
158 struct skiplist
*sl
= rn
->info
;
159 struct skiplist
*dsl
= rn
->aggregate
;
160 uint32_t ri_active
= 0;
161 uint32_t ri_deleted
= 0;
164 ri_active
= skiplist_count(sl
);
166 t_ri_active
+= ri_active
;
172 ri_deleted
= skiplist_count(dsl
);
173 t_ri_deleted
+= ri_deleted
;
178 for (rn
= route_top(rfd
->rib_pending
[afi
]); rn
;
179 rn
= route_next(rn
)) {
181 struct list
*l
= rn
->info
; /* sorted by cost */
182 struct skiplist
*sl
= rn
->aggregate
;
183 uint32_t ri_pend_cost
= 0;
184 uint32_t ri_pend_uniq
= 0;
187 ri_pend_uniq
= skiplist_count(sl
);
190 if (l
&& (l
!= (void *)1)) {
191 ri_pend_cost
= l
->count
;
192 t_ri_pend
+= l
->count
;
195 assert(ri_pend_uniq
== ri_pend_cost
);
200 if (pfx_active
!= rfd
->rib_prefix_count
) {
201 vnc_zlog_debug_verbose(
202 "%s: rfd %p actual pfx count %u != running %u",
203 __func__
, rfd
, pfx_active
,
204 rfd
->rib_prefix_count
);
210 if (checkstats
&& bgp
->rfapi
) {
211 if (t_pfx_active
!= bgp
->rfapi
->rib_prefix_count_total
) {
212 vnc_zlog_debug_verbose(
213 "%s: actual total pfx count %u != running %u",
214 __func__
, t_pfx_active
,
215 bgp
->rfapi
->rib_prefix_count_total
);
221 * Check against memory allocation count
223 alloc_count
= mtype_stats_alloc(MTYPE_RFAPI_INFO
);
224 assert(t_ri_active
+ t_ri_deleted
+ t_ri_pend
+ offset
== alloc_count
);
227 static struct rfapi_info
*rfapi_info_new()
229 return XCALLOC(MTYPE_RFAPI_INFO
, sizeof(struct rfapi_info
));
232 void rfapiFreeRfapiUnOptionChain(struct rfapi_un_option
*p
)
235 struct rfapi_un_option
*next
;
238 XFREE(MTYPE_RFAPI_UN_OPTION
, p
);
243 void rfapiFreeRfapiVnOptionChain(struct rfapi_vn_option
*p
)
246 struct rfapi_vn_option
*next
;
249 XFREE(MTYPE_RFAPI_VN_OPTION
, p
);
255 static void rfapi_info_free(struct rfapi_info
*goner
)
258 if (goner
->tea_options
) {
259 rfapiFreeBgpTeaOptionChain(goner
->tea_options
);
260 goner
->tea_options
= NULL
;
262 if (goner
->un_options
) {
263 rfapiFreeRfapiUnOptionChain(goner
->un_options
);
264 goner
->un_options
= NULL
;
266 if (goner
->vn_options
) {
267 rfapiFreeRfapiVnOptionChain(goner
->vn_options
);
268 goner
->vn_options
= NULL
;
271 struct rfapi_rib_tcb
*tcb
;
273 tcb
= ((struct thread
*)goner
->timer
)->arg
;
274 thread_cancel((struct thread
*)goner
->timer
);
275 XFREE(MTYPE_RFAPI_RECENT_DELETE
, tcb
);
278 XFREE(MTYPE_RFAPI_INFO
, goner
);
283 * Timer control block for recently-deleted and expired routes
285 struct rfapi_rib_tcb
{
286 struct rfapi_descriptor
*rfd
;
288 struct rfapi_info
*ri
;
289 struct route_node
*rn
;
291 #define RFAPI_RIB_TCB_FLAG_DELETED 0x00000001
295 * remove route from rib
297 static int rfapiRibExpireTimer(struct thread
*t
)
299 struct rfapi_rib_tcb
*tcb
= t
->arg
;
301 RFAPI_RIB_CHECK_COUNTS(1, 0);
304 * Forget reference to thread. Otherwise rfapi_info_free() will
305 * attempt to free thread pointer as an option chain
307 tcb
->ri
->timer
= NULL
;
309 /* "deleted" skiplist frees ri, "active" doesn't */
310 assert(!skiplist_delete(tcb
->sl
, &tcb
->ri
->rk
, NULL
));
313 * XXX in this case, skiplist has no delete function: we must
314 * therefore delete rfapi_info explicitly.
316 rfapi_info_free(tcb
->ri
);
319 if (skiplist_empty(tcb
->sl
)) {
320 if (CHECK_FLAG(tcb
->flags
, RFAPI_RIB_TCB_FLAG_DELETED
))
321 tcb
->rn
->aggregate
= NULL
;
323 struct bgp
*bgp
= bgp_get_default();
324 tcb
->rn
->info
= NULL
;
325 RFAPI_RIB_PREFIX_COUNT_DECR(tcb
->rfd
, bgp
->rfapi
);
327 skiplist_free(tcb
->sl
);
328 route_unlock_node(tcb
->rn
);
331 XFREE(MTYPE_RFAPI_RECENT_DELETE
, tcb
);
333 RFAPI_RIB_CHECK_COUNTS(1, 0);
339 rfapiRibStartTimer(struct rfapi_descriptor
*rfd
, struct rfapi_info
*ri
,
340 struct route_node
*rn
, /* route node attached to */
343 struct thread
*t
= ri
->timer
;
344 struct rfapi_rib_tcb
*tcb
= NULL
;
345 char buf_prefix
[PREFIX_STRLEN
];
352 tcb
= XCALLOC(MTYPE_RFAPI_RECENT_DELETE
,
353 sizeof(struct rfapi_rib_tcb
));
359 tcb
->sl
= (struct skiplist
*)rn
->aggregate
;
360 SET_FLAG(tcb
->flags
, RFAPI_RIB_TCB_FLAG_DELETED
);
362 tcb
->sl
= (struct skiplist
*)rn
->info
;
363 UNSET_FLAG(tcb
->flags
, RFAPI_RIB_TCB_FLAG_DELETED
);
366 prefix2str(&rn
->p
, buf_prefix
, sizeof(buf_prefix
));
367 vnc_zlog_debug_verbose("%s: rfd %p pfx %s life %u", __func__
, rfd
,
368 buf_prefix
, ri
->lifetime
);
370 thread_add_timer(bm
->master
, rfapiRibExpireTimer
, tcb
, ri
->lifetime
,
375 extern void rfapi_rib_key_init(struct prefix
*prefix
, /* may be NULL */
376 struct prefix_rd
*rd
, /* may be NULL */
377 struct prefix
*aux
, /* may be NULL */
378 struct rfapi_rib_key
*rk
)
381 memset((void *)rk
, 0, sizeof(struct rfapi_rib_key
));
387 rk
->aux_prefix
= *aux
;
391 * Compares two <struct rfapi_rib_key>s
393 int rfapi_rib_key_cmp(void *k1
, void *k2
)
395 struct rfapi_rib_key
*a
= (struct rfapi_rib_key
*)k1
;
396 struct rfapi_rib_key
*b
= (struct rfapi_rib_key
*)k2
;
402 ret
= vnc_prefix_cmp(&a
->vn
, &b
->vn
);
406 ret
= vnc_prefix_cmp(&a
->rd
, &b
->rd
);
410 ret
= vnc_prefix_cmp(&a
->aux_prefix
, &b
->aux_prefix
);
417 * Note: this function will claim that two option chains are
418 * different unless their option items are in identical order.
419 * The consequence is that RFP updated responses can be sent
420 * unnecessarily, or that they might contain nexthop items
421 * that are not strictly needed.
423 * This function could be modified to compare option chains more
424 * thoroughly, but it's not clear that the extra compuation would
427 static int bgp_tea_options_cmp(struct bgp_tea_options
*a
,
428 struct bgp_tea_options
*b
)
436 if (a
->type
!= b
->type
)
437 return (a
->type
- b
->type
);
438 if (a
->length
!= b
->length
)
439 return (a
->length
= b
->length
);
440 if ((rc
= memcmp(a
->value
, b
->value
, a
->length
)))
442 if (!a
->next
!= !b
->next
) { /* logical xor */
443 return (a
->next
- b
->next
);
446 return bgp_tea_options_cmp(a
->next
, b
->next
);
450 static int rfapi_info_cmp(struct rfapi_info
*a
, struct rfapi_info
*b
)
457 if ((rc
= rfapi_rib_key_cmp(&a
->rk
, &b
->rk
)))
460 if ((rc
= vnc_prefix_cmp(&a
->un
, &b
->un
)))
463 if (a
->cost
!= b
->cost
)
464 return (a
->cost
- b
->cost
);
466 if (a
->lifetime
!= b
->lifetime
)
467 return (a
->lifetime
- b
->lifetime
);
469 if ((rc
= bgp_tea_options_cmp(a
->tea_options
, b
->tea_options
)))
475 void rfapiRibClear(struct rfapi_descriptor
*rfd
)
483 bgp
= bgp_get_default();
485 vnc_zlog_debug_verbose("%s: rfd=%p", __func__
, rfd
);
488 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
489 struct route_node
*pn
;
490 struct route_node
*rn
;
492 if (rfd
->rib_pending
[afi
]) {
493 for (pn
= route_top(rfd
->rib_pending
[afi
]); pn
;
494 pn
= route_next(pn
)) {
497 * free references into the rfapi_info
499 * freeing the structures themselves
504 pn
->aggregate
= NULL
;
506 pn
); /* skiplist deleted */
509 * free the rfapi_info structures
512 if (pn
->info
!= (void *)1) {
513 list_delete_and_null(
518 /* linklist or 1 deleted */
519 route_unlock_node(pn
);
524 for (rn
= route_top(rfd
->rib
[afi
]); rn
;
525 rn
= route_next(rn
)) {
528 struct rfapi_info
*ri
;
530 while (0 == skiplist_first(
537 skiplist_delete_first(
542 (struct skiplist
*)rn
->info
);
544 route_unlock_node(rn
);
545 RFAPI_RIB_PREFIX_COUNT_DECR(rfd
,
550 struct rfapi_info
*ri_del
;
552 /* delete skiplist & contents */
553 while (!skiplist_first(
556 NULL
, (void **)&ri_del
)) {
558 /* sl->del takes care of ri_del
560 skiplist_delete_first((
568 rn
->aggregate
= NULL
;
569 route_unlock_node(rn
);
574 if (rfd
->updated_responses_queue
)
575 work_queue_free_and_null(&rfd
->updated_responses_queue
);
579 * Release all dynamically-allocated memory that is part of an HD's RIB
581 void rfapiRibFree(struct rfapi_descriptor
*rfd
)
587 * NB rfd is typically detached from master list, so is not included
588 * in the count performed by RFAPI_RIB_CHECK_COUNTS
592 * Free routes attached to radix trees
596 /* Now the uncounted rfapi_info's are freed, so the check should succeed
598 RFAPI_RIB_CHECK_COUNTS(1, 0);
603 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
604 route_table_finish(rfd
->rib_pending
[afi
]);
605 rfd
->rib_pending
[afi
] = NULL
;
607 route_table_finish(rfd
->rib
[afi
]);
608 rfd
->rib
[afi
] = NULL
;
610 /* NB route_table_finish frees only prefix nodes, not chained
612 route_table_finish(rfd
->rsp_times
[afi
]);
613 rfd
->rib
[afi
] = NULL
;
618 * Copies struct bgp_info to struct rfapi_info, except for rk fields and un
620 static void rfapiRibBi2Ri(struct bgp_info
*bi
, struct rfapi_info
*ri
,
623 struct bgp_attr_encap_subtlv
*pEncap
;
625 ri
->cost
= rfapiRfpCost(bi
->attr
);
626 ri
->lifetime
= lifetime
;
628 /* This loop based on rfapiRouteInfo2NextHopEntry() */
629 for (pEncap
= bi
->attr
->vnc_subtlvs
; pEncap
; pEncap
= pEncap
->next
) {
630 struct bgp_tea_options
*hop
;
632 switch (pEncap
->type
) {
633 case BGP_VNC_SUBTLV_TYPE_LIFETIME
:
634 /* use configured lifetime, not attr lifetime */
637 case BGP_VNC_SUBTLV_TYPE_RFPOPTION
:
638 hop
= XCALLOC(MTYPE_BGP_TEA_OPTIONS
,
639 sizeof(struct bgp_tea_options
));
641 hop
->type
= pEncap
->value
[0];
642 hop
->length
= pEncap
->value
[1];
643 hop
->value
= XCALLOC(MTYPE_BGP_TEA_OPTIONS_VALUE
,
646 memcpy(hop
->value
, pEncap
->value
+ 2,
648 if (hop
->length
> pEncap
->length
- 2) {
650 "%s: VNC subtlv length mismatch: "
651 "RFP option says %d, attr says %d "
653 __func__
, hop
->length
,
655 hop
->length
= pEncap
->length
- 2;
657 hop
->next
= ri
->tea_options
;
658 ri
->tea_options
= hop
;
666 rfapi_un_options_free(ri
->un_options
); /* maybe free old version */
667 ri
->un_options
= rfapi_encap_tlv_to_un_option(bi
->attr
);
673 && decode_rd_type(bi
->extra
->vnc
.import
.rd
.val
)
674 == RD_TYPE_VNC_ETH
) {
677 struct rfapi_vn_option
*vo
;
679 vo
= XCALLOC(MTYPE_RFAPI_VN_OPTION
,
680 sizeof(struct rfapi_vn_option
));
683 vo
->type
= RFAPI_VN_OPTION_TYPE_L2ADDR
;
685 /* copy from RD already stored in bi, so we don't need it_node
687 memcpy(&vo
->v
.l2addr
.macaddr
, bi
->extra
->vnc
.import
.rd
.val
+ 2,
690 (void)rfapiEcommunityGetLNI(bi
->attr
->ecommunity
,
691 &vo
->v
.l2addr
.logical_net_id
);
692 (void)rfapiEcommunityGetEthernetTag(bi
->attr
->ecommunity
,
693 &vo
->v
.l2addr
.tag_id
);
695 /* local_nve_id comes from RD */
696 vo
->v
.l2addr
.local_nve_id
= bi
->extra
->vnc
.import
.rd
.val
[1];
698 /* label comes from MP_REACH_NLRI label */
699 vo
->v
.l2addr
.label
= decode_label(&bi
->extra
->label
[0]);
701 rfapi_vn_options_free(
702 ri
->vn_options
); /* maybe free old version */
707 * If there is an auxiliary IP address (L2 can have it), copy it
709 if (bi
->extra
&& bi
->extra
->vnc
.import
.aux_prefix
.family
) {
710 ri
->rk
.aux_prefix
= bi
->extra
->vnc
.import
.aux_prefix
;
717 * Install route into NVE RIB model so as to be consistent with
718 * caller's response to rfapi_query().
720 * Also: return indication to caller whether this specific route
721 * should be included in the response to the NVE according to
722 * the following tests:
724 * 1. If there were prior duplicates of this route in this same
725 * query response, don't include the route.
729 * 0 OK to include route in response
730 * !0 do not include route in response
732 int rfapiRibPreloadBi(
733 struct route_node
*rfd_rib_node
, /* NULL = don't preload or filter */
734 struct prefix
*pfx_vn
, struct prefix
*pfx_un
, uint32_t lifetime
,
737 struct rfapi_descriptor
*rfd
;
738 struct skiplist
*slRibPt
= NULL
;
739 struct rfapi_info
*ori
= NULL
;
740 struct rfapi_rib_key rk
;
741 struct route_node
*trn
;
747 afi
= family2afi(rfd_rib_node
->p
.family
);
749 rfd
= (struct rfapi_descriptor
*)(rfd_rib_node
->table
->info
);
751 memset((void *)&rk
, 0, sizeof(rk
));
753 rk
.rd
= bi
->extra
->vnc
.import
.rd
;
756 * If there is an auxiliary IP address (L2 can have it), copy it
758 if (bi
->extra
->vnc
.import
.aux_prefix
.family
) {
759 rk
.aux_prefix
= bi
->extra
->vnc
.import
.aux_prefix
;
763 * is this route already in NVE's RIB?
765 slRibPt
= (struct skiplist
*)rfd_rib_node
->info
;
767 if (slRibPt
&& !skiplist_search(slRibPt
, &rk
, (void **)&ori
)) {
769 if ((ori
->rsp_counter
== rfd
->rsp_counter
)
770 && (ori
->last_sent_time
== rfd
->rsp_time
)) {
771 return -1; /* duplicate in this response */
774 /* found: update contents of existing route in RIB */
776 rfapiRibBi2Ri(bi
, ori
, lifetime
);
778 /* not found: add new route to RIB */
779 ori
= rfapi_info_new();
782 rfapiRibBi2Ri(bi
, ori
, lifetime
);
785 slRibPt
= skiplist_new(0, rfapi_rib_key_cmp
, NULL
);
786 rfd_rib_node
->info
= slRibPt
;
787 route_lock_node(rfd_rib_node
);
788 RFAPI_RIB_PREFIX_COUNT_INCR(rfd
, rfd
->bgp
->rfapi
);
790 skiplist_insert(slRibPt
, &ori
->rk
, ori
);
793 ori
->last_sent_time
= rfapi_time(NULL
);
798 RFAPI_RIB_CHECK_COUNTS(0, 0);
799 rfapiRibStartTimer(rfd
, ori
, rfd_rib_node
, 0);
800 RFAPI_RIB_CHECK_COUNTS(0, 0);
803 * Update last sent time for prefix
805 trn
= route_node_get(rfd
->rsp_times
[afi
],
806 &rfd_rib_node
->p
); /* locks trn */
807 trn
->info
= (void *)(uintptr_t)bgp_clock();
809 route_unlock_node(trn
);
815 * Frees rfapi_info items at node
817 * Adjust 'rib' and 'rib_pending' as follows:
819 * If rib_pending node->info is 1 (magic value):
820 * callback: NHL = RIB NHL with lifetime = withdraw_lifetime_value
821 * RIB = remove all routes at the node
824 * For each item at rib node:
825 * if not present in pending node, move RIB item to "delete list"
827 * For each item at pending rib node:
828 * if present (same vn/un) in rib node with same lifetime & options, drop
829 * matching item from pending node
831 * For each remaining item at pending rib node, add or replace item
834 * Construct NHL as concatenation of pending list + delete list
838 static void process_pending_node(struct bgp
*bgp
, struct rfapi_descriptor
*rfd
,
840 struct route_node
*pn
, /* pending node */
841 struct rfapi_next_hop_entry
**head
,
842 struct rfapi_next_hop_entry
**tail
)
844 struct listnode
*node
= NULL
;
845 struct listnode
*nnode
= NULL
;
846 struct rfapi_info
*ri
= NULL
; /* happy valgrind */
847 struct rfapi_ip_prefix hp
= {0}; /* pfx to put in NHE */
848 struct route_node
*rn
= NULL
;
849 struct skiplist
*slRibPt
= NULL
; /* rib list */
850 struct skiplist
*slPendPt
= NULL
;
851 struct list
*lPendCost
= NULL
;
852 struct list
*delete_list
= NULL
;
853 int printedprefix
= 0;
854 char buf_prefix
[PREFIX_STRLEN
];
855 int rib_node_started_nonempty
= 0;
856 int sendingsomeroutes
= 0;
858 #if DEBUG_PROCESS_PENDING_NODE
859 unsigned int count_rib_initial
= 0;
860 unsigned int count_pend_vn_initial
= 0;
861 unsigned int count_pend_cost_initial
= 0;
865 prefix2str(&pn
->p
, buf_prefix
, sizeof(buf_prefix
));
866 vnc_zlog_debug_verbose("%s: afi=%d, %s pn->info=%p", __func__
, afi
,
867 buf_prefix
, pn
->info
);
869 if (AFI_L2VPN
!= afi
) {
870 rfapiQprefix2Rprefix(&pn
->p
, &hp
);
873 RFAPI_RIB_CHECK_COUNTS(1, 0);
876 * Find corresponding RIB node
878 rn
= route_node_get(rfd
->rib
[afi
], &pn
->p
); /* locks rn */
881 * RIB skiplist has key=rfapi_addr={vn,un}, val = rfapi_info,
882 * skiplist.del = NULL
884 slRibPt
= (struct skiplist
*)rn
->info
;
886 rib_node_started_nonempty
= 1;
888 slPendPt
= (struct skiplist
*)(pn
->aggregate
);
889 lPendCost
= (struct list
*)(pn
->info
);
891 #if DEBUG_PROCESS_PENDING_NODE
894 count_rib_initial
= skiplist_count(slRibPt
);
897 count_pend_vn_initial
= skiplist_count(slPendPt
);
899 if (lPendCost
&& lPendCost
!= (struct list
*)1)
900 count_pend_cost_initial
= lPendCost
->count
;
905 * Handle special case: delete all routes at prefix
907 if (lPendCost
== (struct list
*)1) {
908 vnc_zlog_debug_verbose("%s: lPendCost=1 => delete all",
910 if (slRibPt
&& !skiplist_empty(slRibPt
)) {
911 delete_list
= list_new();
913 == skiplist_first(slRibPt
, NULL
, (void **)&ri
)) {
915 char buf
[PREFIX_STRLEN
];
916 char buf2
[PREFIX_STRLEN
];
918 listnode_add(delete_list
, ri
);
919 vnc_zlog_debug_verbose(
920 "%s: after listnode_add, delete_list->count=%d",
921 __func__
, delete_list
->count
);
922 rfapiFreeBgpTeaOptionChain(ri
->tea_options
);
923 ri
->tea_options
= NULL
;
926 struct rfapi_rib_tcb
*tcb
;
928 tcb
= ((struct thread
*)ri
->timer
)->arg
;
929 thread_cancel(ri
->timer
);
930 XFREE(MTYPE_RFAPI_RECENT_DELETE
, tcb
);
934 prefix2str(&ri
->rk
.vn
, buf
, sizeof(buf
));
935 prefix2str(&ri
->un
, buf2
, sizeof(buf2
));
936 vnc_zlog_debug_verbose(
937 "%s: put dl pfx=%s vn=%s un=%s cost=%d life=%d vn_options=%p",
938 __func__
, buf_prefix
, buf
, buf2
,
939 ri
->cost
, ri
->lifetime
, ri
->vn_options
);
941 skiplist_delete_first(slRibPt
);
944 assert(skiplist_empty(slRibPt
));
946 skiplist_free(slRibPt
);
947 rn
->info
= slRibPt
= NULL
;
948 route_unlock_node(rn
);
950 lPendCost
= pn
->info
= NULL
;
951 route_unlock_node(pn
);
956 skiplist_free(slRibPt
);
958 route_unlock_node(rn
);
962 if (slPendPt
) { /* TBD I think we can toss this block */
963 skiplist_free(slPendPt
);
964 pn
->aggregate
= NULL
;
965 route_unlock_node(pn
);
969 route_unlock_node(pn
);
971 route_unlock_node(rn
); /* route_node_get() */
973 if (rib_node_started_nonempty
) {
974 RFAPI_RIB_PREFIX_COUNT_DECR(rfd
, bgp
->rfapi
);
977 RFAPI_RIB_CHECK_COUNTS(1, 0);
982 vnc_zlog_debug_verbose("%s: lPendCost->count=%d, slRibPt->count=%d",
984 (lPendCost
? (int)lPendCost
->count
: -1),
985 (slRibPt
? (int)slRibPt
->count
: -1));
988 * Iterate over routes at RIB Node.
989 * If not found at Pending Node, delete from RIB Node and add to
991 * If found at Pending Node
992 * If identical rfapi_info, delete from Pending Node
996 struct rfapi_info
*ori
;
999 * Iterate over RIB List
1002 while (!skiplist_next(slRibPt
, NULL
, (void **)&ori
, &cursor
)) {
1004 if (skiplist_search(slPendPt
, &ori
->rk
, (void **)&ri
)) {
1006 * Not in Pending list, so it should be deleted
1009 delete_list
= list_new();
1010 listnode_add(delete_list
, ori
);
1011 rfapiFreeBgpTeaOptionChain(ori
->tea_options
);
1012 ori
->tea_options
= NULL
;
1014 struct rfapi_rib_tcb
*tcb
;
1016 tcb
= ((struct thread
*)ori
->timer
)
1018 thread_cancel(ori
->timer
);
1019 XFREE(MTYPE_RFAPI_RECENT_DELETE
, tcb
);
1023 #if DEBUG_PROCESS_PENDING_NODE
1024 /* deleted from slRibPt below, after we're done
1026 vnc_zlog_debug_verbose(
1027 "%s: slRibPt ri %p not matched in pending list, delete",
1033 * Found in pending list. If same lifetime,
1035 * then remove from pending list because the
1039 if (!rfapi_info_cmp(ori
, ri
)) {
1040 skiplist_delete(slPendPt
, &ri
->rk
,
1044 /* linear walk: might need
1046 listnode_delete(lPendCost
,
1056 #if DEBUG_PROCESS_PENDING_NODE
1057 vnc_zlog_debug_verbose(
1058 "%s: slRibPt ri %p matched in pending list, %s",
1061 : "different info"));
1066 * Go back and delete items from RIB
1069 for (ALL_LIST_ELEMENTS_RO(delete_list
, node
, ri
)) {
1070 vnc_zlog_debug_verbose(
1071 "%s: deleting ri %p from slRibPt",
1073 assert(!skiplist_delete(slRibPt
, &ri
->rk
,
1076 if (skiplist_empty(slRibPt
)) {
1077 skiplist_free(slRibPt
);
1078 slRibPt
= rn
->info
= NULL
;
1079 route_unlock_node(rn
);
1084 RFAPI_RIB_CHECK_COUNTS(0, (delete_list
? delete_list
->count
: 0));
1087 * Iterate over routes at Pending Node
1089 * If {vn} found at RIB Node, update RIB Node route contents to match PN
1090 * If {vn} NOT found at RIB Node, add copy to RIB Node
1093 for (ALL_LIST_ELEMENTS_RO(lPendCost
, node
, ri
)) {
1095 struct rfapi_info
*ori
;
1098 && !skiplist_search(slRibPt
, &ri
->rk
,
1101 /* found: update contents of existing route in
1104 ori
->cost
= ri
->cost
;
1105 ori
->lifetime
= ri
->lifetime
;
1106 rfapiFreeBgpTeaOptionChain(ori
->tea_options
);
1108 rfapiOptionsDup(ri
->tea_options
);
1109 ori
->last_sent_time
= rfapi_time(NULL
);
1111 rfapiFreeRfapiVnOptionChain(ori
->vn_options
);
1113 rfapiVnOptionsDup(ri
->vn_options
);
1115 rfapiFreeRfapiUnOptionChain(ori
->un_options
);
1117 rfapiUnOptionsDup(ri
->un_options
);
1119 vnc_zlog_debug_verbose(
1120 "%s: matched lPendCost item %p in slRibPt, rewrote",
1125 char buf_rd
[RD_ADDRSTRLEN
];
1127 /* not found: add new route to RIB */
1128 ori
= rfapi_info_new();
1131 ori
->cost
= ri
->cost
;
1132 ori
->lifetime
= ri
->lifetime
;
1134 rfapiOptionsDup(ri
->tea_options
);
1135 ori
->last_sent_time
= rfapi_time(NULL
);
1137 rfapiVnOptionsDup(ri
->vn_options
);
1139 rfapiUnOptionsDup(ri
->un_options
);
1142 slRibPt
= skiplist_new(
1143 0, rfapi_rib_key_cmp
, NULL
);
1145 route_lock_node(rn
);
1147 skiplist_insert(slRibPt
, &ori
->rk
, ori
);
1150 prefix_rd2str(&ori
->rk
.rd
, buf_rd
,
1156 vnc_zlog_debug_verbose(
1157 "%s: nomatch lPendCost item %p in slRibPt, added (rd=%s)",
1158 __func__
, ri
, buf_rd
);
1164 RFAPI_RIB_CHECK_COUNTS(
1165 0, (delete_list
? delete_list
->count
: 0));
1166 rfapiRibStartTimer(rfd
, ori
, rn
, 0);
1167 RFAPI_RIB_CHECK_COUNTS(
1168 0, (delete_list
? delete_list
->count
: 0));
1175 * Construct NHL as concatenation of pending list + delete list
1179 RFAPI_RIB_CHECK_COUNTS(0, (delete_list
? delete_list
->count
: 0));
1186 vnc_zlog_debug_verbose("%s: lPendCost->count now %d", __func__
,
1188 vnc_zlog_debug_verbose("%s: For prefix %s (a)", __func__
,
1192 for (ALL_LIST_ELEMENTS(lPendCost
, node
, nnode
, ri
)) {
1194 struct rfapi_next_hop_entry
*new;
1195 struct route_node
*trn
;
1197 new = XCALLOC(MTYPE_RFAPI_NEXTHOP
,
1198 sizeof(struct rfapi_next_hop_entry
));
1201 if (ri
->rk
.aux_prefix
.family
) {
1202 rfapiQprefix2Rprefix(&ri
->rk
.aux_prefix
,
1206 if (AFI_L2VPN
== afi
) {
1207 /* hp is 0; need to set length to match
1209 new->prefix
.length
=
1210 (ri
->rk
.vn
.family
== AF_INET
)
1215 new->prefix
.cost
= ri
->cost
;
1216 new->lifetime
= ri
->lifetime
;
1217 rfapiQprefix2Raddr(&ri
->rk
.vn
, &new->vn_address
);
1218 rfapiQprefix2Raddr(&ri
->un
, &new->un_address
);
1219 /* free option chain from ri */
1220 rfapiFreeBgpTeaOptionChain(ri
->tea_options
);
1223 NULL
; /* option chain was transferred to NHL */
1225 new->vn_options
= ri
->vn_options
;
1227 NULL
; /* option chain was transferred to NHL */
1229 new->un_options
= ri
->un_options
;
1231 NULL
; /* option chain was transferred to NHL */
1234 (*tail
)->next
= new;
1239 sendingsomeroutes
= 1;
1241 ++rfd
->stat_count_nh_reachable
;
1242 ++bgp
->rfapi
->stat
.count_updated_response_updates
;
1245 * update this NVE's timestamp for this prefix
1247 trn
= route_node_get(rfd
->rsp_times
[afi
],
1248 &pn
->p
); /* locks trn */
1249 trn
->info
= (void *)(uintptr_t)bgp_clock();
1251 route_unlock_node(trn
);
1253 rfapiRfapiIpAddr2Str(&new->vn_address
, buf
, BUFSIZ
);
1254 rfapiRfapiIpAddr2Str(&new->un_address
, buf2
, BUFSIZ
);
1255 vnc_zlog_debug_verbose(
1256 "%s: add vn=%s un=%s cost=%d life=%d",
1257 __func__
, buf
, buf2
, new->prefix
.cost
,
1262 RFAPI_RIB_CHECK_COUNTS(0, (delete_list
? delete_list
->count
: 0));
1269 if (!printedprefix
) {
1270 vnc_zlog_debug_verbose("%s: For prefix %s (d)",
1271 __func__
, buf_prefix
);
1273 vnc_zlog_debug_verbose("%s: delete_list has %d elements",
1274 __func__
, delete_list
->count
);
1276 RFAPI_RIB_CHECK_COUNTS(0, delete_list
->count
);
1277 if (!CHECK_FLAG(bgp
->rfapi_cfg
->flags
,
1278 BGP_VNC_CONFIG_RESPONSE_REMOVAL_DISABLE
)) {
1280 for (ALL_LIST_ELEMENTS(delete_list
, node
, nnode
, ri
)) {
1282 struct rfapi_next_hop_entry
*new;
1283 struct rfapi_info
*ri_del
;
1285 RFAPI_RIB_CHECK_COUNTS(0, delete_list
->count
);
1287 MTYPE_RFAPI_NEXTHOP
,
1288 sizeof(struct rfapi_next_hop_entry
));
1291 if (ri
->rk
.aux_prefix
.family
) {
1292 rfapiQprefix2Rprefix(&ri
->rk
.aux_prefix
,
1296 if (AFI_L2VPN
== afi
) {
1297 /* hp is 0; need to set length
1298 * to match AF of vn */
1299 new->prefix
.length
=
1307 new->prefix
.cost
= ri
->cost
;
1308 new->lifetime
= RFAPI_REMOVE_RESPONSE_LIFETIME
;
1309 rfapiQprefix2Raddr(&ri
->rk
.vn
,
1311 rfapiQprefix2Raddr(&ri
->un
, &new->un_address
);
1313 new->vn_options
= ri
->vn_options
;
1314 ri
->vn_options
= NULL
; /* option chain was
1315 transferred to NHL */
1317 new->un_options
= ri
->un_options
;
1318 ri
->un_options
= NULL
; /* option chain was
1319 transferred to NHL */
1322 (*tail
)->next
= new;
1327 ++rfd
->stat_count_nh_removal
;
1329 .count_updated_response_deletes
;
1331 rfapiRfapiIpAddr2Str(&new->vn_address
, buf
,
1333 rfapiRfapiIpAddr2Str(&new->un_address
, buf2
,
1335 vnc_zlog_debug_verbose(
1336 "%s: DEL vn=%s un=%s cost=%d life=%d",
1337 __func__
, buf
, buf2
, new->prefix
.cost
,
1340 RFAPI_RIB_CHECK_COUNTS(0, delete_list
->count
);
1342 * Update/add to list of recent deletions at
1345 if (!rn
->aggregate
) {
1346 rn
->aggregate
= skiplist_new(
1347 0, rfapi_rib_key_cmp
,
1350 route_lock_node(rn
);
1352 RFAPI_RIB_CHECK_COUNTS(0, delete_list
->count
);
1354 /* sanity check lifetime */
1356 > RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
1358 RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
;
1360 RFAPI_RIB_CHECK_COUNTS(0, delete_list
->count
);
1361 /* cancel normal expire timer */
1363 struct rfapi_rib_tcb
*tcb
;
1365 tcb
= ((struct thread
*)ri
->timer
)->arg
;
1367 (struct thread
*)ri
->timer
);
1368 XFREE(MTYPE_RFAPI_RECENT_DELETE
, tcb
);
1371 RFAPI_RIB_CHECK_COUNTS(0, delete_list
->count
);
1374 * Look in "recently-deleted" list
1376 if (skiplist_search(
1377 (struct skiplist
*)(rn
->aggregate
),
1378 &ri
->rk
, (void **)&ri_del
)) {
1382 RFAPI_RIB_CHECK_COUNTS(
1383 0, delete_list
->count
);
1385 * NOT in "recently-deleted" list
1389 node
); /* does not free ri */
1390 rc
= skiplist_insert(
1396 RFAPI_RIB_CHECK_COUNTS(
1397 0, delete_list
->count
);
1398 rfapiRibStartTimer(rfd
, ri
, rn
, 1);
1399 RFAPI_RIB_CHECK_COUNTS(
1400 0, delete_list
->count
);
1401 ri
->last_sent_time
= rfapi_time(NULL
);
1404 char buf_rd
[RD_ADDRSTRLEN
];
1406 vnc_zlog_debug_verbose(
1407 "%s: move route to recently deleted list, rd=%s",
1418 * IN "recently-deleted" list
1420 RFAPI_RIB_CHECK_COUNTS(
1421 0, delete_list
->count
);
1422 rfapiRibStartTimer(rfd
, ri_del
, rn
, 1);
1423 RFAPI_RIB_CHECK_COUNTS(
1424 0, delete_list
->count
);
1425 ri
->last_sent_time
= rfapi_time(NULL
);
1429 vnc_zlog_debug_verbose(
1430 "%s: response removal disabled, omitting removals",
1434 delete_list
->del
= (void (*)(void *))rfapi_info_free
;
1435 list_delete_and_null(&delete_list
);
1438 RFAPI_RIB_CHECK_COUNTS(0, 0);
1441 * Reset pending lists. The final route_unlock_node() will probably
1442 * cause the pending node to be released.
1445 skiplist_free(slPendPt
);
1446 pn
->aggregate
= NULL
;
1447 route_unlock_node(pn
);
1450 list_delete_and_null(&lPendCost
);
1452 route_unlock_node(pn
);
1454 RFAPI_RIB_CHECK_COUNTS(0, 0);
1456 if (rib_node_started_nonempty
) {
1458 RFAPI_RIB_PREFIX_COUNT_DECR(rfd
, bgp
->rfapi
);
1462 RFAPI_RIB_PREFIX_COUNT_INCR(rfd
, bgp
->rfapi
);
1466 if (sendingsomeroutes
)
1467 rfapiMonitorTimersRestart(rfd
, &pn
->p
);
1469 route_unlock_node(rn
); /* route_node_get() */
1471 RFAPI_RIB_CHECK_COUNTS(1, 0);
1475 * regardless of targets, construct a single callback by doing
1476 * only one traversal of the pending RIB
1482 static void rib_do_callback_onepass(struct rfapi_descriptor
*rfd
, afi_t afi
)
1484 struct bgp
*bgp
= bgp_get_default();
1485 struct rfapi_next_hop_entry
*head
= NULL
;
1486 struct rfapi_next_hop_entry
*tail
= NULL
;
1487 struct route_node
*rn
;
1490 vnc_zlog_debug_verbose("%s: rfd=%p, afi=%d", __func__
, rfd
, afi
);
1493 if (!rfd
->rib_pending
[afi
])
1498 for (rn
= route_top(rfd
->rib_pending
[afi
]); rn
; rn
= route_next(rn
)) {
1499 process_pending_node(bgp
, rfd
, afi
, rn
, &head
, &tail
);
1503 rfapi_response_cb_t
*f
;
1506 vnc_zlog_debug_verbose("%s: response callback NHL follows:",
1508 rfapiPrintNhl(NULL
, head
);
1511 if (rfd
->response_cb
)
1512 f
= rfd
->response_cb
;
1514 f
= bgp
->rfapi
->rfp_methods
.response_cb
;
1516 bgp
->rfapi
->flags
|= RFAPI_INCALLBACK
;
1517 vnc_zlog_debug_verbose("%s: invoking updated response callback",
1519 (*f
)(head
, rfd
->cookie
);
1520 bgp
->rfapi
->flags
&= ~RFAPI_INCALLBACK
;
1521 ++bgp
->rfapi
->response_updated_count
;
1525 static wq_item_status
rfapiRibDoQueuedCallback(struct work_queue
*wq
,
1528 struct rfapi_descriptor
*rfd
;
1530 uint32_t queued_flag
;
1532 RFAPI_RIB_CHECK_COUNTS(1, 0);
1534 rfd
= ((struct rfapi_updated_responses_queue
*)data
)->rfd
;
1535 afi
= ((struct rfapi_updated_responses_queue
*)data
)->afi
;
1537 /* Make sure the HD wasn't closed after the work item was scheduled */
1538 if (rfapi_check(rfd
))
1541 rib_do_callback_onepass(rfd
, afi
);
1543 queued_flag
= RFAPI_QUEUED_FLAG(afi
);
1545 UNSET_FLAG(rfd
->flags
, queued_flag
);
1547 RFAPI_RIB_CHECK_COUNTS(1, 0);
1552 static void rfapiRibQueueItemDelete(struct work_queue
*wq
, void *data
)
1554 XFREE(MTYPE_RFAPI_UPDATED_RESPONSE_QUEUE
, data
);
1557 static void updated_responses_queue_init(struct rfapi_descriptor
*rfd
)
1559 if (rfd
->updated_responses_queue
)
1562 rfd
->updated_responses_queue
=
1563 work_queue_new(bm
->master
, "rfapi updated responses");
1564 assert(rfd
->updated_responses_queue
);
1566 rfd
->updated_responses_queue
->spec
.workfunc
= rfapiRibDoQueuedCallback
;
1567 rfd
->updated_responses_queue
->spec
.del_item_data
=
1568 rfapiRibQueueItemDelete
;
1569 rfd
->updated_responses_queue
->spec
.max_retries
= 0;
1570 rfd
->updated_responses_queue
->spec
.hold
= 1;
1574 * Called when an import table node is modified. Construct a
1575 * new complete nexthop list, sorted by cost (lowest first),
1576 * based on the import table node.
1578 * Filter out duplicate nexthops (vn address). There should be
1579 * only one UN address per VN address from the point of view of
1580 * a given import table, so we can probably ignore UN addresses
1583 * Based on rfapiNhlAddNodeRoutes()
1585 void rfapiRibUpdatePendingNode(
1586 struct bgp
*bgp
, struct rfapi_descriptor
*rfd
,
1587 struct rfapi_import_table
*it
, /* needed for L2 */
1588 struct route_node
*it_node
, uint32_t lifetime
)
1590 struct prefix
*prefix
;
1591 struct bgp_info
*bi
;
1592 struct route_node
*pn
;
1594 uint32_t queued_flag
;
1596 char buf
[PREFIX_STRLEN
];
1598 vnc_zlog_debug_verbose("%s: entry", __func__
);
1600 if (CHECK_FLAG(bgp
->rfapi_cfg
->flags
, BGP_VNC_CONFIG_CALLBACK_DISABLE
))
1603 vnc_zlog_debug_verbose("%s: callbacks are not disabled", __func__
);
1605 RFAPI_RIB_CHECK_COUNTS(1, 0);
1607 prefix
= &it_node
->p
;
1608 afi
= family2afi(prefix
->family
);
1609 prefix2str(prefix
, buf
, sizeof(buf
));
1610 vnc_zlog_debug_verbose("%s: prefix=%s", __func__
, buf
);
1612 pn
= route_node_get(rfd
->rib_pending
[afi
], prefix
);
1615 vnc_zlog_debug_verbose("%s: pn->info=%p, pn->aggregate=%p", __func__
,
1616 pn
->info
, pn
->aggregate
);
1618 if (pn
->aggregate
) {
1620 * free references into the rfapi_info structures before
1621 * freeing the structures themselves
1623 skiplist_free((struct skiplist
*)(pn
->aggregate
));
1624 pn
->aggregate
= NULL
;
1625 route_unlock_node(pn
); /* skiplist deleted */
1630 * free the rfapi_info structures
1633 if (pn
->info
!= (void *)1) {
1634 list_delete_and_null((struct list
**)(&pn
->info
));
1637 route_unlock_node(pn
); /* linklist or 1 deleted */
1641 * The BIs in the import table are already sorted by cost
1643 for (bi
= it_node
->info
; bi
; bi
= bi
->next
) {
1645 struct rfapi_info
*ri
;
1646 struct prefix pfx_nh
;
1649 /* shouldn't happen */
1650 /* TBD increment error stats counter */
1654 /* shouldn't happen */
1655 /* TBD increment error stats counter */
1659 rfapiNexthop2Prefix(bi
->attr
, &pfx_nh
);
1662 * Omit route if nexthop is self
1664 if (CHECK_FLAG(bgp
->rfapi_cfg
->flags
,
1665 BGP_VNC_CONFIG_FILTER_SELF_FROM_RSP
)) {
1667 struct prefix pfx_vn
;
1669 assert(!rfapiRaddr2Qprefix(&rfd
->vn_addr
, &pfx_vn
));
1670 if (prefix_same(&pfx_vn
, &pfx_nh
))
1674 ri
= rfapi_info_new();
1676 ri
->rk
.rd
= bi
->extra
->vnc
.import
.rd
;
1678 * If there is an auxiliary IP address (L2 can have it), copy it
1680 if (bi
->extra
->vnc
.import
.aux_prefix
.family
) {
1681 ri
->rk
.aux_prefix
= bi
->extra
->vnc
.import
.aux_prefix
;
1684 if (rfapiGetUnAddrOfVpnBi(bi
, &ri
->un
)) {
1685 rfapi_info_free(ri
);
1689 if (!pn
->aggregate
) {
1691 skiplist_new(0, rfapi_rib_key_cmp
, NULL
);
1692 route_lock_node(pn
);
1696 * If we have already added this nexthop, the insert will fail.
1697 * Note that the skiplist key is a pointer INTO the rfapi_info
1698 * structure which will be added to the "info" list.
1699 * The skiplist entry VALUE is not used for anything but
1700 * might be useful during debugging.
1702 if (skiplist_insert((struct skiplist
*)pn
->aggregate
, &ri
->rk
,
1708 rfapi_info_free(ri
);
1712 rfapiRibBi2Ri(bi
, ri
, lifetime
);
1715 pn
->info
= list_new();
1716 ((struct list
*)(pn
->info
))->del
=
1717 (void (*)(void *))rfapi_info_free
;
1718 route_lock_node(pn
);
1721 listnode_add((struct list
*)(pn
->info
), ri
);
1725 count
= ((struct list
*)(pn
->info
))->count
;
1730 assert(!pn
->aggregate
);
1731 pn
->info
= (void *)1; /* magic value means this node has no
1733 route_lock_node(pn
);
1736 route_unlock_node(pn
); /* route_node_get */
1738 queued_flag
= RFAPI_QUEUED_FLAG(afi
);
1740 if (!CHECK_FLAG(rfd
->flags
, queued_flag
)) {
1742 struct rfapi_updated_responses_queue
*urq
;
1744 urq
= XCALLOC(MTYPE_RFAPI_UPDATED_RESPONSE_QUEUE
,
1745 sizeof(struct rfapi_updated_responses_queue
));
1747 if (!rfd
->updated_responses_queue
)
1748 updated_responses_queue_init(rfd
);
1750 SET_FLAG(rfd
->flags
, queued_flag
);
1753 work_queue_add(rfd
->updated_responses_queue
, urq
);
1755 RFAPI_RIB_CHECK_COUNTS(1, 0);
1758 void rfapiRibUpdatePendingNodeSubtree(
1759 struct bgp
*bgp
, struct rfapi_descriptor
*rfd
,
1760 struct rfapi_import_table
*it
, struct route_node
*it_node
,
1761 struct route_node
*omit_subtree
, /* may be NULL */
1764 /* FIXME: need to find a better way here to work without sticking our
1765 * hands in node->link */
1766 if (it_node
->l_left
&& (it_node
->l_left
!= omit_subtree
)) {
1767 if (it_node
->l_left
->info
)
1768 rfapiRibUpdatePendingNode(bgp
, rfd
, it
, it_node
->l_left
,
1770 rfapiRibUpdatePendingNodeSubtree(bgp
, rfd
, it
, it_node
->l_left
,
1771 omit_subtree
, lifetime
);
1774 if (it_node
->l_right
&& (it_node
->l_right
!= omit_subtree
)) {
1775 if (it_node
->l_right
->info
)
1776 rfapiRibUpdatePendingNode(bgp
, rfd
, it
,
1777 it_node
->l_right
, lifetime
);
1778 rfapiRibUpdatePendingNodeSubtree(bgp
, rfd
, it
, it_node
->l_right
,
1779 omit_subtree
, lifetime
);
1786 * 0 allow prefix to be included in response
1787 * !0 don't allow prefix to be included in response
1789 int rfapiRibFTDFilterRecentPrefix(
1790 struct rfapi_descriptor
*rfd
,
1791 struct route_node
*it_rn
, /* import table node */
1792 struct prefix
*pfx_target_original
) /* query target */
1794 struct bgp
*bgp
= rfd
->bgp
;
1795 afi_t afi
= family2afi(it_rn
->p
.family
);
1797 struct route_node
*trn
;
1800 * Not in FTD mode, so allow prefix
1802 if (bgp
->rfapi_cfg
->rfp_cfg
.download_type
!= RFAPI_RFP_DOWNLOAD_FULL
)
1807 * This matches behavior of now-obsolete rfapiRibFTDFilterRecent(),
1808 * but we need to decide if that is correct.
1810 if (it_rn
->p
.family
== AF_ETHERNET
)
1813 #if DEBUG_FTD_FILTER_RECENT
1815 char buf_pfx
[PREFIX_STRLEN
];
1817 prefix2str(&it_rn
->p
, buf_pfx
, sizeof(buf_pfx
));
1818 vnc_zlog_debug_verbose("%s: prefix %s", __func__
, buf_pfx
);
1823 * prefix covers target address, so allow prefix
1825 if (prefix_match(&it_rn
->p
, pfx_target_original
)) {
1826 #if DEBUG_FTD_FILTER_RECENT
1827 vnc_zlog_debug_verbose("%s: prefix covers target, allowed",
1834 * check this NVE's timestamp for this prefix
1836 trn
= route_node_get(rfd
->rsp_times
[afi
], &it_rn
->p
); /* locks trn */
1837 prefix_time
= (time_t)trn
->info
;
1839 route_unlock_node(trn
);
1841 #if DEBUG_FTD_FILTER_RECENT
1842 vnc_zlog_debug_verbose("%s: last sent time %lu, last allowed time %lu",
1843 __func__
, prefix_time
,
1844 rfd
->ftd_last_allowed_time
);
1848 * haven't sent this prefix, which doesn't cover target address,
1849 * to NVE since ftd_advertisement_interval, so OK to send now.
1851 if (prefix_time
<= rfd
->ftd_last_allowed_time
)
1858 * Call when rfapi returns from rfapi_query() so the RIB reflects
1859 * the routes sent to the NVE before the first updated response
1861 * Also: remove duplicates from response. Caller should use returned
1862 * value of nexthop chain.
1864 struct rfapi_next_hop_entry
*
1865 rfapiRibPreload(struct bgp
*bgp
, struct rfapi_descriptor
*rfd
,
1866 struct rfapi_next_hop_entry
*response
, int use_eth_resolution
)
1868 struct rfapi_next_hop_entry
*nhp
;
1869 struct rfapi_next_hop_entry
*nhp_next
;
1870 struct rfapi_next_hop_entry
*head
= NULL
;
1871 struct rfapi_next_hop_entry
*tail
= NULL
;
1872 time_t new_last_sent_time
;
1874 vnc_zlog_debug_verbose("%s: loading response=%p, use_eth_resolution=%d",
1875 __func__
, response
, use_eth_resolution
);
1877 new_last_sent_time
= rfapi_time(NULL
);
1879 for (nhp
= response
; nhp
; nhp
= nhp_next
) {
1882 struct rfapi_rib_key rk
;
1884 struct rfapi_info
*ri
;
1886 struct route_node
*rn
;
1887 int rib_node_started_nonempty
= 0;
1888 struct route_node
*trn
;
1891 /* save in case we delete nhp */
1892 nhp_next
= nhp
->next
;
1894 if (nhp
->lifetime
== RFAPI_REMOVE_RESPONSE_LIFETIME
) {
1896 * weird, shouldn't happen
1898 vnc_zlog_debug_verbose(
1899 "%s: got nhp->lifetime == RFAPI_REMOVE_RESPONSE_LIFETIME",
1905 if (use_eth_resolution
) {
1906 /* get the prefix of the ethernet address in the L2
1908 struct rfapi_l2address_option
*pL2o
;
1909 struct rfapi_vn_option
*vo
;
1912 * Look for VN option of type
1913 * RFAPI_VN_OPTION_TYPE_L2ADDR
1915 for (pL2o
= NULL
, vo
= nhp
->vn_options
; vo
;
1917 if (RFAPI_VN_OPTION_TYPE_L2ADDR
== vo
->type
) {
1918 pL2o
= &vo
->v
.l2addr
;
1925 * not supposed to happen
1927 vnc_zlog_debug_verbose("%s: missing L2 info",
1933 rfapiL2o2Qprefix(pL2o
, &pfx
);
1935 rfapiRprefix2Qprefix(&nhp
->prefix
, &pfx
);
1936 afi
= family2afi(pfx
.family
);
1940 * TBD for ethernet, rib must know the right way to distinguish
1943 * Current approach: prefix is key to radix tree; then
1944 * each prefix has a set of routes with unique VN addrs
1948 * Look up prefix in RIB
1950 rn
= route_node_get(rfd
->rib
[afi
], &pfx
); /* locks rn */
1953 rib_node_started_nonempty
= 1;
1955 rn
->info
= skiplist_new(0, rfapi_rib_key_cmp
, NULL
);
1956 route_lock_node(rn
);
1960 * Look up route at prefix
1963 memset((void *)&rk
, 0, sizeof(rk
));
1964 assert(!rfapiRaddr2Qprefix(&nhp
->vn_address
, &rk
.vn
));
1966 if (use_eth_resolution
) {
1967 /* copy what came from aux_prefix to rk.aux_prefix */
1968 rfapiRprefix2Qprefix(&nhp
->prefix
, &rk
.aux_prefix
);
1969 if (RFAPI_0_PREFIX(&rk
.aux_prefix
)
1970 && RFAPI_HOST_PREFIX(&rk
.aux_prefix
)) {
1971 /* mark as "none" if nhp->prefix is 0/32 or
1973 rk
.aux_prefix
.family
= 0;
1979 char str_vn
[PREFIX_STRLEN
];
1980 char str_aux_prefix
[PREFIX_STRLEN
];
1983 str_aux_prefix
[0] = 0;
1985 prefix2str(&rk
.vn
, str_vn
, sizeof(str_vn
));
1986 prefix2str(&rk
.aux_prefix
, str_aux_prefix
,
1987 sizeof(str_aux_prefix
));
1989 if (!rk
.aux_prefix
.family
) {
1991 vnc_zlog_debug_verbose(
1992 "%s: rk.vn=%s rk.aux_prefix=%s", __func__
,
1994 (rk
.aux_prefix
.family
? str_aux_prefix
: "-"));
1996 vnc_zlog_debug_verbose(
1997 "%s: RIB skiplist for this prefix follows", __func__
);
1998 rfapiRibShowRibSl(NULL
, &rn
->p
, (struct skiplist
*)rn
->info
);
2002 if (!skiplist_search((struct skiplist
*)rn
->info
, &rk
,
2005 * Already have this route; make values match
2007 rfapiFreeRfapiUnOptionChain(ri
->un_options
);
2008 ri
->un_options
= NULL
;
2009 rfapiFreeRfapiVnOptionChain(ri
->vn_options
);
2010 ri
->vn_options
= NULL
;
2013 vnc_zlog_debug_verbose("%s: found in RIB", __func__
);
2017 * Filter duplicate routes from initial response.
2018 * Check timestamps to avoid wraparound problems
2020 if ((ri
->rsp_counter
!= rfd
->rsp_counter
)
2021 || (ri
->last_sent_time
!= new_last_sent_time
)) {
2024 vnc_zlog_debug_verbose(
2025 "%s: allowed due to counter/timestamp diff",
2034 vnc_zlog_debug_verbose(
2035 "%s: allowed due to not yet in RIB", __func__
);
2037 /* not found: add new route to RIB */
2038 ri
= rfapi_info_new();
2044 assert(!rfapiRaddr2Qprefix(&nhp
->un_address
, &ri
->un
));
2045 ri
->cost
= nhp
->prefix
.cost
;
2046 ri
->lifetime
= nhp
->lifetime
;
2047 ri
->vn_options
= rfapiVnOptionsDup(nhp
->vn_options
);
2048 ri
->rsp_counter
= rfd
->rsp_counter
;
2049 ri
->last_sent_time
= rfapi_time(NULL
);
2053 rc
= skiplist_insert((struct skiplist
*)rn
->info
,
2058 if (!rib_node_started_nonempty
) {
2059 RFAPI_RIB_PREFIX_COUNT_INCR(rfd
, bgp
->rfapi
);
2062 RFAPI_RIB_CHECK_COUNTS(0, 0);
2063 rfapiRibStartTimer(rfd
, ri
, rn
, 0);
2064 RFAPI_RIB_CHECK_COUNTS(0, 0);
2066 route_unlock_node(rn
);
2069 * update this NVE's timestamp for this prefix
2071 trn
= route_node_get(rfd
->rsp_times
[afi
], &pfx
); /* locks trn */
2072 trn
->info
= (void *)(uintptr_t)bgp_clock();
2074 route_unlock_node(trn
);
2077 char str_pfx
[PREFIX_STRLEN
];
2078 char str_pfx_vn
[PREFIX_STRLEN
];
2080 prefix2str(&pfx
, str_pfx
, sizeof(str_pfx
));
2081 prefix2str(&rk
.vn
, str_pfx_vn
, sizeof(str_pfx_vn
));
2082 vnc_zlog_debug_verbose(
2083 "%s: added pfx=%s nh[vn]=%s, cost=%u, lifetime=%u, allowed=%d",
2084 __func__
, str_pfx
, str_pfx_vn
, nhp
->prefix
.cost
,
2085 nhp
->lifetime
, allowed
);
2096 rfapi_un_options_free(nhp
->un_options
);
2097 nhp
->un_options
= NULL
;
2098 rfapi_vn_options_free(nhp
->vn_options
);
2099 nhp
->vn_options
= NULL
;
2101 XFREE(MTYPE_RFAPI_NEXTHOP
, nhp
);
2111 void rfapiRibPendingDeleteRoute(struct bgp
*bgp
, struct rfapi_import_table
*it
,
2112 afi_t afi
, struct route_node
*it_node
)
2114 struct rfapi_descriptor
*rfd
;
2115 struct listnode
*node
;
2116 char buf
[PREFIX_STRLEN
];
2118 prefix2str(&it_node
->p
, buf
, sizeof(buf
));
2119 vnc_zlog_debug_verbose("%s: entry, it=%p, afi=%d, it_node=%p, pfx=%s",
2120 __func__
, it
, afi
, it_node
, buf
);
2122 if (AFI_L2VPN
== afi
) {
2124 * ethernet import tables are per-LNI and each ethernet monitor
2125 * identifies the rfd that owns it.
2127 struct rfapi_monitor_eth
*m
;
2128 struct route_node
*rn
;
2129 struct skiplist
*sl
;
2134 * route-specific monitors
2136 if ((sl
= RFAPI_MONITOR_ETH(it_node
))) {
2138 vnc_zlog_debug_verbose(
2139 "%s: route-specific skiplist: %p", __func__
,
2143 rc
= skiplist_next(sl
, NULL
, (void **)&m
,
2145 !rc
; rc
= skiplist_next(sl
, NULL
, (void **)&m
,
2146 (void **)&cursor
)) {
2148 #if DEBUG_PENDING_DELETE_ROUTE
2149 vnc_zlog_debug_verbose("%s: eth monitor rfd=%p",
2153 * If we have already sent a route with this
2155 * NVE, it's OK to send an update with the
2158 if ((rn
= route_node_lookup(m
->rfd
->rib
[afi
],
2160 rfapiRibUpdatePendingNode(
2161 bgp
, m
->rfd
, it
, it_node
,
2162 m
->rfd
->response_lifetime
);
2163 route_unlock_node(rn
);
2169 * all-routes/FTD monitors
2171 for (m
= it
->eth0_queries
; m
; m
= m
->next
) {
2172 #if DEBUG_PENDING_DELETE_ROUTE
2173 vnc_zlog_debug_verbose("%s: eth0 monitor rfd=%p",
2177 * If we have already sent a route with this prefix to
2179 * NVE, it's OK to send an update with the delete
2181 if ((rn
= route_node_lookup(m
->rfd
->rib
[afi
],
2183 rfapiRibUpdatePendingNode(
2184 bgp
, m
->rfd
, it
, it_node
,
2185 m
->rfd
->response_lifetime
);
2191 * Find RFDs that reference this import table
2193 for (ALL_LIST_ELEMENTS_RO(&bgp
->rfapi
->descriptors
, node
,
2196 struct route_node
*rn
;
2198 vnc_zlog_debug_verbose(
2199 "%s: comparing rfd(%p)->import_table=%p to it=%p",
2200 __func__
, rfd
, rfd
->import_table
, it
);
2202 if (rfd
->import_table
!= it
)
2205 vnc_zlog_debug_verbose("%s: matched rfd %p", __func__
,
2209 * If we have sent a response to this NVE with this
2211 * previously, we should send an updated response.
2213 if ((rn
= route_node_lookup(rfd
->rib
[afi
],
2215 rfapiRibUpdatePendingNode(
2216 bgp
, rfd
, it
, it_node
,
2217 rfd
->response_lifetime
);
2218 route_unlock_node(rn
);
2224 void rfapiRibShowResponsesSummary(void *stream
)
2226 int (*fp
)(void *, const char *, ...);
2229 const char *vty_newline
;
2230 struct bgp
*bgp
= bgp_get_default();
2233 int nves_with_nonempty_ribs
= 0;
2234 struct rfapi_descriptor
*rfd
;
2235 struct listnode
*node
;
2237 if (rfapiStream2Vty(stream
, &fp
, &vty
, &out
, &vty_newline
) == 0)
2240 fp(out
, "Unable to find default BGP instance\n");
2244 fp(out
, "%-24s ", "Responses: (Prefixes)");
2245 fp(out
, "%-8s %-8u ", "Active:", bgp
->rfapi
->rib_prefix_count_total
);
2246 fp(out
, "%-8s %-8u",
2247 "Maximum:", bgp
->rfapi
->rib_prefix_count_total_max
);
2250 fp(out
, "%-24s ", " (Updated)");
2251 fp(out
, "%-8s %-8u ",
2252 "Update:", bgp
->rfapi
->stat
.count_updated_response_updates
);
2253 fp(out
, "%-8s %-8u",
2254 "Remove:", bgp
->rfapi
->stat
.count_updated_response_deletes
);
2255 fp(out
, "%-8s %-8u", "Total:",
2256 bgp
->rfapi
->stat
.count_updated_response_updates
2257 + bgp
->rfapi
->stat
.count_updated_response_deletes
);
2260 fp(out
, "%-24s ", " (NVEs)");
2261 for (ALL_LIST_ELEMENTS_RO(&bgp
->rfapi
->descriptors
, node
, rfd
)) {
2263 if (rfd
->rib_prefix_count
)
2264 ++nves_with_nonempty_ribs
;
2266 fp(out
, "%-8s %-8u ", "Active:", nves_with_nonempty_ribs
);
2267 fp(out
, "%-8s %-8u", "Total:", nves
);
2271 void rfapiRibShowResponsesSummaryClear(void)
2273 struct bgp
*bgp
= bgp_get_default();
2275 bgp
->rfapi
->rib_prefix_count_total_max
=
2276 bgp
->rfapi
->rib_prefix_count_total
;
2279 static int print_rib_sl(int (*fp
)(void *, const char *, ...), struct vty
*vty
,
2280 void *out
, struct skiplist
*sl
, int deleted
,
2281 char *str_pfx
, int *printedprefix
)
2283 struct rfapi_info
*ri
;
2286 int routes_displayed
= 0;
2289 for (rc
= skiplist_next(sl
, NULL
, (void **)&ri
, &cursor
); !rc
;
2290 rc
= skiplist_next(sl
, NULL
, (void **)&ri
, &cursor
)) {
2292 char str_vn
[PREFIX_STRLEN
];
2293 char str_un
[PREFIX_STRLEN
];
2294 char str_lifetime
[BUFSIZ
];
2295 char str_age
[BUFSIZ
];
2297 char str_rd
[RD_ADDRSTRLEN
];
2301 prefix2str(&ri
->rk
.vn
, str_vn
, sizeof(str_vn
));
2302 p
= index(str_vn
, '/');
2306 prefix2str(&ri
->un
, str_un
, sizeof(str_un
));
2307 p
= index(str_un
, '/');
2311 rfapiFormatSeconds(ri
->lifetime
, str_lifetime
, BUFSIZ
);
2312 #if RFAPI_REGISTRATIONS_REPORT_AGE
2313 rfapiFormatAge(ri
->last_sent_time
, str_age
, BUFSIZ
);
2316 time_t now
= rfapi_time(NULL
);
2318 ri
->last_sent_time
+ (time_t)ri
->lifetime
;
2319 /* allow for delayed/async removal */
2320 rfapiFormatSeconds((expire
> now
? expire
- now
: 1),
2325 str_rd
[0] = 0; /* start empty */
2327 prefix_rd2str(&ri
->rk
.rd
, str_rd
, sizeof(str_rd
));
2330 fp(out
, " %c %-20s %-15s %-15s %-4u %-8s %-8s %s\n",
2331 deleted
? 'r' : ' ', *printedprefix
? "" : str_pfx
, str_vn
,
2332 str_un
, ri
->cost
, str_lifetime
, str_age
, str_rd
);
2334 if (!*printedprefix
)
2337 return routes_displayed
;
2342 * This one is for debugging (set stream to NULL to send output to log)
2344 static void rfapiRibShowRibSl(void *stream
, struct prefix
*pfx
,
2345 struct skiplist
*sl
)
2347 int (*fp
)(void *, const char *, ...);
2350 const char *vty_newline
;
2352 int nhs_displayed
= 0;
2353 char str_pfx
[PREFIX_STRLEN
];
2354 int printedprefix
= 0;
2356 if (rfapiStream2Vty(stream
, &fp
, &vty
, &out
, &vty_newline
) == 0)
2359 prefix2str(pfx
, str_pfx
, sizeof(str_pfx
));
2362 print_rib_sl(fp
, vty
, out
, sl
, 0, str_pfx
, &printedprefix
);
2366 void rfapiRibShowResponses(void *stream
, struct prefix
*pfx_match
,
2369 int (*fp
)(void *, const char *, ...);
2372 const char *vty_newline
;
2374 struct rfapi_descriptor
*rfd
;
2375 struct listnode
*node
;
2377 struct bgp
*bgp
= bgp_get_default();
2378 int printedheader
= 0;
2379 int routes_total
= 0;
2381 int prefixes_total
= 0;
2382 int prefixes_displayed
= 0;
2384 int nves_with_routes
= 0;
2385 int nves_displayed
= 0;
2386 int routes_displayed
= 0;
2387 int nhs_displayed
= 0;
2389 if (rfapiStream2Vty(stream
, &fp
, &vty
, &out
, &vty_newline
) == 0)
2392 fp(out
, "Unable to find default BGP instance\n");
2399 for (ALL_LIST_ELEMENTS_RO(&bgp
->rfapi
->descriptors
, node
, rfd
)) {
2405 if (rfd
->rib_prefix_count
)
2408 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
2410 struct route_node
*rn
;
2415 for (rn
= route_top(rfd
->rib
[afi
]); rn
;
2416 rn
= route_next(rn
)) {
2418 struct skiplist
*sl
;
2419 char str_pfx
[PREFIX_STRLEN
];
2420 int printedprefix
= 0;
2431 nhs_total
+= skiplist_count(sl
);
2435 && !prefix_match(pfx_match
, &rn
->p
)
2436 && !prefix_match(&rn
->p
, pfx_match
))
2439 ++prefixes_displayed
;
2441 if (!printedheader
) {
2445 show_removed
? "Removed" : "Active");
2446 fp(out
, "%-15s %-15s\n", "Querying VN",
2449 " %-20s %-15s %-15s %4s %-8s %-8s\n",
2450 "Prefix", "Registered VN",
2451 "Registered UN", "Cost", "Lifetime",
2452 #if RFAPI_REGISTRATIONS_REPORT_AGE
2460 char str_vn
[BUFSIZ
];
2461 char str_un
[BUFSIZ
];
2466 fp(out
, "%-15s %-15s\n",
2467 rfapiRfapiIpAddr2Str(&rfd
->vn_addr
,
2469 rfapiRfapiIpAddr2Str(&rfd
->un_addr
,
2473 prefix2str(&rn
->p
, str_pfx
, sizeof(str_pfx
));
2474 // fp(out, " %s\n", buf); /* prefix */
2477 nhs_displayed
+= print_rib_sl(
2478 fp
, vty
, out
, sl
, show_removed
, str_pfx
,
2486 fp(out
, "Displayed %u NVEs, and %u out of %u %s prefixes",
2487 nves_displayed
, routes_displayed
, routes_total
,
2488 show_removed
? "removed" : "active");
2489 if (nhs_displayed
!= routes_displayed
2490 || nhs_total
!= routes_total
)
2491 fp(out
, " with %u out of %u next hops", nhs_displayed
,