1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Copyright 2009-2016, LabN Consulting, L.L.C.
8 * Purpose: Handle import of routes from BGP to RFAPI
11 #include "lib/zebra.h"
12 #include "lib/prefix.h"
13 #include "lib/agg_table.h"
15 #include "lib/memory.h"
17 #include "lib/skiplist.h"
18 #include "lib/thread.h"
19 #include "lib/stream.h"
20 #include "lib/lib_errors.h"
22 #include "bgpd/bgpd.h"
23 #include "bgpd/bgp_ecommunity.h"
24 #include "bgpd/bgp_attr.h"
25 #include "bgpd/bgp_route.h"
26 #include "bgpd/bgp_mplsvpn.h" /* prefix_rd2str() */
27 #include "bgpd/bgp_vnc_types.h"
28 #include "bgpd/bgp_rd.h"
30 #include "bgpd/rfapi/rfapi.h"
31 #include "bgpd/rfapi/bgp_rfapi_cfg.h"
32 #include "bgpd/rfapi/rfapi_backend.h"
33 #include "bgpd/rfapi/rfapi_import.h"
34 #include "bgpd/rfapi/rfapi_private.h"
35 #include "bgpd/rfapi/rfapi_monitor.h"
36 #include "bgpd/rfapi/rfapi_nve_addr.h"
37 #include "bgpd/rfapi/rfapi_vty.h"
38 #include "bgpd/rfapi/vnc_export_bgp.h"
39 #include "bgpd/rfapi/vnc_export_bgp_p.h"
40 #include "bgpd/rfapi/vnc_zebra.h"
41 #include "bgpd/rfapi/vnc_import_bgp.h"
42 #include "bgpd/rfapi/vnc_import_bgp_p.h"
43 #include "bgpd/rfapi/rfapi_rib.h"
44 #include "bgpd/rfapi/rfapi_encap_tlv.h"
45 #include "bgpd/rfapi/vnc_debug.h"
47 #ifdef HAVE_GLIBC_BACKTRACE
48 /* for backtrace and friends */
50 #endif /* HAVE_GLIBC_BACKTRACE */
52 #undef DEBUG_MONITOR_MOVE_SHORTER
53 #undef DEBUG_RETURNED_NHL
54 #undef DEBUG_ROUTE_COUNTERS
55 #undef DEBUG_ENCAP_MONITOR
58 #undef DEBUG_BI_SEARCH
61 * Allocated for each withdraw timer instance; freed when the timer
62 * expires or is canceled
64 struct rfapi_withdraw
{
65 struct rfapi_import_table
*import_table
;
66 struct agg_node
*node
;
67 struct bgp_path_info
*info
;
68 safi_t safi
; /* used only for bulk operations */
70 * For import table node reference count checking (i.e., debugging).
71 * Normally when a timer expires, lockoffset should be 0. However, if
72 * the timer expiration function is called directly (e.g.,
73 * rfapiExpireVpnNow), the node could be locked by a preceding
74 * agg_route_top() or agg_route_next() in a loop, so we need to pass
82 * It's evil and fiendish. It's compiler-dependent.
83 * ? Might need LDFLAGS -rdynamic to produce all function names
85 void rfapiDebugBacktrace(void)
87 #ifdef HAVE_GLIBC_BACKTRACE
88 #define RFAPI_DEBUG_BACKTRACE_NENTRIES 200
89 void *buf
[RFAPI_DEBUG_BACKTRACE_NENTRIES
];
94 size
= backtrace(buf
, RFAPI_DEBUG_BACKTRACE_NENTRIES
);
95 syms
= backtrace_symbols(buf
, size
);
97 for (i
= 0; i
< size
&& i
< RFAPI_DEBUG_BACKTRACE_NENTRIES
; ++i
) {
98 vnc_zlog_debug_verbose("backtrace[%2zu]: %s", i
, syms
[i
]);
108 * Count remote routes and compare with actively-maintained values.
109 * Abort if they disagree.
111 void rfapiCheckRouteCount(void)
113 struct bgp
*bgp
= bgp_get_default();
115 struct rfapi_import_table
*it
;
123 for (it
= h
->imports
; it
; it
= it
->next
) {
124 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
126 struct agg_table
*rt
;
129 int holddown_count
= 0;
131 int imported_count
= 0;
132 int remote_count
= 0;
134 rt
= it
->imported_vpn
[afi
];
136 for (rn
= agg_route_top(rt
); rn
;
137 rn
= agg_route_next(rn
)) {
138 struct bgp_path_info
*bpi
;
139 struct bgp_path_info
*next
;
141 for (bpi
= rn
->info
; bpi
; bpi
= next
) {
144 if (CHECK_FLAG(bpi
->flags
,
149 if (RFAPI_LOCAL_BI(bpi
)) {
152 if (RFAPI_DIRECT_IMPORT_BI(
163 if (it
->holddown_count
[afi
] != holddown_count
) {
164 vnc_zlog_debug_verbose(
165 "%s: it->holddown_count %d != holddown_count %d",
166 __func__
, it
->holddown_count
[afi
],
170 if (it
->remote_count
[afi
] != remote_count
) {
171 vnc_zlog_debug_verbose(
172 "%s: it->remote_count %d != remote_count %d",
173 __func__
, it
->remote_count
[afi
],
177 if (it
->imported_count
[afi
] != imported_count
) {
178 vnc_zlog_debug_verbose(
179 "%s: it->imported_count %d != imported_count %d",
180 __func__
, it
->imported_count
[afi
],
188 #ifdef DEBUG_ROUTE_COUNTERS
189 #define VNC_ITRCCK do {rfapiCheckRouteCount();} while (0)
195 * Validate reference count for a node in an import table
197 * Normally lockoffset is 0 for nodes in quiescent state. However,
198 * agg_unlock_node will delete the node if it is called when
199 * node->lock == 1, and we have to validate the refcount before
200 * the node is deleted. In this case, we specify lockoffset 1.
202 void rfapiCheckRefcount(struct agg_node
*rn
, safi_t safi
, int lockoffset
)
204 unsigned int count_bpi
= 0;
205 unsigned int count_monitor
= 0;
206 struct bgp_path_info
*bpi
;
207 struct rfapi_monitor_encap
*hme
;
208 struct rfapi_monitor_vpn
*hmv
;
210 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
)
215 ++count_monitor
; /* rfapi_it_extra */
222 for (hme
= RFAPI_MONITOR_ENCAP(rn
); hme
;
229 for (hmv
= RFAPI_MONITOR_VPN(rn
); hmv
; hmv
= hmv
->next
)
232 if (RFAPI_MONITOR_EXTERIOR(rn
)->source
) {
233 ++count_monitor
; /* sl */
235 for (rc
= skiplist_next(
236 RFAPI_MONITOR_EXTERIOR(rn
)->source
,
237 NULL
, NULL
, &cursor
);
240 RFAPI_MONITOR_EXTERIOR(rn
)->source
,
241 NULL
, NULL
, &cursor
)) {
243 ++count_monitor
; /* sl entry */
252 case SAFI_LABELED_UNICAST
:
255 assert(!"Passed in safi should be impossible");
259 if (count_bpi
+ count_monitor
+ lockoffset
260 != agg_node_get_lock_count(rn
)) {
261 vnc_zlog_debug_verbose(
262 "%s: count_bpi=%d, count_monitor=%d, lockoffset=%d, rn->lock=%d",
263 __func__
, count_bpi
, count_monitor
, lockoffset
,
264 agg_node_get_lock_count(rn
));
270 * Perform deferred rfapi_close operations that were queued
273 static wq_item_status
rfapi_deferred_close_workfunc(struct work_queue
*q
,
276 struct rfapi_descriptor
*rfd
= data
;
277 struct rfapi
*h
= q
->spec
.data
;
279 assert(!(h
->flags
& RFAPI_INCALLBACK
));
281 vnc_zlog_debug_verbose("%s: completed deferred close on handle %p",
287 * Extract layer 2 option from Encap TLVS in BGP attrs
289 int rfapiGetL2o(struct attr
*attr
, struct rfapi_l2address_option
*l2o
)
292 struct bgp_attr_encap_subtlv
*pEncap
;
294 for (pEncap
= bgp_attr_get_vnc_subtlvs(attr
); pEncap
;
295 pEncap
= pEncap
->next
) {
297 if (pEncap
->type
== BGP_VNC_SUBTLV_TYPE_RFPOPTION
) {
299 == RFAPI_VN_OPTION_TYPE_L2ADDR
) {
301 if (pEncap
->value
[1] == 14) {
302 memcpy(l2o
->macaddr
.octet
,
319 l2o
->logical_net_id
=
322 + ((pEncap
->value
[14]
325 + ((pEncap
->value
[13]
340 * Extract the lifetime from the Tunnel Encap attribute of a route in
343 int rfapiGetVncLifetime(struct attr
*attr
, uint32_t *lifetime
)
345 struct bgp_attr_encap_subtlv
*pEncap
;
347 *lifetime
= RFAPI_INFINITE_LIFETIME
; /* default to infinite */
351 for (pEncap
= bgp_attr_get_vnc_subtlvs(attr
); pEncap
;
352 pEncap
= pEncap
->next
) {
355 == BGP_VNC_SUBTLV_TYPE_LIFETIME
) { /* lifetime */
356 if (pEncap
->length
== 4) {
357 memcpy(lifetime
, pEncap
->value
, 4);
358 *lifetime
= ntohl(*lifetime
);
369 * Look for UN address in Encap attribute
371 int rfapiGetVncTunnelUnAddr(struct attr
*attr
, struct prefix
*p
)
373 struct bgp_attr_encap_subtlv
*pEncap
;
374 bgp_encap_types tun_type
= BGP_ENCAP_TYPE_MPLS
;/*Default tunnel type*/
376 bgp_attr_extcom_tunnel_type(attr
, &tun_type
);
377 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
380 /* MPLS carries UN address in next hop */
381 rfapiNexthop2Prefix(attr
, p
);
382 if (p
->family
!= AF_UNSPEC
)
388 for (pEncap
= attr
->encap_subtlvs
; pEncap
;
389 pEncap
= pEncap
->next
) {
392 == BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
) { /* un
395 switch (pEncap
->length
) {
399 p
->prefixlen
= IPV4_MAX_BITLEN
;
400 memcpy(p
->u
.val
, pEncap
->value
,
407 p
->family
= AF_INET6
;
408 p
->prefixlen
= IPV6_MAX_BITLEN
;
409 memcpy(p
->u
.val
, pEncap
->value
,
422 * Get UN address wherever it might be
424 int rfapiGetUnAddrOfVpnBi(struct bgp_path_info
*bpi
, struct prefix
*p
)
426 /* If it's in this route's VNC attribute, we're done */
427 if (!rfapiGetVncTunnelUnAddr(bpi
->attr
, p
))
430 * Otherwise, see if it's cached from a corresponding ENCAP SAFI
434 switch (bpi
->extra
->vnc
.import
.un_family
) {
437 p
->family
= bpi
->extra
->vnc
.import
.un_family
;
438 p
->u
.prefix4
= bpi
->extra
->vnc
.import
.un
.addr4
;
439 p
->prefixlen
= IPV4_MAX_BITLEN
;
444 p
->family
= bpi
->extra
->vnc
.import
.un_family
;
445 p
->u
.prefix6
= bpi
->extra
->vnc
.import
.un
.addr6
;
446 p
->prefixlen
= IPV6_MAX_BITLEN
;
451 p
->family
= AF_UNSPEC
;
452 #ifdef DEBUG_ENCAP_MONITOR
453 vnc_zlog_debug_verbose(
454 "%s: bpi->extra->vnc.import.un_family is 0, no UN addr",
466 * Make a new bgp_path_info from gathered parameters
468 static struct bgp_path_info
*rfapiBgpInfoCreate(struct attr
*attr
,
469 struct peer
*peer
, void *rfd
,
470 struct prefix_rd
*prd
,
471 uint8_t type
, uint8_t sub_type
,
474 struct bgp_path_info
*new;
476 new = info_make(type
, sub_type
, 0, peer
, attr
, NULL
);
478 new->attr
= bgp_attr_intern(attr
);
480 bgp_path_info_extra_get(new);
482 new->extra
->vnc
.import
.rd
= *prd
;
483 new->extra
->vnc
.import
.create_time
= monotime(NULL
);
486 encode_label(*label
, &new->extra
->label
[0]);
494 * Frees bgp_path_info as used in import tables (parts are not
495 * allocated exactly the way they are in the main RIBs)
497 static void rfapiBgpInfoFree(struct bgp_path_info
*goner
)
503 vnc_zlog_debug_verbose("%s: calling peer_unlock(%p), #%d",
504 __func__
, goner
->peer
,
506 peer_unlock(goner
->peer
);
509 bgp_attr_unintern(&goner
->attr
);
512 bgp_path_info_extra_free(&goner
->extra
);
513 XFREE(MTYPE_BGP_ROUTE
, goner
);
516 struct rfapi_import_table
*rfapiMacImportTableGetNoAlloc(struct bgp
*bgp
,
520 struct rfapi_import_table
*it
= NULL
;
521 uintptr_t lni_as_ptr
= lni
;
530 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
))
536 struct rfapi_import_table
*rfapiMacImportTableGet(struct bgp
*bgp
, uint32_t lni
)
539 struct rfapi_import_table
*it
= NULL
;
540 uintptr_t lni_as_ptr
= lni
;
545 if (!h
->import_mac
) {
546 /* default cmp is good enough for LNI */
547 h
->import_mac
= skiplist_new(0, NULL
, NULL
);
550 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
)) {
552 struct ecommunity
*enew
;
553 struct ecommunity_val eval
;
556 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
557 sizeof(struct rfapi_import_table
));
558 /* set RT list of new import table based on LNI */
559 memset((char *)&eval
, 0, sizeof(eval
));
560 eval
.val
[0] = 0; /* VNC L2VPN */
561 eval
.val
[1] = 2; /* VNC L2VPN */
562 eval
.val
[5] = (lni
>> 16) & 0xff;
563 eval
.val
[6] = (lni
>> 8) & 0xff;
564 eval
.val
[7] = (lni
>> 0) & 0xff;
566 enew
= ecommunity_new();
567 ecommunity_add_val(enew
, &eval
, false, false);
568 it
->rt_import_list
= enew
;
570 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
571 it
->imported_vpn
[afi
] = agg_table_init();
572 it
->imported_encap
[afi
] = agg_table_init();
575 it
->l2_logical_net_id
= lni
;
577 skiplist_insert(h
->import_mac
, (void *)lni_as_ptr
, it
);
585 * Implement MONITOR_MOVE_SHORTER(original_node) from
586 * RFAPI-Import-Event-Handling.txt
588 * Returns pointer to the list of moved monitors
590 static struct rfapi_monitor_vpn
*
591 rfapiMonitorMoveShorter(struct agg_node
*original_vpn_node
, int lockoffset
)
593 struct bgp_path_info
*bpi
;
594 struct agg_node
*par
;
595 struct rfapi_monitor_vpn
*m
;
596 struct rfapi_monitor_vpn
*mlast
;
597 struct rfapi_monitor_vpn
*moved
;
599 int parent_already_refcounted
= 0;
601 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
, lockoffset
);
603 #ifdef DEBUG_MONITOR_MOVE_SHORTER
605 vnc_zlog_debug_verbose("%s: called with node pfx=%pFX",
606 __func__
, &original_vpn_node
->p
);
611 * 1. If there is at least one bpi (either regular route or
612 * route marked as withdrawn, with a pending timer) at
613 * original_node with a valid UN address, we're done. Return.
615 for (bpi
= original_vpn_node
->info
; bpi
; bpi
= bpi
->next
) {
618 if (!rfapiGetUnAddrOfVpnBi(bpi
, &pfx
)) {
619 #ifdef DEBUG_MONITOR_MOVE_SHORTER
620 vnc_zlog_debug_verbose(
621 "%s: have valid UN at original node, no change",
629 * 2. Travel up the tree (toward less-specific prefixes) from
630 * original_node to find the first node that has at least
631 * one route (even if it is only a withdrawn route) with a
632 * valid UN address. Call this node "Node P."
634 for (par
= agg_node_parent(original_vpn_node
); par
;
635 par
= agg_node_parent(par
)) {
636 for (bpi
= par
->info
; bpi
; bpi
= bpi
->next
) {
638 if (!rfapiGetUnAddrOfVpnBi(bpi
, &pfx
)) {
647 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 0);
651 * If no less-specific routes, try to use the 0/0 node
654 const struct prefix
*p
;
655 /* this isn't necessarily 0/0 */
656 par
= agg_route_table_top(original_vpn_node
);
659 p
= agg_node_get_prefix(par
);
661 * If we got the top node but it wasn't 0/0,
664 if (par
&& p
->prefixlen
) {
665 agg_unlock_node(par
); /* maybe free */
670 ++parent_already_refcounted
;
675 * Create 0/0 node if it isn't there
678 struct prefix pfx_default
;
679 const struct prefix
*p
= agg_node_get_prefix(original_vpn_node
);
681 memset(&pfx_default
, 0, sizeof(pfx_default
));
682 pfx_default
.family
= p
->family
;
684 /* creates default node if none exists */
685 par
= agg_node_get(agg_get_table(original_vpn_node
),
687 ++parent_already_refcounted
;
691 * 3. Move each of the monitors found at original_node to Node P.
692 * These are "Moved Monitors."
697 * Attach at end so that the list pointer we return points
698 * only to the moved routes
700 for (m
= RFAPI_MONITOR_VPN(par
), mlast
= NULL
; m
;
701 mlast
= m
, m
= m
->next
)
705 moved
= mlast
->next
= RFAPI_MONITOR_VPN(original_vpn_node
);
707 moved
= RFAPI_MONITOR_VPN_W_ALLOC(par
) =
708 RFAPI_MONITOR_VPN(original_vpn_node
);
710 if (RFAPI_MONITOR_VPN(
711 original_vpn_node
)) /* check agg, so not allocated */
712 RFAPI_MONITOR_VPN_W_ALLOC(original_vpn_node
) = NULL
;
715 * update the node pointers on the monitors
717 for (m
= moved
; m
; m
= m
->next
) {
722 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
,
723 parent_already_refcounted
- movecount
);
724 while (movecount
> parent_already_refcounted
) {
726 ++parent_already_refcounted
;
728 while (movecount
< parent_already_refcounted
) {
729 /* unlikely, but code defensively */
730 agg_unlock_node(par
);
731 --parent_already_refcounted
;
733 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
,
734 movecount
+ lockoffset
);
735 while (movecount
--) {
736 agg_unlock_node(original_vpn_node
);
739 #ifdef DEBUG_MONITOR_MOVE_SHORTER
741 vnc_zlog_debug_verbose("%s: moved to node pfx=%pFX", __func__
,
751 * Implement MONITOR_MOVE_LONGER(new_node) from
752 * RFAPI-Import-Event-Handling.txt
754 static void rfapiMonitorMoveLonger(struct agg_node
*new_vpn_node
)
756 struct rfapi_monitor_vpn
*monitor
;
757 struct rfapi_monitor_vpn
*mlast
;
758 struct bgp_path_info
*bpi
;
759 struct agg_node
*par
;
760 const struct prefix
*new_vpn_node_p
= agg_node_get_prefix(new_vpn_node
);
762 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
765 * Make sure we have at least one valid route at the new node
767 for (bpi
= new_vpn_node
->info
; bpi
; bpi
= bpi
->next
) {
769 if (!rfapiGetUnAddrOfVpnBi(bpi
, &pfx
))
774 vnc_zlog_debug_verbose(
775 "%s: no valid routes at node %p, so not attempting moves",
776 __func__
, new_vpn_node
);
781 * Find first parent node that has monitors
783 for (par
= agg_node_parent(new_vpn_node
); par
;
784 par
= agg_node_parent(par
)) {
785 if (RFAPI_MONITOR_VPN(par
))
790 vnc_zlog_debug_verbose(
791 "%s: no parent nodes with monitors, done", __func__
);
796 * Check each of these monitors to see of their longest-match
797 * is now the updated node. Move any such monitors to the more-
798 * specific updated node
800 for (mlast
= NULL
, monitor
= RFAPI_MONITOR_VPN(par
); monitor
;) {
802 * If new longest match for monitor prefix is the new
803 * route's prefix, move monitor to new route's prefix
805 if (prefix_match(new_vpn_node_p
, &monitor
->p
)) {
808 mlast
->next
= monitor
->next
;
810 RFAPI_MONITOR_VPN_W_ALLOC(par
) = monitor
->next
;
815 monitor
->next
= RFAPI_MONITOR_VPN(new_vpn_node
);
816 RFAPI_MONITOR_VPN_W_ALLOC(new_vpn_node
) = monitor
;
817 monitor
->node
= new_vpn_node
;
819 agg_lock_node(new_vpn_node
); /* incr refcount */
821 monitor
= mlast
? mlast
->next
: RFAPI_MONITOR_VPN(par
);
823 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 1);
824 /* decr refcount after we're done with par as this might
826 agg_unlock_node(par
);
831 monitor
= monitor
->next
;
834 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
838 static void rfapiBgpInfoChainFree(struct bgp_path_info
*bpi
)
840 struct bgp_path_info
*next
;
845 * If there is a timer waiting to delete this bpi, cancel
846 * the timer and delete immediately
848 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
849 && bpi
->extra
->vnc
.import
.timer
) {
850 struct rfapi_withdraw
*wcb
=
851 THREAD_ARG(bpi
->extra
->vnc
.import
.timer
);
853 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
854 THREAD_OFF(bpi
->extra
->vnc
.import
.timer
);
859 rfapiBgpInfoFree(bpi
);
864 static void rfapiImportTableFlush(struct rfapi_import_table
*it
)
871 ecommunity_free(&it
->rt_import_list
);
872 it
->rt_import_list
= NULL
;
874 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
877 struct agg_table
*at
;
879 at
= it
->imported_vpn
[afi
];
881 for (rn
= agg_route_top(at
); rn
;
882 rn
= agg_route_next(rn
)) {
884 * Each route_node has:
885 * aggregate: points to rfapi_it_extra with
887 * info: points to chain of bgp_path_info
889 /* free bgp_path_info and its children */
890 rfapiBgpInfoChainFree(rn
->info
);
893 rfapiMonitorExtraFlush(SAFI_MPLS_VPN
, rn
);
895 agg_table_finish(at
);
899 at
= it
->imported_encap
[afi
];
900 for (rn
= agg_route_top(at
); rn
;
901 rn
= agg_route_next(rn
)) {
902 /* free bgp_path_info and its children */
903 rfapiBgpInfoChainFree(rn
->info
);
906 rfapiMonitorExtraFlush(SAFI_ENCAP
, rn
);
908 agg_table_finish(at
);
911 if (it
->monitor_exterior_orphans
) {
912 skiplist_free(it
->monitor_exterior_orphans
);
916 void rfapiImportTableRefDelByIt(struct bgp
*bgp
,
917 struct rfapi_import_table
*it_target
)
920 struct rfapi_import_table
*it
;
921 struct rfapi_import_table
*prev
= NULL
;
928 for (it
= h
->imports
; it
; prev
= it
, it
= it
->next
) {
934 assert(it
->refcount
);
940 prev
->next
= it
->next
;
942 h
->imports
= it
->next
;
944 rfapiImportTableFlush(it
);
945 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
949 #ifdef RFAPI_REQUIRE_ENCAP_BEEC
951 * Look for magic BGP Encapsulation Extended Community value
952 * Format in RFC 5512 Sect. 4.5
954 static int rfapiEcommunitiesMatchBeec(struct ecommunity
*ecom
,
955 bgp_encap_types type
)
962 for (i
= 0; i
< (ecom
->size
* ECOMMUNITY_SIZE
); i
+= ECOMMUNITY_SIZE
) {
968 if (ep
[0] == ECOMMUNITY_ENCODE_OPAQUE
969 && ep
[1] == ECOMMUNITY_OPAQUE_SUBTYPE_ENCAP
970 && ep
[6] == ((type
&& 0xff00) >> 8)
971 && ep
[7] == (type
& 0xff)) {
980 int rfapiEcommunitiesIntersect(struct ecommunity
*e1
, struct ecommunity
*e2
)
989 s1
= ecommunity_ecom2str(e1
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
990 s2
= ecommunity_ecom2str(e2
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
991 vnc_zlog_debug_verbose("%s: e1[%s], e2[%s]", __func__
, s1
, s2
);
992 XFREE(MTYPE_ECOMMUNITY_STR
, s1
);
993 XFREE(MTYPE_ECOMMUNITY_STR
, s2
);
996 for (i
= 0; i
< e1
->size
; ++i
) {
997 for (j
= 0; j
< e2
->size
; ++j
) {
998 if (!memcmp(e1
->val
+ (i
* ECOMMUNITY_SIZE
),
999 e2
->val
+ (j
* ECOMMUNITY_SIZE
),
1009 int rfapiEcommunityGetLNI(struct ecommunity
*ecom
, uint32_t *lni
)
1014 for (i
= 0; i
< ecom
->size
; ++i
) {
1015 uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1017 if ((*(p
+ 0) == 0x00) && (*(p
+ 1) == 0x02)) {
1019 *lni
= (*(p
+ 5) << 16) | (*(p
+ 6) << 8)
1028 int rfapiEcommunityGetEthernetTag(struct ecommunity
*ecom
, uint16_t *tag_id
)
1030 struct bgp
*bgp
= bgp_get_default();
1031 *tag_id
= 0; /* default to untagged */
1035 for (i
= 0; i
< ecom
->size
; ++i
) {
1038 const uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1040 /* High-order octet of type. */
1043 if (*p
++ == ECOMMUNITY_ROUTE_TARGET
) {
1044 if (encode
== ECOMMUNITY_ENCODE_AS4
) {
1045 p
= ptr_get_be32(p
, &as
);
1046 } else if (encode
== ECOMMUNITY_ENCODE_AS
) {
1049 p
+= 2; /* skip next two, tag/vid
1050 always in lowest bytes */
1052 if (as
== bgp
->as
) {
1053 *tag_id
= *p
++ << 8;
1063 static int rfapiVpnBiNhEqualsPt(struct bgp_path_info
*bpi
,
1064 struct rfapi_ip_addr
*hpt
)
1071 family
= BGP_MP_NEXTHOP_FAMILY(bpi
->attr
->mp_nexthop_len
);
1073 if (hpt
->addr_family
!= family
)
1078 if (bpi
->attr
->mp_nexthop_global_in
.s_addr
1079 != hpt
->addr
.v4
.s_addr
)
1084 if (IPV6_ADDR_CMP(&bpi
->attr
->mp_nexthop_global
, &hpt
->addr
.v6
))
1097 * Compare 2 VPN BIs. Return true if they have the same VN and UN addresses
1099 static int rfapiVpnBiSamePtUn(struct bgp_path_info
*bpi1
,
1100 struct bgp_path_info
*bpi2
)
1102 struct prefix pfx_un1
;
1103 struct prefix pfx_un2
;
1109 * VN address comparisons
1112 if (BGP_MP_NEXTHOP_FAMILY(bpi1
->attr
->mp_nexthop_len
)
1113 != BGP_MP_NEXTHOP_FAMILY(bpi2
->attr
->mp_nexthop_len
)) {
1117 switch (BGP_MP_NEXTHOP_FAMILY(bpi1
->attr
->mp_nexthop_len
)) {
1119 if (bpi1
->attr
->mp_nexthop_global_in
.s_addr
1120 != bpi2
->attr
->mp_nexthop_global_in
.s_addr
)
1125 if (IPV6_ADDR_CMP(&bpi1
->attr
->mp_nexthop_global
,
1126 &bpi2
->attr
->mp_nexthop_global
))
1134 memset(&pfx_un1
, 0, sizeof(pfx_un1
));
1135 memset(&pfx_un2
, 0, sizeof(pfx_un2
));
1138 * UN address comparisons
1140 if (rfapiGetVncTunnelUnAddr(bpi1
->attr
, &pfx_un1
)) {
1142 pfx_un1
.family
= bpi1
->extra
->vnc
.import
.un_family
;
1143 switch (bpi1
->extra
->vnc
.import
.un_family
) {
1146 bpi1
->extra
->vnc
.import
.un
.addr4
;
1150 bpi1
->extra
->vnc
.import
.un
.addr6
;
1153 pfx_un1
.family
= AF_UNSPEC
;
1159 if (rfapiGetVncTunnelUnAddr(bpi2
->attr
, &pfx_un2
)) {
1161 pfx_un2
.family
= bpi2
->extra
->vnc
.import
.un_family
;
1162 switch (bpi2
->extra
->vnc
.import
.un_family
) {
1165 bpi2
->extra
->vnc
.import
.un
.addr4
;
1169 bpi2
->extra
->vnc
.import
.un
.addr6
;
1172 pfx_un2
.family
= AF_UNSPEC
;
1178 if (pfx_un1
.family
== AF_UNSPEC
|| pfx_un2
.family
== AF_UNSPEC
)
1181 if (pfx_un1
.family
!= pfx_un2
.family
)
1184 switch (pfx_un1
.family
) {
1186 if (!IPV4_ADDR_SAME(&pfx_un1
.u
.prefix4
, &pfx_un2
.u
.prefix4
))
1190 if (!IPV6_ADDR_SAME(&pfx_un1
.u
.prefix6
, &pfx_un2
.u
.prefix6
))
1199 uint8_t rfapiRfpCost(struct attr
*attr
)
1201 if (attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_LOCAL_PREF
)) {
1202 if (attr
->local_pref
> 255) {
1205 return 255 - attr
->local_pref
;
1211 /*------------------------------------------
1214 * Find Layer 2 options in an option chain
1220 * l2o layer 2 options extracted
1224 * 1 no options found
1226 --------------------------------------------*/
1227 int rfapi_extract_l2o(
1228 struct bgp_tea_options
*pHop
, /* chain of options */
1229 struct rfapi_l2address_option
*l2o
) /* return extracted value */
1231 struct bgp_tea_options
*p
;
1233 for (p
= pHop
; p
; p
= p
->next
) {
1234 if ((p
->type
== RFAPI_VN_OPTION_TYPE_L2ADDR
)
1235 && (p
->length
>= 8)) {
1239 memcpy(&l2o
->macaddr
, v
, 6);
1241 l2o
->label
= ((v
[6] << 12) & 0xff000)
1242 + ((v
[7] << 4) & 0xff0)
1243 + ((v
[8] >> 4) & 0xf);
1245 l2o
->local_nve_id
= (uint8_t)v
[10];
1247 l2o
->logical_net_id
=
1248 (v
[11] << 16) + (v
[12] << 8) + (v
[13] << 0);
1256 static struct rfapi_next_hop_entry
*
1257 rfapiRouteInfo2NextHopEntry(struct rfapi_ip_prefix
*rprefix
,
1258 struct bgp_path_info
*bpi
, /* route to encode */
1259 uint32_t lifetime
, /* use this in nhe */
1260 struct agg_node
*rn
) /* req for L2 eth addr */
1262 struct rfapi_next_hop_entry
*new;
1263 int have_vnc_tunnel_un
= 0;
1264 const struct prefix
*p
= agg_node_get_prefix(rn
);
1266 #ifdef DEBUG_ENCAP_MONITOR
1267 vnc_zlog_debug_verbose("%s: entry, bpi %p, rn %p", __func__
, bpi
, rn
);
1270 new = XCALLOC(MTYPE_RFAPI_NEXTHOP
, sizeof(struct rfapi_next_hop_entry
));
1272 new->prefix
= *rprefix
;
1275 && decode_rd_type(bpi
->extra
->vnc
.import
.rd
.val
)
1276 == RD_TYPE_VNC_ETH
) {
1279 struct rfapi_vn_option
*vo
;
1281 vo
= XCALLOC(MTYPE_RFAPI_VN_OPTION
,
1282 sizeof(struct rfapi_vn_option
));
1284 vo
->type
= RFAPI_VN_OPTION_TYPE_L2ADDR
;
1286 memcpy(&vo
->v
.l2addr
.macaddr
, &p
->u
.prefix_eth
.octet
, ETH_ALEN
);
1287 /* only low 3 bytes of this are significant */
1288 (void)rfapiEcommunityGetLNI(bgp_attr_get_ecommunity(bpi
->attr
),
1289 &vo
->v
.l2addr
.logical_net_id
);
1290 (void)rfapiEcommunityGetEthernetTag(
1291 bgp_attr_get_ecommunity(bpi
->attr
),
1292 &vo
->v
.l2addr
.tag_id
);
1294 /* local_nve_id comes from lower byte of RD type */
1295 vo
->v
.l2addr
.local_nve_id
= bpi
->extra
->vnc
.import
.rd
.val
[1];
1297 /* label comes from MP_REACH_NLRI label */
1298 vo
->v
.l2addr
.label
= decode_label(&bpi
->extra
->label
[0]);
1300 new->vn_options
= vo
;
1303 * If there is an auxiliary prefix (i.e., host IP address),
1304 * use it as the nexthop prefix instead of the query prefix
1306 if (bpi
->extra
->vnc
.import
.aux_prefix
.family
) {
1307 rfapiQprefix2Rprefix(&bpi
->extra
->vnc
.import
.aux_prefix
,
1312 bgp_encap_types tun_type
= BGP_ENCAP_TYPE_MPLS
; /*Default*/
1313 new->prefix
.cost
= rfapiRfpCost(bpi
->attr
);
1315 struct bgp_attr_encap_subtlv
*pEncap
;
1317 switch (BGP_MP_NEXTHOP_FAMILY(bpi
->attr
->mp_nexthop_len
)) {
1319 new->vn_address
.addr_family
= AF_INET
;
1320 new->vn_address
.addr
.v4
= bpi
->attr
->mp_nexthop_global_in
;
1324 new->vn_address
.addr_family
= AF_INET6
;
1325 new->vn_address
.addr
.v6
= bpi
->attr
->mp_nexthop_global
;
1329 zlog_warn("%s: invalid vpn nexthop length: %d", __func__
,
1330 bpi
->attr
->mp_nexthop_len
);
1331 rfapi_free_next_hop_list(new);
1335 for (pEncap
= bgp_attr_get_vnc_subtlvs(bpi
->attr
); pEncap
;
1336 pEncap
= pEncap
->next
) {
1337 switch (pEncap
->type
) {
1338 case BGP_VNC_SUBTLV_TYPE_LIFETIME
:
1339 /* use configured lifetime, not attr lifetime */
1343 zlog_warn("%s: unknown VNC option type %d", __func__
,
1350 bgp_attr_extcom_tunnel_type(bpi
->attr
, &tun_type
);
1351 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
1353 /* MPLS carries UN address in next hop */
1354 rfapiNexthop2Prefix(bpi
->attr
, &p
);
1355 if (p
.family
!= AF_UNSPEC
) {
1356 rfapiQprefix2Raddr(&p
, &new->un_address
);
1357 have_vnc_tunnel_un
= 1;
1361 for (pEncap
= bpi
->attr
->encap_subtlvs
; pEncap
; pEncap
= pEncap
->next
) {
1362 switch (pEncap
->type
) {
1363 case BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
:
1365 * Overrides ENCAP UN address, if any
1367 switch (pEncap
->length
) {
1370 new->un_address
.addr_family
= AF_INET
;
1371 memcpy(&new->un_address
.addr
.v4
, pEncap
->value
,
1373 have_vnc_tunnel_un
= 1;
1377 new->un_address
.addr_family
= AF_INET6
;
1378 memcpy(&new->un_address
.addr
.v6
, pEncap
->value
,
1380 have_vnc_tunnel_un
= 1;
1385 "%s: invalid tunnel subtlv UN addr length (%d) for bpi %p",
1386 __func__
, pEncap
->length
, bpi
);
1391 zlog_warn("%s: unknown Encap Attribute option type %d",
1392 __func__
, pEncap
->type
);
1397 new->un_options
= rfapi_encap_tlv_to_un_option(bpi
->attr
);
1399 #ifdef DEBUG_ENCAP_MONITOR
1400 vnc_zlog_debug_verbose("%s: line %d: have_vnc_tunnel_un=%d", __func__
,
1401 __LINE__
, have_vnc_tunnel_un
);
1404 if (!have_vnc_tunnel_un
&& bpi
->extra
) {
1406 * use cached UN address from ENCAP route
1408 new->un_address
.addr_family
= bpi
->extra
->vnc
.import
.un_family
;
1409 switch (new->un_address
.addr_family
) {
1411 new->un_address
.addr
.v4
=
1412 bpi
->extra
->vnc
.import
.un
.addr4
;
1415 new->un_address
.addr
.v6
=
1416 bpi
->extra
->vnc
.import
.un
.addr6
;
1419 zlog_warn("%s: invalid UN addr family (%d) for bpi %p",
1420 __func__
, new->un_address
.addr_family
, bpi
);
1421 rfapi_free_next_hop_list(new);
1426 new->lifetime
= lifetime
;
1430 int rfapiHasNonRemovedRoutes(struct agg_node
*rn
)
1432 struct bgp_path_info
*bpi
;
1434 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
1437 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
1438 && (bpi
->extra
&& !rfapiGetUnAddrOfVpnBi(bpi
, &pfx
))) {
1446 #ifdef DEBUG_IT_NODES
1450 void rfapiDumpNode(struct agg_node
*rn
)
1452 struct bgp_path_info
*bpi
;
1454 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
1455 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
1457 int ctrc
= rfapiGetUnAddrOfVpnBi(bpi
, &pfx
);
1460 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
1461 && (bpi
->extra
&& !ctrc
)) {
1468 vnc_zlog_debug_verbose(
1469 " bpi=%p, nr=%d, flags=0x%x, extra=%p, ctrc=%d", bpi
,
1470 nr
, bpi
->flags
, bpi
->extra
, ctrc
);
1475 static int rfapiNhlAddNodeRoutes(
1476 struct agg_node
*rn
, /* in */
1477 struct rfapi_ip_prefix
*rprefix
, /* in */
1478 uint32_t lifetime
, /* in */
1479 int removed
, /* in */
1480 struct rfapi_next_hop_entry
**head
, /* in/out */
1481 struct rfapi_next_hop_entry
**tail
, /* in/out */
1482 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1483 struct agg_node
*rfd_rib_node
, /* preload this NVE rib node */
1484 struct prefix
*pfx_target_original
) /* query target */
1486 struct bgp_path_info
*bpi
;
1487 struct rfapi_next_hop_entry
*new;
1488 struct prefix pfx_un
;
1489 struct skiplist
*seen_nexthops
;
1491 const struct prefix
*p
= agg_node_get_prefix(rn
);
1492 int is_l2
= (p
->family
== AF_ETHERNET
);
1495 struct agg_table
*atable
= agg_get_table(rfd_rib_node
);
1496 struct rfapi_descriptor
*rfd
;
1499 rfd
= agg_get_table_info(atable
);
1501 if (rfapiRibFTDFilterRecentPrefix(rfd
, rn
,
1502 pfx_target_original
))
1508 skiplist_new(0, vnc_prefix_cmp
, prefix_free_lists
);
1510 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
1512 struct prefix pfx_vn
;
1513 struct prefix
*newpfx
;
1515 if (removed
&& !CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
1516 #ifdef DEBUG_RETURNED_NHL
1517 vnc_zlog_debug_verbose(
1518 "%s: want holddown, this route not holddown, skip",
1523 if (!removed
&& CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
1532 * Check for excluded VN address
1534 if (rfapiVpnBiNhEqualsPt(bpi
, exclude_vnaddr
))
1538 * Check for VN address (nexthop) copied already
1541 /* L2 routes: semantic nexthop in aux_prefix; VN addr
1543 pfx_vn
= bpi
->extra
->vnc
.import
.aux_prefix
;
1545 rfapiNexthop2Prefix(bpi
->attr
, &pfx_vn
);
1547 if (!skiplist_search(seen_nexthops
, &pfx_vn
, NULL
)) {
1548 #ifdef DEBUG_RETURNED_NHL
1549 vnc_zlog_debug_verbose(
1550 "%s: already put VN/nexthop %pFX, skip",
1556 if (rfapiGetUnAddrOfVpnBi(bpi
, &pfx_un
)) {
1557 #ifdef DEBUG_ENCAP_MONITOR
1558 vnc_zlog_debug_verbose(
1559 "%s: failed to get UN address of this VPN bpi",
1565 newpfx
= prefix_new();
1567 skiplist_insert(seen_nexthops
, newpfx
, newpfx
);
1569 new = rfapiRouteInfo2NextHopEntry(rprefix
, bpi
, lifetime
, rn
);
1571 if (rfapiRibPreloadBi(rfd_rib_node
, &pfx_vn
, &pfx_un
,
1573 /* duplicate filtered by RIB */
1574 rfapi_free_next_hop_list(new);
1581 (*tail
)->next
= new;
1590 skiplist_free(seen_nexthops
);
1599 * omit_node is meant for the situation where we are adding a subtree
1600 * of a parent of some original requested node. The response already
1601 * contains the original requested node, and we don't want to duplicate
1602 * its routes in the list, so we skip it if the right or left node
1603 * matches (of course, we still travel down its child subtrees).
1605 static int rfapiNhlAddSubtree(
1606 struct agg_node
*rn
, /* in */
1607 uint32_t lifetime
, /* in */
1608 struct rfapi_next_hop_entry
**head
, /* in/out */
1609 struct rfapi_next_hop_entry
**tail
, /* in/out */
1610 struct agg_node
*omit_node
, /* in */
1611 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1612 struct agg_table
*rfd_rib_table
, /* preload here */
1613 struct prefix
*pfx_target_original
) /* query target */
1615 struct rfapi_ip_prefix rprefix
;
1618 /* FIXME: need to find a better way here to work without sticking our
1619 * hands in node->link */
1620 if (agg_node_left(rn
) && agg_node_left(rn
) != omit_node
) {
1621 if (agg_node_left(rn
)->info
) {
1622 const struct prefix
*p
=
1623 agg_node_get_prefix(agg_node_left(rn
));
1625 struct agg_node
*rib_rn
= NULL
;
1627 rfapiQprefix2Rprefix(p
, &rprefix
);
1629 rib_rn
= agg_node_get(rfd_rib_table
, p
);
1631 count
= rfapiNhlAddNodeRoutes(
1632 agg_node_left(rn
), &rprefix
, lifetime
, 0, head
,
1633 tail
, exclude_vnaddr
, rib_rn
,
1634 pfx_target_original
);
1636 count
= rfapiNhlAddNodeRoutes(
1637 agg_node_left(rn
), &rprefix
, lifetime
,
1638 1, head
, tail
, exclude_vnaddr
, rib_rn
,
1639 pfx_target_original
);
1643 agg_unlock_node(rib_rn
);
1647 if (agg_node_right(rn
) && agg_node_right(rn
) != omit_node
) {
1648 if (agg_node_right(rn
)->info
) {
1649 const struct prefix
*p
=
1650 agg_node_get_prefix(agg_node_right(rn
));
1652 struct agg_node
*rib_rn
= NULL
;
1654 rfapiQprefix2Rprefix(p
, &rprefix
);
1656 rib_rn
= agg_node_get(rfd_rib_table
, p
);
1658 count
= rfapiNhlAddNodeRoutes(
1659 agg_node_right(rn
), &rprefix
, lifetime
, 0, head
,
1660 tail
, exclude_vnaddr
, rib_rn
,
1661 pfx_target_original
);
1663 count
= rfapiNhlAddNodeRoutes(
1664 agg_node_right(rn
), &rprefix
, lifetime
,
1665 1, head
, tail
, exclude_vnaddr
, rib_rn
,
1666 pfx_target_original
);
1670 agg_unlock_node(rib_rn
);
1674 if (agg_node_left(rn
)) {
1675 rcount
+= rfapiNhlAddSubtree(
1676 agg_node_left(rn
), lifetime
, head
, tail
, omit_node
,
1677 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1679 if (agg_node_right(rn
)) {
1680 rcount
+= rfapiNhlAddSubtree(
1681 agg_node_right(rn
), lifetime
, head
, tail
, omit_node
,
1682 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1689 * Implementation of ROUTE_LIST(node) from RFAPI-Import-Event-Handling.txt
1691 * Construct an rfapi nexthop list based on the routes attached to
1692 * the specified node.
1694 * If there are any routes that do NOT have BGP_PATH_REMOVED set,
1695 * return those only. If there are ONLY routes with BGP_PATH_REMOVED,
1696 * then return those, and also include all the non-removed routes from the
1697 * next less-specific node (i.e., this node's parent) at the end.
1699 struct rfapi_next_hop_entry
*rfapiRouteNode2NextHopList(
1700 struct agg_node
*rn
, uint32_t lifetime
, /* put into nexthop entries */
1701 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1702 struct agg_table
*rfd_rib_table
, /* preload here */
1703 struct prefix
*pfx_target_original
) /* query target */
1705 struct rfapi_ip_prefix rprefix
;
1706 struct rfapi_next_hop_entry
*answer
= NULL
;
1707 struct rfapi_next_hop_entry
*last
= NULL
;
1708 struct agg_node
*parent
;
1709 const struct prefix
*p
= agg_node_get_prefix(rn
);
1711 struct agg_node
*rib_rn
;
1713 #ifdef DEBUG_RETURNED_NHL
1714 vnc_zlog_debug_verbose("%s: called with node pfx=%rRN", __func__
, rn
);
1715 rfapiDebugBacktrace();
1718 rfapiQprefix2Rprefix(p
, &rprefix
);
1720 rib_rn
= rfd_rib_table
? agg_node_get(rfd_rib_table
, p
) : NULL
;
1723 * Add non-withdrawn routes at this node
1725 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 0, &answer
, &last
,
1726 exclude_vnaddr
, rib_rn
,
1727 pfx_target_original
);
1730 * If the list has at least one entry, it's finished
1733 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
, &last
, NULL
,
1734 exclude_vnaddr
, rfd_rib_table
,
1735 pfx_target_original
);
1736 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
,
1738 #ifdef DEBUG_RETURNED_NHL
1739 rfapiPrintNhl(NULL
, answer
);
1742 agg_unlock_node(rib_rn
);
1747 * Add withdrawn routes at this node
1749 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 1, &answer
, &last
,
1750 exclude_vnaddr
, rib_rn
,
1751 pfx_target_original
);
1753 agg_unlock_node(rib_rn
);
1755 // rfapiPrintNhl(NULL, answer);
1758 * walk up the tree until we find a node with non-deleted
1759 * routes, then add them
1761 for (parent
= agg_node_parent(rn
); parent
;
1762 parent
= agg_node_parent(parent
)) {
1763 if (rfapiHasNonRemovedRoutes(parent
)) {
1769 * Add non-withdrawn routes from less-specific prefix
1772 const struct prefix
*p
= agg_node_get_prefix(parent
);
1774 rib_rn
= rfd_rib_table
? agg_node_get(rfd_rib_table
, p
) : NULL
;
1775 rfapiQprefix2Rprefix(p
, &rprefix
);
1776 count
+= rfapiNhlAddNodeRoutes(parent
, &rprefix
, lifetime
, 0,
1777 &answer
, &last
, exclude_vnaddr
,
1778 rib_rn
, pfx_target_original
);
1779 count
+= rfapiNhlAddSubtree(parent
, lifetime
, &answer
, &last
,
1780 rn
, exclude_vnaddr
, rfd_rib_table
,
1781 pfx_target_original
);
1783 agg_unlock_node(rib_rn
);
1786 * There is no parent with non-removed routes. Still need to
1787 * add subtree of original node if it contributed routes to the
1791 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
,
1792 &last
, rn
, exclude_vnaddr
,
1794 pfx_target_original
);
1797 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
, count
,
1799 #ifdef DEBUG_RETURNED_NHL
1800 rfapiPrintNhl(NULL
, answer
);
1806 * Construct nexthop list of all routes in table
1808 struct rfapi_next_hop_entry
*rfapiRouteTable2NextHopList(
1809 struct agg_table
*rt
, uint32_t lifetime
, /* put into nexthop entries */
1810 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1811 struct agg_table
*rfd_rib_table
, /* preload this NVE rib table */
1812 struct prefix
*pfx_target_original
) /* query target */
1814 struct agg_node
*rn
;
1815 struct rfapi_next_hop_entry
*biglist
= NULL
;
1816 struct rfapi_next_hop_entry
*nhl
;
1817 struct rfapi_next_hop_entry
*tail
= NULL
;
1820 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
1822 nhl
= rfapiRouteNode2NextHopList(rn
, lifetime
, exclude_vnaddr
,
1824 pfx_target_original
);
1826 tail
= biglist
= nhl
;
1833 while (tail
->next
) {
1840 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1844 struct rfapi_next_hop_entry
*rfapiEthRouteNode2NextHopList(
1845 struct agg_node
*rn
, struct rfapi_ip_prefix
*rprefix
,
1846 uint32_t lifetime
, /* put into nexthop entries */
1847 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1848 struct agg_table
*rfd_rib_table
, /* preload NVE rib table */
1849 struct prefix
*pfx_target_original
) /* query target */
1852 struct rfapi_next_hop_entry
*answer
= NULL
;
1853 struct rfapi_next_hop_entry
*last
= NULL
;
1854 struct agg_node
*rib_rn
;
1856 rib_rn
= rfd_rib_table
1857 ? agg_node_get(rfd_rib_table
, agg_node_get_prefix(rn
))
1860 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 0, &answer
, &last
,
1861 NULL
, rib_rn
, pfx_target_original
);
1863 #ifdef DEBUG_ENCAP_MONITOR
1864 vnc_zlog_debug_verbose("%s: node %p: %d non-holddown routes", __func__
,
1869 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 1, &answer
,
1870 &last
, exclude_vnaddr
, rib_rn
,
1871 pfx_target_original
);
1872 vnc_zlog_debug_verbose("%s: node %p: %d holddown routes",
1873 __func__
, rn
, count
);
1877 agg_unlock_node(rib_rn
);
1879 #ifdef DEBUG_RETURNED_NHL
1880 rfapiPrintNhl(NULL
, answer
);
1888 * Construct nexthop list of all routes in table
1890 struct rfapi_next_hop_entry
*rfapiEthRouteTable2NextHopList(
1891 uint32_t logical_net_id
, struct rfapi_ip_prefix
*rprefix
,
1892 uint32_t lifetime
, /* put into nexthop entries */
1893 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1894 struct agg_table
*rfd_rib_table
, /* preload NVE rib node */
1895 struct prefix
*pfx_target_original
) /* query target */
1897 struct rfapi_import_table
*it
;
1898 struct bgp
*bgp
= bgp_get_default();
1899 struct agg_table
*rt
;
1900 struct agg_node
*rn
;
1901 struct rfapi_next_hop_entry
*biglist
= NULL
;
1902 struct rfapi_next_hop_entry
*nhl
;
1903 struct rfapi_next_hop_entry
*tail
= NULL
;
1907 it
= rfapiMacImportTableGet(bgp
, logical_net_id
);
1908 rt
= it
->imported_vpn
[AFI_L2VPN
];
1910 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
1912 nhl
= rfapiEthRouteNode2NextHopList(
1913 rn
, rprefix
, lifetime
, exclude_vnaddr
, rfd_rib_table
,
1914 pfx_target_original
);
1916 tail
= biglist
= nhl
;
1923 while (tail
->next
) {
1930 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1935 * Insert a new bpi to the imported route table node,
1936 * keeping the list of BPIs sorted best route first
1938 static void rfapiBgpInfoAttachSorted(struct agg_node
*rn
,
1939 struct bgp_path_info
*info_new
, afi_t afi
,
1943 struct bgp_path_info
*prev
;
1944 struct bgp_path_info
*next
;
1945 char pfx_buf
[PREFIX2STR_BUFFER
];
1948 bgp
= bgp_get_default(); /* assume 1 instance for now */
1950 if (VNC_DEBUG(IMPORT_BI_ATTACH
)) {
1951 vnc_zlog_debug_verbose("%s: info_new->peer=%p", __func__
,
1953 vnc_zlog_debug_verbose("%s: info_new->peer->su_remote=%p",
1954 __func__
, info_new
->peer
->su_remote
);
1957 for (prev
= NULL
, next
= rn
->info
; next
;
1958 prev
= next
, next
= next
->next
) {
1959 enum bgp_path_selection_reason reason
;
1962 || (!CHECK_FLAG(info_new
->flags
, BGP_PATH_REMOVED
)
1963 && CHECK_FLAG(next
->flags
, BGP_PATH_REMOVED
))
1964 || bgp_path_info_cmp_compatible(bgp
, info_new
, next
,
1967 == -1) { /* -1 if 1st is better */
1971 vnc_zlog_debug_verbose("%s: prev=%p, next=%p", __func__
, prev
, next
);
1973 prev
->next
= info_new
;
1975 rn
->info
= info_new
;
1977 info_new
->prev
= prev
;
1978 info_new
->next
= next
;
1980 next
->prev
= info_new
;
1981 bgp_attr_intern(info_new
->attr
);
1984 static void rfapiBgpInfoDetach(struct agg_node
*rn
, struct bgp_path_info
*bpi
)
1987 * Remove the route (doubly-linked)
1989 // bgp_attr_unintern (&bpi->attr);
1991 bpi
->next
->prev
= bpi
->prev
;
1993 bpi
->prev
->next
= bpi
->next
;
1995 rn
->info
= bpi
->next
;
1999 * For L3-indexed import tables
2001 static int rfapi_bi_peer_rd_cmp(const void *b1
, const void *b2
)
2003 const struct bgp_path_info
*bpi1
= b1
;
2004 const struct bgp_path_info
*bpi2
= b2
;
2009 if (bpi1
->peer
< bpi2
->peer
)
2011 if (bpi1
->peer
> bpi2
->peer
)
2017 return vnc_prefix_cmp(
2018 (const struct prefix
*)&bpi1
->extra
->vnc
.import
.rd
,
2019 (const struct prefix
*)&bpi2
->extra
->vnc
.import
.rd
);
2023 * For L2-indexed import tables
2024 * The BPIs in these tables should ALWAYS have an aux_prefix set because
2025 * they arrive via IPv4 or IPv6 advertisements.
2027 static int rfapi_bi_peer_rd_aux_cmp(const void *b1
, const void *b2
)
2029 const struct bgp_path_info
*bpi1
= b1
;
2030 const struct bgp_path_info
*bpi2
= b2
;
2036 if (bpi1
->peer
< bpi2
->peer
)
2038 if (bpi1
->peer
> bpi2
->peer
)
2044 rc
= vnc_prefix_cmp((struct prefix
*)&bpi1
->extra
->vnc
.import
.rd
,
2045 (struct prefix
*)&bpi2
->extra
->vnc
.import
.rd
);
2051 * L2 import tables can have multiple entries with the
2052 * same MAC address, same RD, but different L3 addresses.
2054 * Use presence of aux_prefix with AF=ethernet and prefixlen=1
2055 * as magic value to signify explicit wildcarding of the aux_prefix.
2056 * This magic value will not appear in bona fide bpi entries in
2057 * the import table, but is allowed in the "fake" bpi used to
2058 * probe the table when searching. (We have to test both b1 and b2
2059 * because there is no guarantee of the order the test key and
2060 * the real key will be passed)
2062 if ((bpi1
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2063 && (bpi1
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))
2064 || (bpi2
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2065 && (bpi2
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))) {
2068 * wildcard aux address specified
2073 return vnc_prefix_cmp(&bpi1
->extra
->vnc
.import
.aux_prefix
,
2074 &bpi2
->extra
->vnc
.import
.aux_prefix
);
2079 * Index on RD and Peer
2081 static void rfapiItBiIndexAdd(struct agg_node
*rn
, /* Import table VPN node */
2082 struct bgp_path_info
*bpi
) /* new BPI */
2084 struct skiplist
*sl
;
2085 const struct prefix
*p
;
2091 vnc_zlog_debug_verbose("%s: bpi %p, peer %p, rd %pRDP", __func__
, bpi
,
2092 bpi
->peer
, &bpi
->extra
->vnc
.import
.rd
);
2094 sl
= RFAPI_RDINDEX_W_ALLOC(rn
);
2096 p
= agg_node_get_prefix(rn
);
2097 if (AF_ETHERNET
== p
->family
) {
2098 sl
= skiplist_new(0, rfapi_bi_peer_rd_aux_cmp
, NULL
);
2100 sl
= skiplist_new(0, rfapi_bi_peer_rd_cmp
, NULL
);
2102 RFAPI_IT_EXTRA_GET(rn
)->u
.vpn
.idx_rd
= sl
;
2103 agg_lock_node(rn
); /* for skiplist */
2105 assert(!skiplist_insert(sl
, (void *)bpi
, (void *)bpi
));
2106 agg_lock_node(rn
); /* for skiplist entry */
2108 /* NB: BPIs in import tables are not refcounted */
2111 static void rfapiItBiIndexDump(struct agg_node
*rn
)
2113 struct skiplist
*sl
;
2114 void *cursor
= NULL
;
2115 struct bgp_path_info
*k
;
2116 struct bgp_path_info
*v
;
2119 sl
= RFAPI_RDINDEX(rn
);
2123 for (rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
); !rc
;
2124 rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
)) {
2126 char buf
[RD_ADDRSTRLEN
];
2127 char buf_aux_pfx
[PREFIX_STRLEN
];
2130 &k
->extra
->vnc
.import
.rd
, buf
, sizeof(buf
),
2131 bgp_get_asnotation(k
->peer
? k
->peer
->bgp
: NULL
));
2132 if (k
->extra
->vnc
.import
.aux_prefix
.family
) {
2133 prefix2str(&k
->extra
->vnc
.import
.aux_prefix
,
2134 buf_aux_pfx
, sizeof(buf_aux_pfx
));
2136 strlcpy(buf_aux_pfx
, "(none)", sizeof(buf_aux_pfx
));
2138 vnc_zlog_debug_verbose("bpi %p, peer %p, rd %s, aux_prefix %s",
2139 k
, k
->peer
, buf
, buf_aux_pfx
);
2143 static struct bgp_path_info
*rfapiItBiIndexSearch(
2144 struct agg_node
*rn
, /* Import table VPN node */
2145 struct prefix_rd
*prd
, struct peer
*peer
,
2146 const struct prefix
*aux_prefix
) /* optional L3 addr for L2 ITs */
2148 struct skiplist
*sl
;
2150 struct bgp_path_info bpi_fake
= {0};
2151 struct bgp_path_info_extra bpi_extra
= {0};
2152 struct bgp_path_info
*bpi_result
;
2154 sl
= RFAPI_RDINDEX(rn
);
2158 #ifdef DEBUG_BI_SEARCH
2160 char buf_aux_pfx
[PREFIX_STRLEN
];
2163 prefix2str(aux_prefix
, buf_aux_pfx
,
2164 sizeof(buf_aux_pfx
));
2166 strlcpy(buf_aux_pfx
, "(nil)", sizeof(buf_aux_pfx
));
2168 vnc_zlog_debug_verbose(
2169 "%s want prd=%pRDP, peer=%p, aux_prefix=%s", __func__
,
2170 prd
, peer
, buf_aux_pfx
);
2171 rfapiItBiIndexDump(rn
);
2175 /* threshold is a WAG */
2176 if (sl
->count
< 3) {
2177 #ifdef DEBUG_BI_SEARCH
2178 vnc_zlog_debug_verbose("%s: short list algorithm", __func__
);
2180 /* if short list, linear search might be faster */
2181 for (bpi_result
= rn
->info
; bpi_result
;
2182 bpi_result
= bpi_result
->next
) {
2183 #ifdef DEBUG_BI_SEARCH
2184 vnc_zlog_debug_verbose(
2185 "%s: bpi has prd=%pRDP, peer=%p", __func__
,
2186 &bpi_result
->extra
->vnc
.import
.rd
,
2189 if (peer
== bpi_result
->peer
2190 && !prefix_cmp((struct prefix
*)&bpi_result
->extra
2192 (struct prefix
*)prd
)) {
2194 #ifdef DEBUG_BI_SEARCH
2195 vnc_zlog_debug_verbose(
2196 "%s: peer and RD same, doing aux_prefix check",
2202 &bpi_result
->extra
->vnc
.import
2205 #ifdef DEBUG_BI_SEARCH
2206 vnc_zlog_debug_verbose("%s: match",
2216 bpi_fake
.peer
= peer
;
2217 bpi_fake
.extra
= &bpi_extra
;
2218 bpi_fake
.extra
->vnc
.import
.rd
= *prd
;
2220 bpi_fake
.extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
2223 bpi_fake
.extra
->vnc
.import
.aux_prefix
.family
= AF_ETHERNET
;
2224 bpi_fake
.extra
->vnc
.import
.aux_prefix
.prefixlen
= 1;
2227 rc
= skiplist_search(sl
, (void *)&bpi_fake
, (void *)&bpi_result
);
2230 #ifdef DEBUG_BI_SEARCH
2231 vnc_zlog_debug_verbose("%s: no match", __func__
);
2236 #ifdef DEBUG_BI_SEARCH
2237 vnc_zlog_debug_verbose("%s: matched bpi=%p", __func__
, bpi_result
);
2243 static void rfapiItBiIndexDel(struct agg_node
*rn
, /* Import table VPN node */
2244 struct bgp_path_info
*bpi
) /* old BPI */
2246 struct skiplist
*sl
;
2249 vnc_zlog_debug_verbose("%s: bpi %p, peer %p, rd %pRDP", __func__
, bpi
,
2250 bpi
->peer
, &bpi
->extra
->vnc
.import
.rd
);
2252 sl
= RFAPI_RDINDEX(rn
);
2255 rc
= skiplist_delete(sl
, (void *)(bpi
), (void *)bpi
);
2257 rfapiItBiIndexDump(rn
);
2261 agg_unlock_node(rn
); /* for skiplist entry */
2263 /* NB: BPIs in import tables are not refcounted */
2267 * Add a backreference at the ENCAP node to the VPN route that
2271 rfapiMonitorEncapAdd(struct rfapi_import_table
*import_table
,
2272 struct prefix
*p
, /* VN address */
2273 struct agg_node
*vpn_rn
, /* VPN node */
2274 struct bgp_path_info
*vpn_bpi
) /* VPN bpi/route */
2276 afi_t afi
= family2afi(p
->family
);
2277 struct agg_node
*rn
;
2278 struct rfapi_monitor_encap
*m
;
2281 rn
= agg_node_get(import_table
->imported_encap
[afi
], p
); /* locks rn */
2284 m
= XCALLOC(MTYPE_RFAPI_MONITOR_ENCAP
,
2285 sizeof(struct rfapi_monitor_encap
));
2291 /* insert to encap node's list */
2292 m
->next
= RFAPI_MONITOR_ENCAP(rn
);
2295 RFAPI_MONITOR_ENCAP_W_ALLOC(rn
) = m
;
2297 /* for easy lookup when deleting vpn route */
2298 vpn_bpi
->extra
->vnc
.import
.hme
= m
;
2300 vnc_zlog_debug_verbose(
2301 "%s: it=%p, vpn_bpi=%p, afi=%d, encap rn=%p, setting vpn_bpi->extra->vnc.import.hme=%p",
2302 __func__
, import_table
, vpn_bpi
, afi
, rn
, m
);
2304 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
2305 bgp_attr_intern(vpn_bpi
->attr
);
2308 static void rfapiMonitorEncapDelete(struct bgp_path_info
*vpn_bpi
)
2311 * Remove encap monitor
2313 vnc_zlog_debug_verbose("%s: vpn_bpi=%p", __func__
, vpn_bpi
);
2314 if (vpn_bpi
->extra
) {
2315 struct rfapi_monitor_encap
*hme
=
2316 vpn_bpi
->extra
->vnc
.import
.hme
;
2320 vnc_zlog_debug_verbose("%s: hme=%p", __func__
, hme
);
2322 /* Refcount checking takes too long here */
2323 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 0);
2325 hme
->next
->prev
= hme
->prev
;
2327 hme
->prev
->next
= hme
->next
;
2329 RFAPI_MONITOR_ENCAP_W_ALLOC(hme
->rn
) =
2331 /* Refcount checking takes too long here */
2332 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 1);
2334 /* see if the struct rfapi_it_extra is empty and can be
2336 rfapiMonitorExtraPrune(SAFI_ENCAP
, hme
->rn
);
2338 agg_unlock_node(hme
->rn
); /* decr ref count */
2339 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, hme
);
2340 vpn_bpi
->extra
->vnc
.import
.hme
= NULL
;
2346 * Timer callback for withdraw
2348 static void rfapiWithdrawTimerVPN(struct thread
*t
)
2350 struct rfapi_withdraw
*wcb
= THREAD_ARG(t
);
2351 struct bgp_path_info
*bpi
= wcb
->info
;
2352 struct bgp
*bgp
= bgp_get_default();
2353 const struct prefix
*p
;
2354 struct rfapi_monitor_vpn
*moved
;
2356 bool early_exit
= false;
2359 vnc_zlog_debug_verbose(
2360 "%s: NULL BGP pointer, assume shutdown race condition!!!",
2364 if (bgp
&& CHECK_FLAG(bgp
->flags
, BGP_FLAG_DELETE_IN_PROGRESS
)) {
2365 vnc_zlog_debug_verbose(
2366 "%s: BGP delete in progress, assume shutdown race condition!!!",
2371 /* This callback is responsible for the withdraw object's memory */
2373 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2379 assert(wcb
->import_table
);
2382 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, wcb
->lockoffset
);
2384 vnc_zlog_debug_verbose("%s: removing bpi %p at prefix %pRN", __func__
,
2388 * Remove the route (doubly-linked)
2390 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_VALID
)
2391 && VALID_INTERIOR_TYPE(bpi
->type
))
2392 RFAPI_MONITOR_EXTERIOR(wcb
->node
)->valid_interior_count
--;
2394 p
= agg_node_get_prefix(wcb
->node
);
2395 afi
= family2afi(p
->family
);
2396 wcb
->import_table
->holddown_count
[afi
] -= 1; /* keep count consistent */
2397 rfapiItBiIndexDel(wcb
->node
, bpi
);
2398 rfapiBgpInfoDetach(wcb
->node
, bpi
); /* with removed bpi */
2400 vnc_import_bgp_exterior_del_route_interior(bgp
, wcb
->import_table
,
2405 * If VNC is configured to send response remove messages, AND
2406 * if the removed route had a UN address, do response removal
2409 if (!(bgp
->rfapi_cfg
->flags
2410 & BGP_VNC_CONFIG_RESPONSE_REMOVAL_DISABLE
)) {
2412 int has_valid_duplicate
= 0;
2413 struct bgp_path_info
*bpii
;
2416 * First check if there are any OTHER routes at this node
2417 * that have the same nexthop and a valid UN address. If
2418 * there are (e.g., from other peers), then the route isn't
2419 * really gone, so skip sending a response removal message.
2421 for (bpii
= wcb
->node
->info
; bpii
; bpii
= bpii
->next
) {
2422 if (rfapiVpnBiSamePtUn(bpi
, bpii
)) {
2423 has_valid_duplicate
= 1;
2428 vnc_zlog_debug_verbose("%s: has_valid_duplicate=%d", __func__
,
2429 has_valid_duplicate
);
2431 if (!has_valid_duplicate
) {
2432 rfapiRibPendingDeleteRoute(bgp
, wcb
->import_table
, afi
,
2437 rfapiMonitorEncapDelete(bpi
);
2440 * If there are no VPN monitors at this VPN Node A,
2443 if (!RFAPI_MONITOR_VPN(wcb
->node
)) {
2444 vnc_zlog_debug_verbose("%s: no VPN monitors at this node",
2450 * rfapiMonitorMoveShorter only moves monitors if there are
2451 * no remaining valid routes at the current node
2453 moved
= rfapiMonitorMoveShorter(wcb
->node
, 1);
2456 rfapiMonitorMovedUp(wcb
->import_table
, wcb
->node
, moved
->node
,
2464 rfapiBgpInfoFree(bpi
);
2468 * If route count at this node has gone to 0, withdraw exported prefix
2470 if (!wcb
->node
->info
) {
2471 /* see if the struct rfapi_it_extra is empty and can be freed */
2472 rfapiMonitorExtraPrune(SAFI_MPLS_VPN
, wcb
->node
);
2473 vnc_direct_bgp_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2474 vnc_zebra_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2477 * nexthop change event
2478 * vnc_direct_bgp_add_prefix() will recompute the VN addr
2481 vnc_direct_bgp_add_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2484 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, 1 + wcb
->lockoffset
);
2485 agg_unlock_node(wcb
->node
); /* decr ref count */
2486 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2490 * This works for multiprotocol extension, but not for plain ol'
2491 * unicast IPv4 because that nexthop is stored in attr->nexthop
2493 void rfapiNexthop2Prefix(struct attr
*attr
, struct prefix
*p
)
2498 memset(p
, 0, sizeof(struct prefix
));
2500 switch (p
->family
= BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2502 p
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2503 p
->prefixlen
= IPV4_MAX_BITLEN
;
2507 p
->u
.prefix6
= attr
->mp_nexthop_global
;
2508 p
->prefixlen
= IPV6_MAX_BITLEN
;
2512 vnc_zlog_debug_verbose("%s: Family is unknown = %d", __func__
,
2517 void rfapiUnicastNexthop2Prefix(afi_t afi
, struct attr
*attr
, struct prefix
*p
)
2519 if (afi
== AFI_IP
) {
2520 p
->family
= AF_INET
;
2521 p
->prefixlen
= IPV4_MAX_BITLEN
;
2522 p
->u
.prefix4
= attr
->nexthop
;
2524 rfapiNexthop2Prefix(attr
, p
);
2528 static int rfapiAttrNexthopAddrDifferent(struct prefix
*p1
, struct prefix
*p2
)
2531 vnc_zlog_debug_verbose("%s: p1 or p2 is NULL", __func__
);
2536 * Are address families the same?
2538 if (p1
->family
!= p2
->family
) {
2542 switch (p1
->family
) {
2544 if (IPV4_ADDR_SAME(&p1
->u
.prefix4
, &p2
->u
.prefix4
))
2549 if (IPV6_ADDR_SAME(&p1
->u
.prefix6
, &p2
->u
.prefix6
))
2560 static void rfapiCopyUnEncap2VPN(struct bgp_path_info
*encap_bpi
,
2561 struct bgp_path_info
*vpn_bpi
)
2563 if (!vpn_bpi
|| !vpn_bpi
->extra
) {
2564 zlog_warn("%s: no vpn bpi attr/extra, can't copy UN address",
2569 switch (BGP_MP_NEXTHOP_FAMILY(encap_bpi
->attr
->mp_nexthop_len
)) {
2573 * instrumentation to debug segfault of 091127
2575 vnc_zlog_debug_verbose("%s: vpn_bpi=%p", __func__
, vpn_bpi
);
2576 vnc_zlog_debug_verbose("%s: vpn_bpi->extra=%p", __func__
,
2579 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_INET
;
2580 vpn_bpi
->extra
->vnc
.import
.un
.addr4
=
2581 encap_bpi
->attr
->mp_nexthop_global_in
;
2585 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_INET6
;
2586 vpn_bpi
->extra
->vnc
.import
.un
.addr6
=
2587 encap_bpi
->attr
->mp_nexthop_global
;
2591 zlog_warn("%s: invalid encap nexthop length: %d", __func__
,
2592 encap_bpi
->attr
->mp_nexthop_len
);
2593 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_UNSPEC
;
2599 * returns 0 on success, nonzero on error
2602 rfapiWithdrawEncapUpdateCachedUn(struct rfapi_import_table
*import_table
,
2603 struct bgp_path_info
*encap_bpi
,
2604 struct agg_node
*vpn_rn
,
2605 struct bgp_path_info
*vpn_bpi
)
2610 * clear cached UN address
2612 if (!vpn_bpi
|| !vpn_bpi
->extra
) {
2614 "%s: missing VPN bpi/extra, can't clear UN addr",
2618 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_UNSPEC
;
2619 memset(&vpn_bpi
->extra
->vnc
.import
.un
, 0,
2620 sizeof(vpn_bpi
->extra
->vnc
.import
.un
));
2621 if (CHECK_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
)) {
2622 if (rfapiGetVncTunnelUnAddr(vpn_bpi
->attr
, NULL
)) {
2623 UNSET_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
);
2624 if (VALID_INTERIOR_TYPE(vpn_bpi
->type
))
2625 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2626 ->valid_interior_count
--;
2627 /* signal interior route withdrawal to
2628 * import-exterior */
2629 vnc_import_bgp_exterior_del_route_interior(
2630 bgp_get_default(), import_table
, vpn_rn
,
2637 zlog_warn("%s: missing VPN bpi, can't clear UN addr",
2641 rfapiCopyUnEncap2VPN(encap_bpi
, vpn_bpi
);
2642 if (!CHECK_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
)) {
2643 SET_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
);
2644 if (VALID_INTERIOR_TYPE(vpn_bpi
->type
))
2645 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2646 ->valid_interior_count
++;
2647 /* signal interior route withdrawal to import-exterior
2649 vnc_import_bgp_exterior_add_route_interior(
2650 bgp_get_default(), import_table
, vpn_rn
,
2657 static void rfapiWithdrawTimerEncap(struct thread
*t
)
2659 struct rfapi_withdraw
*wcb
= THREAD_ARG(t
);
2660 struct bgp_path_info
*bpi
= wcb
->info
;
2661 int was_first_route
= 0;
2662 struct rfapi_monitor_encap
*em
;
2663 struct skiplist
*vpn_node_sl
= skiplist_new(0, NULL
, NULL
);
2667 assert(wcb
->import_table
);
2669 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 0);
2671 if (wcb
->node
->info
== bpi
)
2672 was_first_route
= 1;
2675 * Remove the route/bpi and free it
2677 rfapiBgpInfoDetach(wcb
->node
, bpi
);
2678 rfapiBgpInfoFree(bpi
);
2680 if (!was_first_route
)
2683 for (em
= RFAPI_MONITOR_ENCAP(wcb
->node
); em
; em
= em
->next
) {
2686 * Update monitoring VPN BPIs with new encap info at the
2687 * head of the encap bpi chain (which could be NULL after
2688 * removing the expiring bpi above)
2690 if (rfapiWithdrawEncapUpdateCachedUn(wcb
->import_table
,
2691 wcb
->node
->info
, em
->node
,
2696 * Build a list of unique VPN nodes referenced by these
2698 * Use a skiplist for speed.
2700 skiplist_insert(vpn_node_sl
, em
->node
, em
->node
);
2705 * for each VPN node referenced in the ENCAP monitors:
2707 struct agg_node
*rn
;
2708 while (!skiplist_first(vpn_node_sl
, (void **)&rn
, NULL
)) {
2709 if (!wcb
->node
->info
) {
2710 struct rfapi_monitor_vpn
*moved
;
2712 moved
= rfapiMonitorMoveShorter(rn
, 0);
2714 // rfapiDoRouteCallback(wcb->import_table,
2715 // moved->node, moved);
2716 rfapiMonitorMovedUp(wcb
->import_table
, rn
,
2717 moved
->node
, moved
);
2720 // rfapiDoRouteCallback(wcb->import_table, rn, NULL);
2721 rfapiMonitorItNodeChanged(wcb
->import_table
, rn
, NULL
);
2723 skiplist_delete_first(vpn_node_sl
);
2727 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 1);
2728 agg_unlock_node(wcb
->node
); /* decr ref count */
2729 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2730 skiplist_free(vpn_node_sl
);
2735 * Works for both VPN and ENCAP routes; timer_service_func is different
2739 rfapiBiStartWithdrawTimer(struct rfapi_import_table
*import_table
,
2740 struct agg_node
*rn
, struct bgp_path_info
*bpi
,
2741 afi_t afi
, safi_t safi
,
2742 void (*timer_service_func
)(struct thread
*))
2745 struct rfapi_withdraw
*wcb
;
2747 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
2749 * Already on the path to being withdrawn,
2750 * should already have a timer set up to
2753 vnc_zlog_debug_verbose(
2754 "%s: already being withdrawn, do nothing", __func__
);
2758 rfapiGetVncLifetime(bpi
->attr
, &lifetime
);
2759 vnc_zlog_debug_verbose("%s: VNC lifetime is %u", __func__
, lifetime
);
2762 * withdrawn routes get to hang around for a while
2764 SET_FLAG(bpi
->flags
, BGP_PATH_REMOVED
);
2766 /* set timer to remove the route later */
2767 lifetime
= rfapiGetHolddownFromLifetime(lifetime
);
2768 vnc_zlog_debug_verbose("%s: using timeout %u", __func__
, lifetime
);
2771 * Stash import_table, node, and info for use by timer
2772 * service routine, which is supposed to free the wcb.
2774 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2777 wcb
->import_table
= import_table
;
2778 bgp_attr_intern(bpi
->attr
);
2780 if (VNC_DEBUG(VERBOSE
)) {
2781 vnc_zlog_debug_verbose(
2782 "%s: wcb values: node=%p, info=%p, import_table=%p (bpi follows)",
2783 __func__
, wcb
->node
, wcb
->info
, wcb
->import_table
);
2784 rfapiPrintBi(NULL
, bpi
);
2789 if (lifetime
> UINT32_MAX
/ 1001) {
2790 /* sub-optimal case, but will probably never happen */
2791 bpi
->extra
->vnc
.import
.timer
= NULL
;
2792 thread_add_timer(bm
->master
, timer_service_func
, wcb
, lifetime
,
2793 &bpi
->extra
->vnc
.import
.timer
);
2795 static uint32_t jitter
;
2796 uint32_t lifetime_msec
;
2799 * the goal here is to spread out the timers so they are
2800 * sortable in the skip list
2802 if (++jitter
>= 1000)
2805 lifetime_msec
= (lifetime
* 1000) + jitter
;
2807 bpi
->extra
->vnc
.import
.timer
= NULL
;
2808 thread_add_timer_msec(bm
->master
, timer_service_func
, wcb
,
2810 &bpi
->extra
->vnc
.import
.timer
);
2813 /* re-sort route list (BGP_PATH_REMOVED routes are last) */
2814 if (((struct bgp_path_info
*)rn
->info
)->next
) {
2815 rfapiBgpInfoDetach(rn
, bpi
);
2816 rfapiBgpInfoAttachSorted(rn
, bpi
, afi
, safi
);
2821 typedef void(rfapi_bi_filtered_import_f
)(struct rfapi_import_table
*table
,
2822 int action
, struct peer
*peer
,
2823 void *rfd
, const struct prefix
*prefix
,
2824 const struct prefix
*aux_prefix
,
2825 afi_t afi
, struct prefix_rd
*prd
,
2826 struct attr
*attr
, uint8_t type
,
2827 uint8_t sub_type
, uint32_t *label
);
2830 static void rfapiExpireEncapNow(struct rfapi_import_table
*it
,
2831 struct agg_node
*rn
, struct bgp_path_info
*bpi
)
2833 struct rfapi_withdraw
*wcb
;
2837 * pretend we're an expiring timer
2839 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2842 wcb
->import_table
= it
;
2843 memset(&t
, 0, sizeof(t
));
2845 rfapiWithdrawTimerEncap(&t
); /* frees wcb */
2848 static int rfapiGetNexthop(struct attr
*attr
, struct prefix
*prefix
)
2850 switch (BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2852 prefix
->family
= AF_INET
;
2853 prefix
->prefixlen
= IPV4_MAX_BITLEN
;
2854 prefix
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2857 prefix
->family
= AF_INET6
;
2858 prefix
->prefixlen
= IPV6_MAX_BITLEN
;
2859 prefix
->u
.prefix6
= attr
->mp_nexthop_global
;
2862 vnc_zlog_debug_verbose("%s: unknown attr->mp_nexthop_len %d",
2863 __func__
, attr
->mp_nexthop_len
);
2870 * import a bgp_path_info if its route target list intersects with the
2871 * import table's route target list
2873 static void rfapiBgpInfoFilteredImportEncap(
2874 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
2875 void *rfd
, /* set for looped back routes */
2876 const struct prefix
*p
,
2877 const struct prefix
*aux_prefix
, /* Unused for encap routes */
2878 afi_t afi
, struct prefix_rd
*prd
,
2879 struct attr
*attr
, /* part of bgp_path_info */
2880 uint8_t type
, /* part of bgp_path_info */
2881 uint8_t sub_type
, /* part of bgp_path_info */
2882 uint32_t *label
) /* part of bgp_path_info */
2884 struct agg_table
*rt
= NULL
;
2885 struct agg_node
*rn
;
2886 struct bgp_path_info
*info_new
;
2887 struct bgp_path_info
*bpi
;
2888 struct bgp_path_info
*next
;
2891 struct prefix p_firstbpi_old
;
2892 struct prefix p_firstbpi_new
;
2894 const char *action_str
= NULL
;
2895 struct prefix un_prefix
;
2898 bgp
= bgp_get_default(); /* assume 1 instance for now */
2901 case FIF_ACTION_UPDATE
:
2902 action_str
= "update";
2904 case FIF_ACTION_WITHDRAW
:
2905 action_str
= "withdraw";
2907 case FIF_ACTION_KILL
:
2908 action_str
= "kill";
2915 vnc_zlog_debug_verbose(
2916 "%s: entry: %s: prefix %s/%d", __func__
, action_str
,
2917 inet_ntop(p
->family
, &p
->u
.prefix
, buf
, sizeof(buf
)),
2920 memset(&p_firstbpi_old
, 0, sizeof(p_firstbpi_old
));
2921 memset(&p_firstbpi_new
, 0, sizeof(p_firstbpi_new
));
2923 if (action
== FIF_ACTION_UPDATE
) {
2925 * Compare rt lists. If no intersection, don't import this route
2926 * On a withdraw, peer and RD are sufficient to determine if
2929 if (!attr
|| !bgp_attr_get_ecommunity(attr
)) {
2931 vnc_zlog_debug_verbose(
2932 "%s: attr, extra, or ecommunity missing, not importing",
2936 #ifdef RFAPI_REQUIRE_ENCAP_BEEC
2937 if (!rfapiEcommunitiesMatchBeec(
2938 bgp_attr_get_ecommunity(attr
))) {
2939 vnc_zlog_debug_verbose(
2940 "%s: it=%p: no match for BGP Encapsulation ecommunity",
2941 __func__
, import_table
);
2945 if (!rfapiEcommunitiesIntersect(
2946 import_table
->rt_import_list
,
2947 bgp_attr_get_ecommunity(attr
))) {
2949 vnc_zlog_debug_verbose(
2950 "%s: it=%p: no ecommunity intersection",
2951 __func__
, import_table
);
2956 * Updates must also have a nexthop address
2958 memset(&un_prefix
, 0,
2959 sizeof(un_prefix
)); /* keep valgrind happy */
2960 if (rfapiGetNexthop(attr
, &un_prefix
)) {
2961 vnc_zlog_debug_verbose("%s: missing nexthop address",
2968 * Figure out which radix tree the route would go into
2973 rt
= import_table
->imported_encap
[afi
];
2979 flog_err(EC_LIB_DEVELOPMENT
, "%s: bad afi %d", __func__
, afi
);
2984 * agg_node_lookup returns a node only if there is at least
2985 * one route attached.
2987 rn
= agg_node_lookup(rt
, p
);
2989 #ifdef DEBUG_ENCAP_MONITOR
2990 vnc_zlog_debug_verbose("%s: initial encap lookup(it=%p) rn=%p",
2991 __func__
, import_table
, rn
);
2996 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 1);
2997 agg_unlock_node(rn
); /* undo lock in agg_node_lookup */
3001 * capture nexthop of first bpi
3004 rfapiNexthop2Prefix(
3005 ((struct bgp_path_info
*)(rn
->info
))->attr
,
3009 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
3012 * Does this bgp_path_info refer to the same route
3013 * as we are trying to add?
3015 vnc_zlog_debug_verbose("%s: comparing BPI %p", __func__
,
3022 * RD of import table bpi is in
3023 * bpi->extra->vnc.import.rd RD of info_orig is in prd
3026 vnc_zlog_debug_verbose("%s: no bpi->extra",
3031 (struct prefix
*)&bpi
->extra
->vnc
.import
.rd
,
3032 (struct prefix
*)prd
)) {
3034 vnc_zlog_debug_verbose("%s: prd does not match",
3042 if (bpi
->peer
!= peer
) {
3043 vnc_zlog_debug_verbose(
3044 "%s: peer does not match", __func__
);
3048 vnc_zlog_debug_verbose("%s: found matching bpi",
3051 /* Same route. Delete this bpi, replace with new one */
3053 if (action
== FIF_ACTION_WITHDRAW
) {
3055 vnc_zlog_debug_verbose(
3056 "%s: withdrawing at prefix %pRN",
3059 rfapiBiStartWithdrawTimer(
3060 import_table
, rn
, bpi
, afi
, SAFI_ENCAP
,
3061 rfapiWithdrawTimerEncap
);
3064 vnc_zlog_debug_verbose(
3065 "%s: %s at prefix %pRN", __func__
,
3066 ((action
== FIF_ACTION_KILL
)
3072 * If this route is waiting to be deleted
3074 * a previous withdraw, we must cancel its
3077 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
3078 && bpi
->extra
->vnc
.import
.timer
) {
3079 struct rfapi_withdraw
*wcb
= THREAD_ARG(
3080 bpi
->extra
->vnc
.import
.timer
);
3082 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3084 bpi
->extra
->vnc
.import
.timer
);
3087 if (action
== FIF_ACTION_UPDATE
) {
3088 rfapiBgpInfoDetach(rn
, bpi
);
3089 rfapiBgpInfoFree(bpi
);
3093 * Kill: do export stuff when removing
3096 struct rfapi_withdraw
*wcb
;
3100 * pretend we're an expiring timer
3103 MTYPE_RFAPI_WITHDRAW
,
3104 sizeof(struct rfapi_withdraw
));
3107 wcb
->import_table
= import_table
;
3108 memset(&t
, 0, sizeof(t
));
3110 rfapiWithdrawTimerEncap(
3111 &t
); /* frees wcb */
3120 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, replacing
? 1 : 0);
3122 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
)
3126 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, NULL
);
3130 agg_lock_node(rn
); /* incr ref count for new BPI */
3132 rn
= agg_node_get(rt
, p
);
3135 vnc_zlog_debug_verbose("%s: (afi=%d, rn=%p) inserting at prefix %pRN",
3136 __func__
, afi
, rn
, rn
);
3138 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_ENCAP
);
3141 * Delete holddown routes from same NVE. See details in
3142 * rfapiBgpInfoFilteredImportVPN()
3144 for (bpi
= info_new
->next
; bpi
; bpi
= next
) {
3146 struct prefix pfx_un
;
3150 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
))
3154 * We already match the VN address (it is the prefix
3155 * of the route node)
3158 if (!rfapiGetNexthop(bpi
->attr
, &pfx_un
)
3159 && prefix_same(&pfx_un
, &un_prefix
)) {
3167 vnc_zlog_debug_verbose(
3168 "%s: removing holddown bpi matching NVE of new route",
3170 if (bpi
->extra
->vnc
.import
.timer
) {
3171 struct rfapi_withdraw
*wcb
=
3172 THREAD_ARG(bpi
->extra
->vnc
.import
.timer
);
3174 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3175 THREAD_OFF(bpi
->extra
->vnc
.import
.timer
);
3177 rfapiExpireEncapNow(import_table
, rn
, bpi
);
3180 rfapiNexthop2Prefix(((struct bgp_path_info
*)(rn
->info
))->attr
,
3184 * If the nexthop address of the selected Encap route (i.e.,
3185 * the UN address) has changed, then we must update the VPN
3186 * routes that refer to this Encap route and possibly force
3189 if (rfapiAttrNexthopAddrDifferent(&p_firstbpi_old
, &p_firstbpi_new
)) {
3191 struct rfapi_monitor_encap
*m
;
3192 struct rfapi_monitor_encap
*mnext
;
3194 struct agg_node
*referenced_vpn_prefix
;
3197 * Optimized approach: build radix tree on the fly to
3198 * hold list of VPN nodes referenced by the ENCAP monitors
3200 * The nodes in this table correspond to prefixes of VPN routes.
3201 * The "info" pointer of the node points to a chain of
3202 * struct rfapi_monitor_encap, each of which refers to a
3203 * specific VPN node.
3205 struct agg_table
*referenced_vpn_table
;
3207 referenced_vpn_table
= agg_table_init();
3210 * iterate over the set of monitors at this ENCAP node.
3212 #ifdef DEBUG_ENCAP_MONITOR
3213 vnc_zlog_debug_verbose("%s: examining monitors at rn=%p",
3216 for (m
= RFAPI_MONITOR_ENCAP(rn
); m
; m
= m
->next
) {
3217 const struct prefix
*p
;
3220 * For each referenced bpi/route, copy the ENCAP route's
3221 * nexthop to the VPN route's cached UN address field
3223 * the address family of the cached UN address field.
3225 rfapiCopyUnEncap2VPN(info_new
, m
->bpi
);
3226 if (!CHECK_FLAG(m
->bpi
->flags
, BGP_PATH_VALID
)) {
3227 SET_FLAG(m
->bpi
->flags
, BGP_PATH_VALID
);
3228 if (VALID_INTERIOR_TYPE(m
->bpi
->type
))
3229 RFAPI_MONITOR_EXTERIOR(m
->node
)
3230 ->valid_interior_count
++;
3231 vnc_import_bgp_exterior_add_route_interior(
3232 bgp
, import_table
, m
->node
, m
->bpi
);
3236 * Build a list of unique VPN nodes referenced by these
3239 * There could be more than one VPN node here with a
3241 * prefix. Those are currently in an unsorted linear
3245 p
= agg_node_get_prefix(m
->node
);
3246 referenced_vpn_prefix
=
3247 agg_node_get(referenced_vpn_table
, p
);
3248 assert(referenced_vpn_prefix
);
3249 for (mnext
= referenced_vpn_prefix
->info
; mnext
;
3250 mnext
= mnext
->next
) {
3252 if (mnext
->node
== m
->node
)
3258 * already have an entry for this VPN node
3260 agg_unlock_node(referenced_vpn_prefix
);
3263 MTYPE_RFAPI_MONITOR_ENCAP
,
3264 sizeof(struct rfapi_monitor_encap
));
3265 mnext
->node
= m
->node
;
3266 mnext
->next
= referenced_vpn_prefix
->info
;
3267 referenced_vpn_prefix
->info
= mnext
;
3272 * for each VPN node referenced in the ENCAP monitors:
3274 for (referenced_vpn_prefix
=
3275 agg_route_top(referenced_vpn_table
);
3276 referenced_vpn_prefix
;
3277 referenced_vpn_prefix
=
3278 agg_route_next(referenced_vpn_prefix
)) {
3280 while ((m
= referenced_vpn_prefix
->info
)) {
3284 rfapiMonitorMoveLonger(m
->node
);
3285 for (n
= m
->node
; n
; n
= agg_node_parent(n
)) {
3286 // rfapiDoRouteCallback(import_table, n,
3289 rfapiMonitorItNodeChanged(import_table
, m
->node
,
3292 referenced_vpn_prefix
->info
= m
->next
;
3293 agg_unlock_node(referenced_vpn_prefix
);
3294 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, m
);
3297 agg_table_finish(referenced_vpn_table
);
3300 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
3303 static void rfapiExpireVpnNow(struct rfapi_import_table
*it
,
3304 struct agg_node
*rn
, struct bgp_path_info
*bpi
,
3307 struct rfapi_withdraw
*wcb
;
3311 * pretend we're an expiring timer
3313 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
3316 wcb
->import_table
= it
;
3317 wcb
->lockoffset
= lockoffset
;
3318 memset(&t
, 0, sizeof(t
));
3320 rfapiWithdrawTimerVPN(&t
); /* frees wcb */
3325 * import a bgp_path_info if its route target list intersects with the
3326 * import table's route target list
3328 void rfapiBgpInfoFilteredImportVPN(
3329 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3330 void *rfd
, /* set for looped back routes */
3331 const struct prefix
*p
,
3332 const struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3333 afi_t afi
, struct prefix_rd
*prd
,
3334 struct attr
*attr
, /* part of bgp_path_info */
3335 uint8_t type
, /* part of bgp_path_info */
3336 uint8_t sub_type
, /* part of bgp_path_info */
3337 uint32_t *label
) /* part of bgp_path_info */
3339 struct agg_table
*rt
= NULL
;
3340 struct agg_node
*rn
;
3342 struct bgp_path_info
*info_new
;
3343 struct bgp_path_info
*bpi
;
3344 struct bgp_path_info
*next
;
3346 struct prefix vn_prefix
;
3347 struct prefix un_prefix
;
3348 int un_prefix_valid
= 0;
3349 struct agg_node
*ern
;
3351 int original_had_routes
= 0;
3352 struct prefix original_nexthop
;
3353 const char *action_str
= NULL
;
3357 bgp
= bgp_get_default(); /* assume 1 instance for now */
3360 case FIF_ACTION_UPDATE
:
3361 action_str
= "update";
3363 case FIF_ACTION_WITHDRAW
:
3364 action_str
= "withdraw";
3366 case FIF_ACTION_KILL
:
3367 action_str
= "kill";
3374 if (import_table
== bgp
->rfapi
->it_ce
)
3377 vnc_zlog_debug_verbose("%s: entry: %s%s: prefix %s/%d: it %p, afi %s",
3378 __func__
, (is_it_ce
? "CE-IT " : ""), action_str
,
3379 rfapi_ntop(p
->family
, &p
->u
.prefix
, buf
, BUFSIZ
),
3380 p
->prefixlen
, import_table
, afi2str(afi
));
3385 * Compare rt lists. If no intersection, don't import this route
3386 * On a withdraw, peer and RD are sufficient to determine if
3389 if (action
== FIF_ACTION_UPDATE
) {
3390 if (!attr
|| !bgp_attr_get_ecommunity(attr
)) {
3392 vnc_zlog_debug_verbose(
3393 "%s: attr, extra, or ecommunity missing, not importing",
3397 if ((import_table
!= bgp
->rfapi
->it_ce
) &&
3398 !rfapiEcommunitiesIntersect(
3399 import_table
->rt_import_list
,
3400 bgp_attr_get_ecommunity(attr
))) {
3402 vnc_zlog_debug_verbose(
3403 "%s: it=%p: no ecommunity intersection",
3404 __func__
, import_table
);
3408 memset(&vn_prefix
, 0,
3409 sizeof(vn_prefix
)); /* keep valgrind happy */
3410 if (rfapiGetNexthop(attr
, &vn_prefix
)) {
3411 /* missing nexthop address would be a bad, bad thing */
3412 vnc_zlog_debug_verbose("%s: missing nexthop", __func__
);
3418 * Figure out which radix tree the route would go into
3424 rt
= import_table
->imported_vpn
[afi
];
3429 flog_err(EC_LIB_DEVELOPMENT
, "%s: bad afi %d", __func__
, afi
);
3434 memset(&original_nexthop
, 0, sizeof(original_nexthop
));
3437 * agg_node_lookup returns a node only if there is at least
3438 * one route attached.
3440 rn
= agg_node_lookup(rt
, p
);
3442 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
3446 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
3447 agg_unlock_node(rn
); /* undo lock in agg_node_lookup */
3450 original_had_routes
= 1;
3452 if (VNC_DEBUG(VERBOSE
)) {
3453 vnc_zlog_debug_verbose("%s: showing IT node on entry",
3455 rfapiShowItNode(NULL
, rn
); /* debug */
3459 * Look for same route (will have same RD and peer)
3461 bpi
= rfapiItBiIndexSearch(rn
, prd
, peer
, aux_prefix
);
3466 * This was an old test when we iterated over the
3467 * BPIs linearly. Since we're now looking up with
3468 * RD and peer, comparing types should not be
3469 * needed. Changed to assertion.
3471 * Compare types. Doing so prevents a RFP-originated
3472 * route from matching an imported route, for example.
3474 if (VNC_DEBUG(VERBOSE
) && bpi
->type
!= type
)
3475 /* should be handled by RDs, but warn for now */
3476 zlog_warn("%s: type mismatch! (bpi=%d, arg=%d)",
3477 __func__
, bpi
->type
, type
);
3479 vnc_zlog_debug_verbose("%s: found matching bpi",
3483 * In the special CE table, withdrawals occur without
3486 if (import_table
== bgp
->rfapi
->it_ce
) {
3487 vnc_direct_bgp_del_route_ce(bgp
, rn
, bpi
);
3488 if (action
== FIF_ACTION_WITHDRAW
)
3489 action
= FIF_ACTION_KILL
;
3492 if (action
== FIF_ACTION_WITHDRAW
) {
3494 int washolddown
= CHECK_FLAG(bpi
->flags
,
3497 vnc_zlog_debug_verbose(
3498 "%s: withdrawing at prefix %pRN%s",
3501 ? " (already being withdrawn)"
3506 rfapiBiStartWithdrawTimer(
3507 import_table
, rn
, bpi
, afi
,
3509 rfapiWithdrawTimerVPN
);
3511 RFAPI_UPDATE_ITABLE_COUNT(
3512 bpi
, import_table
, afi
, -1);
3513 import_table
->holddown_count
[afi
] += 1;
3517 vnc_zlog_debug_verbose(
3518 "%s: %s at prefix %pRN", __func__
,
3519 ((action
== FIF_ACTION_KILL
)
3525 * If this route is waiting to be deleted
3527 * a previous withdraw, we must cancel its
3530 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
3531 && bpi
->extra
->vnc
.import
.timer
) {
3532 struct rfapi_withdraw
*wcb
= THREAD_ARG(
3533 bpi
->extra
->vnc
.import
.timer
);
3535 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3537 bpi
->extra
->vnc
.import
.timer
);
3539 import_table
->holddown_count
[afi
] -= 1;
3540 RFAPI_UPDATE_ITABLE_COUNT(
3541 bpi
, import_table
, afi
, 1);
3544 * decrement remote count (if route is remote)
3546 * we are going to remove it below
3548 RFAPI_UPDATE_ITABLE_COUNT(bpi
, import_table
,
3550 if (action
== FIF_ACTION_UPDATE
) {
3554 * make copy of original nexthop so we
3555 * can see if it changed
3557 rfapiGetNexthop(bpi
->attr
,
3561 * remove bpi without doing any export
3564 if (CHECK_FLAG(bpi
->flags
,
3566 && VALID_INTERIOR_TYPE(bpi
->type
))
3567 RFAPI_MONITOR_EXTERIOR(rn
)
3568 ->valid_interior_count
--;
3569 rfapiItBiIndexDel(rn
, bpi
);
3570 rfapiBgpInfoDetach(rn
, bpi
);
3571 rfapiMonitorEncapDelete(bpi
);
3572 vnc_import_bgp_exterior_del_route_interior(
3573 bgp
, import_table
, rn
, bpi
);
3574 rfapiBgpInfoFree(bpi
);
3578 * remove bpi and do export processing
3580 import_table
->holddown_count
[afi
] += 1;
3581 rfapiExpireVpnNow(import_table
, rn
, bpi
,
3589 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, replacing
? 1 : 0);
3591 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
) {
3597 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, label
);
3600 * lookup un address in encap table
3602 ern
= agg_node_match(import_table
->imported_encap
[afi
], &vn_prefix
);
3604 rfapiCopyUnEncap2VPN(ern
->info
, info_new
);
3605 agg_unlock_node(ern
); /* undo lock in route_note_match */
3607 /* Not a big deal, just means VPN route got here first */
3608 vnc_zlog_debug_verbose("%s: no encap route for vn addr %pFX",
3609 __func__
, &vn_prefix
);
3610 info_new
->extra
->vnc
.import
.un_family
= AF_UNSPEC
;
3618 * No need to increment reference count, so only "get"
3619 * if the node is not there already
3621 rn
= agg_node_get(rt
, p
);
3625 * For ethernet routes, if there is an accompanying IP address,
3626 * save it in the bpi
3628 if ((AFI_L2VPN
== afi
) && aux_prefix
) {
3630 vnc_zlog_debug_verbose("%s: setting BPI's aux_prefix",
3632 info_new
->extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
3635 vnc_zlog_debug_verbose("%s: inserting bpi %p at prefix %pRN #%d",
3636 __func__
, info_new
, rn
,
3637 agg_node_get_lock_count(rn
));
3639 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_MPLS_VPN
);
3640 rfapiItBiIndexAdd(rn
, info_new
);
3641 if (!rfapiGetUnAddrOfVpnBi(info_new
, NULL
)) {
3642 if (VALID_INTERIOR_TYPE(info_new
->type
))
3643 RFAPI_MONITOR_EXTERIOR(rn
)->valid_interior_count
++;
3644 SET_FLAG(info_new
->flags
, BGP_PATH_VALID
);
3646 RFAPI_UPDATE_ITABLE_COUNT(info_new
, import_table
, afi
, 1);
3647 vnc_import_bgp_exterior_add_route_interior(bgp
, import_table
, rn
,
3650 if (import_table
== bgp
->rfapi
->it_ce
)
3651 vnc_direct_bgp_add_route_ce(bgp
, rn
, info_new
);
3653 if (VNC_DEBUG(VERBOSE
)) {
3654 vnc_zlog_debug_verbose("%s: showing IT node", __func__
);
3655 rfapiShowItNode(NULL
, rn
); /* debug */
3658 rfapiMonitorEncapAdd(import_table
, &vn_prefix
, rn
, info_new
);
3660 if (!rfapiGetUnAddrOfVpnBi(info_new
, &un_prefix
)) {
3663 * if we have a valid UN address (either via Encap route
3664 * or via tunnel attribute), then we should attempt
3665 * to move any monitors at less-specific nodes to this node
3667 rfapiMonitorMoveLonger(rn
);
3669 un_prefix_valid
= 1;
3673 * 101129 Enhancement: if we add a route (implication: it is not
3674 * in holddown), delete all other routes from this nve at this
3675 * node that are in holddown, regardless of peer.
3677 * Reasons it's OK to do that:
3679 * - if the holddown route being deleted originally came from BGP VPN,
3680 * it is already gone from BGP (implication of holddown), so there
3681 * won't be any added inconsistency with the BGP RIB.
3683 * - once a fresh route is added at a prefix, any routes in holddown
3684 * at that prefix will not show up in RFP responses, so deleting
3685 * the holddown routes won't affect the contents of responses.
3687 * - lifetimes are supposed to be consistent, so there should not
3688 * be a case where the fresh route has a shorter lifetime than
3689 * the holddown route, so we don't expect the fresh route to
3690 * disappear and complete its holddown time before the existing
3691 * holddown routes time out. Therefore, we won't have a situation
3692 * where we expect the existing holddown routes to be hidden and
3693 * then to reappear sometime later (as holddown routes) in a
3696 * Among other things, this would enable us to skirt the problem
3697 * of local holddown routes that refer to NVE descriptors that
3698 * have already been closed (if the same NVE triggers a subsequent
3699 * rfapi_open(), the new peer is different and doesn't match the
3700 * peer of the holddown route, so the stale holddown route still
3701 * hangs around until it times out instead of just being replaced
3702 * by the fresh route).
3705 * We know that the new bpi will have been inserted before any routes
3706 * in holddown, so we can skip any that came before it
3708 for (bpi
= info_new
->next
; bpi
; bpi
= next
) {
3710 struct prefix pfx_vn
;
3711 struct prefix pfx_un
;
3713 int remote_peer_match
= 0;
3720 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
))
3724 * Must match VN address (nexthop of VPN route)
3726 if (rfapiGetNexthop(bpi
->attr
, &pfx_vn
))
3728 if (!prefix_same(&pfx_vn
, &vn_prefix
))
3731 if (un_prefix_valid
&& /* new route UN addr */
3732 !rfapiGetUnAddrOfVpnBi(bpi
, &pfx_un
)
3733 && /* old route UN addr */
3734 prefix_same(&pfx_un
, &un_prefix
)) { /* compare */
3737 if (!RFAPI_LOCAL_BI(bpi
) && !RFAPI_LOCAL_BI(info_new
)
3738 && sockunion_same(&bpi
->peer
->su
, &info_new
->peer
->su
)) {
3739 /* old & new are both remote, same peer */
3740 remote_peer_match
= 1;
3743 if (!un_match
&& !remote_peer_match
)
3746 vnc_zlog_debug_verbose(
3747 "%s: removing holddown bpi matching NVE of new route",
3749 if (bpi
->extra
->vnc
.import
.timer
) {
3750 struct rfapi_withdraw
*wcb
=
3751 THREAD_ARG(bpi
->extra
->vnc
.import
.timer
);
3753 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3754 THREAD_OFF(bpi
->extra
->vnc
.import
.timer
);
3756 rfapiExpireVpnNow(import_table
, rn
, bpi
, 0);
3759 if (!original_had_routes
) {
3761 * We went from 0 usable routes to 1 usable route. Perform the
3762 * "Adding a Route" export process.
3764 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3765 vnc_zebra_add_prefix(bgp
, import_table
, rn
);
3768 * Check for nexthop change event
3769 * Note: the prefix_same() test below detects two situations:
3770 * 1. route is replaced, new route has different nexthop
3771 * 2. new route is added (original_nexthop is 0)
3773 struct prefix new_nexthop
;
3775 rfapiGetNexthop(attr
, &new_nexthop
);
3776 if (!prefix_same(&original_nexthop
, &new_nexthop
)) {
3778 * nexthop change event
3779 * vnc_direct_bgp_add_prefix() will recompute VN addr
3782 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3786 if (!(bgp
->rfapi_cfg
->flags
& BGP_VNC_CONFIG_CALLBACK_DISABLE
)) {
3787 for (n
= rn
; n
; n
= agg_node_parent(n
)) {
3788 // rfapiDoRouteCallback(import_table, n, NULL);
3790 rfapiMonitorItNodeChanged(import_table
, rn
, NULL
);
3792 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 0);
3796 static void rfapiBgpInfoFilteredImportBadSafi(
3797 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3798 void *rfd
, /* set for looped back routes */
3799 const struct prefix
*p
,
3800 const struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3801 afi_t afi
, struct prefix_rd
*prd
,
3802 struct attr
*attr
, /* part of bgp_path_info */
3803 uint8_t type
, /* part of bgp_path_info */
3804 uint8_t sub_type
, /* part of bgp_path_info */
3805 uint32_t *label
) /* part of bgp_path_info */
3807 vnc_zlog_debug_verbose("%s: Error, bad safi", __func__
);
3810 static rfapi_bi_filtered_import_f
*
3811 rfapiBgpInfoFilteredImportFunction(safi_t safi
)
3815 return rfapiBgpInfoFilteredImportVPN
;
3818 return rfapiBgpInfoFilteredImportEncap
;
3822 case SAFI_MULTICAST
:
3824 case SAFI_LABELED_UNICAST
:
3828 flog_err(EC_LIB_DEVELOPMENT
, "%s: bad safi %d", __func__
, safi
);
3829 return rfapiBgpInfoFilteredImportBadSafi
;
3832 assert(!"Reached end of function when we were not expecting to");
3835 void rfapiProcessUpdate(struct peer
*peer
,
3836 void *rfd
, /* set when looped from RFP/RFAPI */
3837 const struct prefix
*p
, struct prefix_rd
*prd
,
3838 struct attr
*attr
, afi_t afi
, safi_t safi
, uint8_t type
,
3839 uint8_t sub_type
, uint32_t *label
)
3843 struct rfapi_import_table
*it
;
3844 int has_ip_route
= 1;
3847 bgp
= bgp_get_default(); /* assume 1 instance for now */
3854 * look at high-order byte of RD. FF means MAC
3855 * address is present (VNC L2VPN)
3857 if ((safi
== SAFI_MPLS_VPN
)
3858 && (decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
)) {
3859 struct prefix pfx_mac_buf
;
3860 struct prefix pfx_nexthop_buf
;
3864 * Set flag if prefix and nexthop are the same - don't
3865 * add the route to normal IP-based import tables
3867 if (!rfapiGetNexthop(attr
, &pfx_nexthop_buf
)) {
3868 if (!prefix_cmp(&pfx_nexthop_buf
, p
)) {
3873 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
3874 pfx_mac_buf
.family
= AF_ETHERNET
;
3875 pfx_mac_buf
.prefixlen
= 48;
3876 memcpy(&pfx_mac_buf
.u
.prefix_eth
.octet
, prd
->val
+ 2, 6);
3879 * Find rt containing LNI (Logical Network ID), which
3880 * _should_ always be present when mac address is present
3882 rc
= rfapiEcommunityGetLNI(bgp_attr_get_ecommunity(attr
), &lni
);
3884 vnc_zlog_debug_verbose(
3885 "%s: rfapiEcommunityGetLNI returned %d, lni=%d, attr=%p",
3886 __func__
, rc
, lni
, attr
);
3888 it
= rfapiMacImportTableGet(bgp
, lni
);
3890 rfapiBgpInfoFilteredImportVPN(
3891 it
, FIF_ACTION_UPDATE
, peer
, rfd
,
3892 &pfx_mac_buf
, /* prefix */
3893 p
, /* aux prefix: IP addr */
3894 AFI_L2VPN
, prd
, attr
, type
, sub_type
, label
);
3902 * Iterate over all import tables; do a filtered import
3903 * for the afi/safi combination
3905 for (it
= h
->imports
; it
; it
= it
->next
) {
3906 (*rfapiBgpInfoFilteredImportFunction(safi
))(
3907 it
, FIF_ACTION_UPDATE
, peer
, rfd
, p
, /* prefix */
3908 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3911 if (safi
== SAFI_MPLS_VPN
) {
3912 vnc_direct_bgp_rh_add_route(bgp
, afi
, p
, peer
, attr
);
3913 rfapiBgpInfoFilteredImportVPN(
3914 bgp
->rfapi
->it_ce
, FIF_ACTION_UPDATE
, peer
, rfd
,
3916 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3921 void rfapiProcessWithdraw(struct peer
*peer
, void *rfd
, const struct prefix
*p
,
3922 struct prefix_rd
*prd
, struct attr
*attr
, afi_t afi
,
3923 safi_t safi
, uint8_t type
, int kill
)
3927 struct rfapi_import_table
*it
;
3929 bgp
= bgp_get_default(); /* assume 1 instance for now */
3936 * look at high-order byte of RD. FF means MAC
3937 * address is present (VNC L2VPN)
3939 if (h
->import_mac
!= NULL
&& safi
== SAFI_MPLS_VPN
3940 && decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
) {
3941 struct prefix pfx_mac_buf
;
3942 void *cursor
= NULL
;
3945 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
3946 pfx_mac_buf
.family
= AF_ETHERNET
;
3947 pfx_mac_buf
.prefixlen
= 48;
3948 memcpy(&pfx_mac_buf
.u
.prefix_eth
, prd
->val
+ 2, 6);
3951 * withdraw does not contain attrs, so we don't have
3952 * access to the route's LNI, which would ordinarily
3953 * select the specific mac-based import table. Instead,
3954 * we must iterate over all mac-based tables and rely
3955 * on the RD to match.
3957 * If this approach is too slow, add an index where
3958 * key is {RD, peer} and value is the import table
3960 for (rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
3962 rc
== 0; rc
= skiplist_next(h
->import_mac
, NULL
,
3963 (void **)&it
, &cursor
)) {
3965 #ifdef DEBUG_L2_EXTRA
3966 vnc_zlog_debug_verbose(
3967 "%s: calling rfapiBgpInfoFilteredImportVPN(it=%p, afi=AFI_L2VPN)",
3971 rfapiBgpInfoFilteredImportVPN(
3973 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
3974 peer
, rfd
, &pfx_mac_buf
, /* prefix */
3975 p
, /* aux_prefix: IP */
3976 AFI_L2VPN
, prd
, attr
, type
, 0,
3977 NULL
); /* sub_type & label unused for withdraw
3983 * XXX For the case where the withdraw involves an L2
3984 * route with no IP information, we rely on the lack
3985 * of RT-list intersection to filter out the withdraw
3986 * from the IP-based import tables below
3990 * Iterate over all import tables; do a filtered import
3991 * for the afi/safi combination
3994 for (it
= h
->imports
; it
; it
= it
->next
) {
3995 (*rfapiBgpInfoFilteredImportFunction(safi
))(
3996 it
, (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
3997 peer
, rfd
, p
, /* prefix */
3998 NULL
, afi
, prd
, attr
, type
, 0,
3999 NULL
); /* sub_type & label unused for withdraw */
4002 /* TBD the deletion should happen after the lifetime expires */
4003 if (safi
== SAFI_MPLS_VPN
)
4004 vnc_direct_bgp_rh_del_route(bgp
, afi
, p
, peer
);
4006 if (safi
== SAFI_MPLS_VPN
) {
4007 rfapiBgpInfoFilteredImportVPN(
4009 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
), peer
,
4010 rfd
, p
, /* prefix */
4011 NULL
, afi
, prd
, attr
, type
, 0,
4012 NULL
); /* sub_type & label unused for withdraw */
4017 * TBD optimized withdraw timer algorithm for case of many
4018 * routes expiring at the same time due to peer drop.
4021 * 1. Visit all BPIs in all ENCAP import tables.
4023 * a. If a bpi's peer is the failed peer, remove the bpi.
4024 * b. If the removed ENCAP bpi was first in the list of
4025 * BPIs at this ENCAP node, loop over all monitors
4028 * (1) for each ENCAP monitor, loop over all its
4029 * VPN node monitors and set their RFAPI_MON_FLAG_NEEDCALLBACK
4032 * 2. Visit all BPIs in all VPN import tables.
4033 * a. If a bpi's peer is the failed peer, remove the bpi.
4034 * b. loop over all the VPN node monitors and set their
4035 * RFAPI_MON_FLAG_NEEDCALLBACK flags
4036 * c. If there are no BPIs left at this VPN node,
4041 /* surprise, this gets called from peer_delete(), from rfapi_close() */
4042 static void rfapiProcessPeerDownRt(struct peer
*peer
,
4043 struct rfapi_import_table
*import_table
,
4044 afi_t afi
, safi_t safi
)
4046 struct agg_node
*rn
;
4047 struct bgp_path_info
*bpi
;
4048 struct agg_table
*rt
= NULL
;
4049 void (*timer_service_func
)(struct thread
*) = NULL
;
4051 assert(afi
== AFI_IP
|| afi
== AFI_IP6
);
4057 rt
= import_table
->imported_vpn
[afi
];
4058 timer_service_func
= rfapiWithdrawTimerVPN
;
4061 rt
= import_table
->imported_encap
[afi
];
4062 timer_service_func
= rfapiWithdrawTimerEncap
;
4066 case SAFI_MULTICAST
:
4068 case SAFI_LABELED_UNICAST
:
4071 /* Suppress uninitialized variable warning */
4073 timer_service_func
= NULL
;
4077 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
4078 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
4079 if (bpi
->peer
== peer
) {
4081 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
4082 /* already in holddown, skip */
4086 if (safi
== SAFI_MPLS_VPN
) {
4087 RFAPI_UPDATE_ITABLE_COUNT(
4088 bpi
, import_table
, afi
, -1);
4089 import_table
->holddown_count
[afi
] += 1;
4091 rfapiBiStartWithdrawTimer(import_table
, rn
, bpi
,
4093 timer_service_func
);
4101 * This gets called when a peer connection drops. We have to remove
4102 * all the routes from this peer.
4104 * Current approach is crude. TBD Optimize by setting fewer timers and
4105 * grouping withdrawn routes so we can generate callbacks more
4108 void rfapiProcessPeerDown(struct peer
*peer
)
4112 struct rfapi_import_table
*it
;
4115 * If this peer is a "dummy" peer structure atached to a RFAPI
4116 * nve_descriptor, we don't need to walk the import tables
4117 * because the routes are already withdrawn by rfapi_close()
4119 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_IS_RFAPI_HD
))
4123 * 1. Visit all BPIs in all ENCAP import tables.
4124 * Start withdraw timer on the BPIs that match peer.
4126 * 2. Visit All BPIs in all VPN import tables.
4127 * Start withdraw timer on the BPIs that match peer.
4130 bgp
= bgp_get_default(); /* assume 1 instance for now */
4137 for (it
= h
->imports
; it
; it
= it
->next
) {
4138 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_ENCAP
);
4139 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_ENCAP
);
4140 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_MPLS_VPN
);
4141 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_MPLS_VPN
);
4145 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4146 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4151 * Import an entire RIB (for an afi/safi) to an import table RIB,
4152 * filtered according to the import table's RT list
4154 * TBD: does this function need additions to match rfapiProcessUpdate()
4155 * for, e.g., L2 handling?
4157 static void rfapiBgpTableFilteredImport(struct bgp
*bgp
,
4158 struct rfapi_import_table
*it
,
4159 afi_t afi
, safi_t safi
)
4161 struct bgp_dest
*dest1
;
4162 struct bgp_dest
*dest2
;
4164 /* Only these SAFIs have 2-level RIBS */
4165 assert(safi
== SAFI_MPLS_VPN
|| safi
== SAFI_ENCAP
);
4168 * Now visit all the rd nodes and the nodes of all the
4169 * route tables attached to them, and import the routes
4170 * if they have matching route targets
4172 for (dest1
= bgp_table_top(bgp
->rib
[afi
][safi
]); dest1
;
4173 dest1
= bgp_route_next(dest1
)) {
4175 if (bgp_dest_has_bgp_path_info_data(dest1
)) {
4177 for (dest2
= bgp_table_top(
4178 bgp_dest_get_bgp_table_info(dest1
));
4179 dest2
; dest2
= bgp_route_next(dest2
)) {
4181 struct bgp_path_info
*bpi
;
4183 for (bpi
= bgp_dest_get_bgp_path_info(dest2
);
4184 bpi
; bpi
= bpi
->next
) {
4187 if (CHECK_FLAG(bpi
->flags
,
4192 label
= decode_label(
4193 &bpi
->extra
->label
[0]);
4194 (*rfapiBgpInfoFilteredImportFunction(
4196 it
, /* which import table */
4197 FIF_ACTION_UPDATE
, bpi
->peer
,
4199 bgp_dest_get_prefix(dest2
),
4201 (struct prefix_rd
*)
4202 bgp_dest_get_prefix(
4204 bpi
->attr
, bpi
->type
,
4205 bpi
->sub_type
, &label
);
4213 /* per-bgp-instance rfapi data */
4214 struct rfapi
*bgp_rfapi_new(struct bgp
*bgp
)
4218 struct rfapi_rfp_cfg
*cfg
= NULL
;
4219 struct rfapi_rfp_cb_methods
*cbm
= NULL
;
4221 assert(bgp
->rfapi_cfg
== NULL
);
4223 h
= XCALLOC(MTYPE_RFAPI
, sizeof(struct rfapi
));
4225 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4226 h
->un
[afi
] = agg_table_init();
4230 * initialize the ce import table
4232 h
->it_ce
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4233 sizeof(struct rfapi_import_table
));
4234 h
->it_ce
->imported_vpn
[AFI_IP
] = agg_table_init();
4235 h
->it_ce
->imported_vpn
[AFI_IP6
] = agg_table_init();
4236 h
->it_ce
->imported_encap
[AFI_IP
] = agg_table_init();
4237 h
->it_ce
->imported_encap
[AFI_IP6
] = agg_table_init();
4238 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4239 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4242 * Set up work queue for deferred rfapi_close operations
4244 h
->deferred_close_q
=
4245 work_queue_new(bm
->master
, "rfapi deferred close");
4246 h
->deferred_close_q
->spec
.workfunc
= rfapi_deferred_close_workfunc
;
4247 h
->deferred_close_q
->spec
.data
= h
;
4249 h
->rfp
= rfp_start(bm
->master
, &cfg
, &cbm
);
4250 bgp
->rfapi_cfg
= bgp_rfapi_cfg_new(cfg
);
4252 h
->rfp_methods
= *cbm
;
4257 void bgp_rfapi_destroy(struct bgp
*bgp
, struct rfapi
*h
)
4261 if (bgp
== NULL
|| h
== NULL
)
4264 if (h
->resolve_nve_nexthop
) {
4265 skiplist_free(h
->resolve_nve_nexthop
);
4266 h
->resolve_nve_nexthop
= NULL
;
4269 rfapiImportTableFlush(h
->it_ce
);
4271 if (h
->import_mac
) {
4272 struct rfapi_import_table
*it
;
4277 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4279 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4282 rfapiImportTableFlush(it
);
4283 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
4285 skiplist_free(h
->import_mac
);
4286 h
->import_mac
= NULL
;
4289 work_queue_free_and_null(&h
->deferred_close_q
);
4294 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4295 agg_table_finish(h
->un
[afi
]);
4298 XFREE(MTYPE_RFAPI_IMPORTTABLE
, h
->it_ce
);
4299 XFREE(MTYPE_RFAPI
, h
);
4302 struct rfapi_import_table
*
4303 rfapiImportTableRefAdd(struct bgp
*bgp
, struct ecommunity
*rt_import_list
,
4304 struct rfapi_nve_group_cfg
*rfg
)
4307 struct rfapi_import_table
*it
;
4313 for (it
= h
->imports
; it
; it
= it
->next
) {
4314 if (ecommunity_cmp(it
->rt_import_list
, rt_import_list
))
4318 vnc_zlog_debug_verbose("%s: matched it=%p", __func__
, it
);
4321 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4322 sizeof(struct rfapi_import_table
));
4323 it
->next
= h
->imports
;
4326 it
->rt_import_list
= ecommunity_dup(rt_import_list
);
4328 it
->monitor_exterior_orphans
=
4329 skiplist_new(0, NULL
, prefix_free_lists
);
4332 * fill import route tables from RIBs
4334 * Potential area for optimization. If this occurs when
4335 * tables are large (e.g., the operator adds a nve group
4336 * with a new RT list to a running system), it could take
4340 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4342 it
->imported_vpn
[afi
] = agg_table_init();
4343 it
->imported_encap
[afi
] = agg_table_init();
4345 rfapiBgpTableFilteredImport(bgp
, it
, afi
,
4347 rfapiBgpTableFilteredImport(bgp
, it
, afi
, SAFI_ENCAP
);
4349 vnc_import_bgp_exterior_redist_enable_it(bgp
, afi
, it
);
4359 * skiplist element free function
4361 static void delete_rem_pfx_na_free(void *na
)
4363 uint32_t *pCounter
= ((struct rfapi_nve_addr
*)na
)->info
;
4366 XFREE(MTYPE_RFAPI_NVE_ADDR
, na
);
4370 * Common deleter for IP and MAC import tables
4372 static void rfapiDeleteRemotePrefixesIt(
4373 struct bgp
*bgp
, struct rfapi_import_table
*it
, struct prefix
*un
,
4374 struct prefix
*vn
, struct prefix
*p
, int delete_active
,
4375 int delete_holddown
, uint32_t *pARcount
, uint32_t *pAHcount
,
4376 uint32_t *pHRcount
, uint32_t *pHHcount
,
4377 struct skiplist
*uniq_active_nves
, struct skiplist
*uniq_holddown_nves
)
4381 #ifdef DEBUG_L2_EXTRA
4383 char buf_pfx
[PREFIX_STRLEN
];
4386 prefix2str(p
, buf_pfx
, sizeof(buf_pfx
));
4392 vnc_zlog_debug_verbose(
4393 "%s: entry, p=%s, delete_active=%d, delete_holddown=%d",
4394 __func__
, buf_pfx
, delete_active
, delete_holddown
);
4398 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4400 struct agg_table
*rt
;
4401 struct agg_node
*rn
;
4403 if (p
&& (family2afi(p
->family
) != afi
)) {
4407 rt
= it
->imported_vpn
[afi
];
4411 vnc_zlog_debug_verbose("%s: scanning rt for afi=%d", __func__
,
4414 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
4415 struct bgp_path_info
*bpi
;
4416 struct bgp_path_info
*next
;
4417 const struct prefix
*rn_p
= agg_node_get_prefix(rn
);
4419 if (p
&& VNC_DEBUG(IMPORT_DEL_REMOTE
))
4420 vnc_zlog_debug_any("%s: want %pFX, have %pRN",
4423 if (p
&& prefix_cmp(p
, rn_p
))
4426 vnc_zlog_debug_verbose("%s: rn pfx=%pRN", __func__
, rn
);
4428 /* TBD is this valid for afi == AFI_L2VPN? */
4429 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
4431 for (bpi
= rn
->info
; bpi
; bpi
= next
) {
4440 vnc_zlog_debug_verbose("%s: examining bpi %p",
4443 if (!rfapiGetNexthop(bpi
->attr
, &qpt
))
4448 || !prefix_match(vn
, &qpt
)) {
4449 #ifdef DEBUG_L2_EXTRA
4450 vnc_zlog_debug_verbose(
4451 "%s: continue at vn && !qpt_valid || !prefix_match(vn, &qpt)",
4458 if (!rfapiGetUnAddrOfVpnBi(bpi
, &qct
))
4463 || !prefix_match(un
, &qct
)) {
4464 #ifdef DEBUG_L2_EXTRA
4465 vnc_zlog_debug_verbose(
4466 "%s: continue at un && !qct_valid || !prefix_match(un, &qct)",
4478 * If this route is waiting to be deleted
4480 * a previous withdraw, we must cancel its
4483 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
4484 if (!delete_holddown
)
4486 if (bpi
->extra
->vnc
.import
.timer
) {
4487 struct rfapi_withdraw
*wcb
=
4494 ->holddown_count
[afi
] -=
4496 RFAPI_UPDATE_ITABLE_COUNT(
4497 bpi
, wcb
->import_table
,
4499 XFREE(MTYPE_RFAPI_WITHDRAW
,
4502 bpi
->extra
->vnc
.import
4511 vnc_zlog_debug_verbose(
4512 "%s: deleting bpi %p (qct_valid=%d, qpt_valid=%d, delete_holddown=%d, delete_active=%d)",
4513 __func__
, bpi
, qct_valid
, qpt_valid
,
4514 delete_holddown
, delete_active
);
4520 if (qct_valid
&& qpt_valid
) {
4522 struct rfapi_nve_addr na
;
4523 struct rfapi_nve_addr
*nap
;
4525 memset(&na
, 0, sizeof(na
));
4526 assert(!rfapiQprefix2Raddr(&qct
,
4528 assert(!rfapiQprefix2Raddr(&qpt
,
4531 if (skiplist_search(
4534 : uniq_holddown_nves
),
4535 &na
, (void **)&nap
)) {
4539 MTYPE_RFAPI_NVE_ADDR
,
4543 nap
->info
= is_active
4549 : uniq_holddown_nves
),
4552 rfapiNveAddr2Str(nap
, line
,
4557 vnc_direct_bgp_rh_del_route(bgp
, afi
, rn_p
,
4560 RFAPI_UPDATE_ITABLE_COUNT(bpi
, it
, afi
, -1);
4561 it
->holddown_count
[afi
] += 1;
4562 rfapiExpireVpnNow(it
, rn
, bpi
, 1);
4564 vnc_zlog_debug_verbose(
4565 "%s: incrementing count (is_active=%d)",
4566 __func__
, is_active
);
4579 * For use by the "clear vnc prefixes" command
4581 /*------------------------------------------
4582 * rfapiDeleteRemotePrefixes
4584 * UI helper: For use by the "clear vnc prefixes" command
4587 * un if set, tunnel must match this prefix
4588 * vn if set, nexthop prefix must match this prefix
4589 * p if set, prefix must match this prefix
4590 * it if set, only look in this import table
4593 * pARcount number of active routes deleted
4594 * pAHcount number of active nves deleted
4595 * pHRcount number of holddown routes deleted
4596 * pHHcount number of holddown nves deleted
4600 --------------------------------------------*/
4601 void rfapiDeleteRemotePrefixes(struct prefix
*un
, struct prefix
*vn
,
4603 struct rfapi_import_table
*arg_it
,
4604 int delete_active
, int delete_holddown
,
4605 uint32_t *pARcount
, uint32_t *pAHcount
,
4606 uint32_t *pHRcount
, uint32_t *pHHcount
)
4610 struct rfapi_import_table
*it
;
4611 uint32_t deleted_holddown_route_count
= 0;
4612 uint32_t deleted_active_route_count
= 0;
4613 uint32_t deleted_holddown_nve_count
= 0;
4614 uint32_t deleted_active_nve_count
= 0;
4615 struct skiplist
*uniq_holddown_nves
;
4616 struct skiplist
*uniq_active_nves
;
4620 bgp
= bgp_get_default(); /* assume 1 instance for now */
4621 /* If no bgp instantiated yet, no vnc prefixes exist */
4628 uniq_holddown_nves
=
4629 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4631 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4634 * Iterate over all import tables; do a filtered import
4635 * for the afi/safi combination
4644 vnc_zlog_debug_verbose(
4645 "%s: calling rfapiDeleteRemotePrefixesIt() on (IP) import %p",
4648 rfapiDeleteRemotePrefixesIt(
4649 bgp
, it
, un
, vn
, p
, delete_active
, delete_holddown
,
4650 &deleted_active_route_count
, &deleted_active_nve_count
,
4651 &deleted_holddown_route_count
,
4652 &deleted_holddown_nve_count
, uniq_active_nves
,
4653 uniq_holddown_nves
);
4662 * Now iterate over L2 import tables
4664 if (h
->import_mac
&& !(p
&& (p
->family
!= AF_ETHERNET
))) {
4666 void *cursor
= NULL
;
4670 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4672 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4675 vnc_zlog_debug_verbose(
4676 "%s: calling rfapiDeleteRemotePrefixesIt() on import_mac %p",
4679 rfapiDeleteRemotePrefixesIt(
4680 bgp
, it
, un
, vn
, p
, delete_active
,
4681 delete_holddown
, &deleted_active_route_count
,
4682 &deleted_active_nve_count
,
4683 &deleted_holddown_route_count
,
4684 &deleted_holddown_nve_count
, uniq_active_nves
,
4685 uniq_holddown_nves
);
4690 * our custom element freeing function above counts as it deletes
4692 skiplist_free(uniq_holddown_nves
);
4693 skiplist_free(uniq_active_nves
);
4696 *pARcount
= deleted_active_route_count
;
4698 *pAHcount
= deleted_active_nve_count
;
4700 *pHRcount
= deleted_holddown_route_count
;
4702 *pHHcount
= deleted_holddown_nve_count
;
4707 /*------------------------------------------
4708 * rfapiCountRemoteRoutes
4710 * UI helper: count VRF routes from BGP side
4715 * pALRcount count of active local routes
4716 * pARRcount count of active remote routes
4717 * pHRcount count of holddown routes
4718 * pIRcount count of direct imported routes
4722 --------------------------------------------*/
4723 void rfapiCountAllItRoutes(int *pALRcount
, /* active local routes */
4724 int *pARRcount
, /* active remote routes */
4725 int *pHRcount
, /* holddown routes */
4726 int *pIRcount
) /* imported routes */
4730 struct rfapi_import_table
*it
;
4733 int total_active_local
= 0;
4734 int total_active_remote
= 0;
4735 int total_holddown
= 0;
4736 int total_imported
= 0;
4738 bgp
= bgp_get_default(); /* assume 1 instance for now */
4745 * Iterate over all import tables; do a filtered import
4746 * for the afi/safi combination
4749 for (it
= h
->imports
; it
; it
= it
->next
) {
4751 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4753 total_active_local
+= it
->local_count
[afi
];
4754 total_active_remote
+= it
->remote_count
[afi
];
4755 total_holddown
+= it
->holddown_count
[afi
];
4756 total_imported
+= it
->imported_count
[afi
];
4763 if (h
->import_mac
) {
4765 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4767 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4770 total_active_local
+= it
->local_count
[AFI_L2VPN
];
4771 total_active_remote
+= it
->remote_count
[AFI_L2VPN
];
4772 total_holddown
+= it
->holddown_count
[AFI_L2VPN
];
4773 total_imported
+= it
->imported_count
[AFI_L2VPN
];
4779 *pALRcount
= total_active_local
;
4782 *pARRcount
= total_active_remote
;
4785 *pHRcount
= total_holddown
;
4788 *pIRcount
= total_imported
;
4792 /*------------------------------------------
4793 * rfapiGetHolddownFromLifetime
4795 * calculate holddown value based on lifetime
4801 * Holddown value based on lifetime, holddown_factor,
4802 * and RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
4804 --------------------------------------------*/
4805 /* hold down time maxes out at RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY */
4806 uint32_t rfapiGetHolddownFromLifetime(uint32_t lifetime
)
4811 bgp
= bgp_get_default();
4812 if (bgp
&& bgp
->rfapi_cfg
)
4813 factor
= bgp
->rfapi_cfg
->rfp_cfg
.holddown_factor
;
4815 factor
= RFAPI_RFP_CFG_DEFAULT_HOLDDOWN_FACTOR
;
4817 if (factor
< 100 || lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4818 lifetime
= lifetime
* factor
/ 100;
4819 if (lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4822 return RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
;