3 * Copyright 2009-2016, LabN Consulting, L.L.C.
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version 2
9 * of the License, or (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License along
17 * with this program; see the file COPYING; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
22 * File: rfapi_import.c
23 * Purpose: Handle import of routes from BGP to RFAPI
28 #include "lib/zebra.h"
29 #include "lib/prefix.h"
30 #include "lib/table.h"
32 #include "lib/memory.h"
34 #include "lib/skiplist.h"
35 #include "lib/thread.h"
36 #include "lib/stream.h"
38 #include "bgpd/bgpd.h"
39 #include "bgpd/bgp_ecommunity.h"
40 #include "bgpd/bgp_attr.h"
41 #include "bgpd/bgp_route.h"
42 #include "bgpd/bgp_mplsvpn.h" /* prefix_rd2str() */
43 #include "bgpd/bgp_vnc_types.h"
44 #include "bgpd/bgp_rd.h"
46 #include "bgpd/rfapi/rfapi.h"
47 #include "bgpd/rfapi/bgp_rfapi_cfg.h"
48 #include "bgpd/rfapi/rfapi_backend.h"
49 #include "bgpd/rfapi/rfapi_import.h"
50 #include "bgpd/rfapi/rfapi_private.h"
51 #include "bgpd/rfapi/rfapi_monitor.h"
52 #include "bgpd/rfapi/rfapi_nve_addr.h"
53 #include "bgpd/rfapi/rfapi_vty.h"
54 #include "bgpd/rfapi/vnc_export_bgp.h"
55 #include "bgpd/rfapi/vnc_export_bgp_p.h"
56 #include "bgpd/rfapi/vnc_zebra.h"
57 #include "bgpd/rfapi/vnc_import_bgp.h"
58 #include "bgpd/rfapi/vnc_import_bgp_p.h"
59 #include "bgpd/rfapi/rfapi_rib.h"
60 #include "bgpd/rfapi/rfapi_encap_tlv.h"
61 #include "bgpd/rfapi/vnc_debug.h"
63 #ifdef HAVE_GLIBC_BACKTRACE
64 /* for backtrace and friends */
66 #endif /* HAVE_GLIBC_BACKTRACE */
68 #undef DEBUG_MONITOR_MOVE_SHORTER
69 #undef DEBUG_RETURNED_NHL
70 #undef DEBUG_ROUTE_COUNTERS
71 #undef DEBUG_ENCAP_MONITOR
74 #undef DEBUG_BI_SEARCH
77 * Allocated for each withdraw timer instance; freed when the timer
78 * expires or is canceled
80 struct rfapi_withdraw
{
81 struct rfapi_import_table
*import_table
;
82 struct route_node
*node
;
83 struct bgp_info
*info
;
84 safi_t safi
; /* used only for bulk operations */
86 * For import table node reference count checking (i.e., debugging).
87 * Normally when a timer expires, lockoffset should be 0. However, if
88 * the timer expiration function is called directly (e.g.,
89 * rfapiExpireVpnNow), the node could be locked by a preceding
90 * route_top() or route_next() in a loop, so we need to pass this
98 * It's evil and fiendish. It's compiler-dependent.
99 * ? Might need LDFLAGS -rdynamic to produce all function names
101 void rfapiDebugBacktrace(void)
103 #ifdef HAVE_GLIBC_BACKTRACE
104 #define RFAPI_DEBUG_BACKTRACE_NENTRIES 200
105 void *buf
[RFAPI_DEBUG_BACKTRACE_NENTRIES
];
110 size
= backtrace(buf
, RFAPI_DEBUG_BACKTRACE_NENTRIES
);
111 syms
= backtrace_symbols(buf
, size
);
113 for (i
= 0; i
< size
&& i
< RFAPI_DEBUG_BACKTRACE_NENTRIES
; ++i
) {
114 vnc_zlog_debug_verbose("backtrace[%2zu]: %s", i
, syms
[i
]);
124 * Count remote routes and compare with actively-maintained values.
125 * Abort if they disagree.
127 void rfapiCheckRouteCount()
129 struct bgp
*bgp
= bgp_get_default();
131 struct rfapi_import_table
*it
;
139 for (it
= h
->imports
; it
; it
= it
->next
) {
140 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
142 struct route_table
*rt
;
143 struct route_node
*rn
;
145 int holddown_count
= 0;
147 int imported_count
= 0;
148 int remote_count
= 0;
150 rt
= it
->imported_vpn
[afi
];
152 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
154 struct bgp_info
*next
;
156 for (bi
= rn
->info
; bi
; bi
= next
) {
159 if (CHECK_FLAG(bi
->flags
,
164 if (RFAPI_LOCAL_BI(bi
)) {
167 if (RFAPI_DIRECT_IMPORT_BI(
178 if (it
->holddown_count
[afi
] != holddown_count
) {
179 vnc_zlog_debug_verbose(
180 "%s: it->holddown_count %d != holddown_count %d",
181 __func__
, it
->holddown_count
[afi
],
185 if (it
->remote_count
[afi
] != remote_count
) {
186 vnc_zlog_debug_verbose(
187 "%s: it->remote_count %d != remote_count %d",
188 __func__
, it
->remote_count
[afi
],
192 if (it
->imported_count
[afi
] != imported_count
) {
193 vnc_zlog_debug_verbose(
194 "%s: it->imported_count %d != imported_count %d",
195 __func__
, it
->imported_count
[afi
],
203 #if DEBUG_ROUTE_COUNTERS
204 #define VNC_ITRCCK do {rfapiCheckRouteCount();} while (0)
210 * Validate reference count for a node in an import table
212 * Normally lockoffset is 0 for nodes in quiescent state. However,
213 * route_unlock_node will delete the node if it is called when
214 * node->lock == 1, and we have to validate the refcount before
215 * the node is deleted. In this case, we specify lockoffset 1.
217 void rfapiCheckRefcount(struct route_node
*rn
, safi_t safi
, int lockoffset
)
219 unsigned int count_bi
= 0;
220 unsigned int count_monitor
= 0;
222 struct rfapi_monitor_encap
*hme
;
223 struct rfapi_monitor_vpn
*hmv
;
225 for (bi
= rn
->info
; bi
; bi
= bi
->next
)
230 ++count_monitor
; /* rfapi_it_extra */
237 for (hme
= RFAPI_MONITOR_ENCAP(rn
); hme
;
244 for (hmv
= RFAPI_MONITOR_VPN(rn
); hmv
; hmv
= hmv
->next
)
247 if (RFAPI_MONITOR_EXTERIOR(rn
)->source
) {
248 ++count_monitor
; /* sl */
250 for (rc
= skiplist_next(
251 RFAPI_MONITOR_EXTERIOR(rn
)->source
,
252 NULL
, NULL
, &cursor
);
255 RFAPI_MONITOR_EXTERIOR(rn
)->source
,
256 NULL
, NULL
, &cursor
)) {
258 ++count_monitor
; /* sl entry */
268 if (count_bi
+ count_monitor
+ lockoffset
!= rn
->lock
) {
269 vnc_zlog_debug_verbose(
270 "%s: count_bi=%d, count_monitor=%d, lockoffset=%d, rn->lock=%d",
271 __func__
, count_bi
, count_monitor
, lockoffset
,
278 * Perform deferred rfapi_close operations that were queued
281 static wq_item_status
rfapi_deferred_close_workfunc(struct work_queue
*q
,
284 struct rfapi_descriptor
*rfd
= data
;
285 struct rfapi
*h
= q
->spec
.data
;
287 assert(!(h
->flags
& RFAPI_INCALLBACK
));
289 vnc_zlog_debug_verbose("%s: completed deferred close on handle %p",
295 * Extract layer 2 option from Encap TLVS in BGP attrs
297 int rfapiGetL2o(struct attr
*attr
, struct rfapi_l2address_option
*l2o
)
301 struct bgp_attr_encap_subtlv
*pEncap
;
303 for (pEncap
= attr
->vnc_subtlvs
; pEncap
;
304 pEncap
= pEncap
->next
) {
306 if (pEncap
->type
== BGP_VNC_SUBTLV_TYPE_RFPOPTION
) {
308 == RFAPI_VN_OPTION_TYPE_L2ADDR
) {
310 if (pEncap
->value
[1] == 14) {
311 memcpy(l2o
->macaddr
.octet
,
328 l2o
->logical_net_id
=
331 + ((pEncap
->value
[14]
334 + ((pEncap
->value
[13]
349 * Extract the lifetime from the Tunnel Encap attribute of a route in
352 int rfapiGetVncLifetime(struct attr
*attr
, uint32_t *lifetime
)
354 struct bgp_attr_encap_subtlv
*pEncap
;
356 *lifetime
= RFAPI_INFINITE_LIFETIME
; /* default to infinite */
360 for (pEncap
= attr
->vnc_subtlvs
; pEncap
;
361 pEncap
= pEncap
->next
) {
364 == BGP_VNC_SUBTLV_TYPE_LIFETIME
) { /* lifetime */
365 if (pEncap
->length
== 4) {
366 memcpy(lifetime
, pEncap
->value
, 4);
367 *lifetime
= ntohl(*lifetime
);
378 * Extract the tunnel type from the extended community
380 int rfapiGetTunnelType(struct attr
*attr
, bgp_encap_types
*type
)
382 *type
= BGP_ENCAP_TYPE_MPLS
; /* default to MPLS */
383 if (attr
&& attr
->ecommunity
) {
384 struct ecommunity
*ecom
= attr
->ecommunity
;
387 for (i
= 0; i
< (ecom
->size
* ECOMMUNITY_SIZE
);
388 i
+= ECOMMUNITY_SIZE
) {
392 if (ep
[0] == ECOMMUNITY_ENCODE_OPAQUE
393 && ep
[1] == ECOMMUNITY_OPAQUE_SUBTYPE_ENCAP
) {
394 *type
= (ep
[6] << 8) + ep
[7];
405 * Look for UN address in Encap attribute
407 int rfapiGetVncTunnelUnAddr(struct attr
*attr
, struct prefix
*p
)
409 struct bgp_attr_encap_subtlv
*pEncap
;
410 bgp_encap_types tun_type
;
412 rfapiGetTunnelType(attr
, &tun_type
);
413 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
416 /* MPLS carries UN address in next hop */
417 rfapiNexthop2Prefix(attr
, p
);
424 for (pEncap
= attr
->encap_subtlvs
; pEncap
;
425 pEncap
= pEncap
->next
) {
428 == BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
) { /* un
431 switch (pEncap
->length
) {
436 memcpy(p
->u
.val
, pEncap
->value
,
443 p
->family
= AF_INET6
;
445 memcpy(p
->u
.val
, pEncap
->value
,
458 * Get UN address wherever it might be
460 int rfapiGetUnAddrOfVpnBi(struct bgp_info
*bi
, struct prefix
*p
)
462 /* If it's in this route's VNC attribute, we're done */
463 if (!rfapiGetVncTunnelUnAddr(bi
->attr
, p
))
466 * Otherwise, see if it's cached from a corresponding ENCAP SAFI
470 switch (bi
->extra
->vnc
.import
.un_family
) {
473 p
->family
= bi
->extra
->vnc
.import
.un_family
;
474 p
->u
.prefix4
= bi
->extra
->vnc
.import
.un
.addr4
;
480 p
->family
= bi
->extra
->vnc
.import
.un_family
;
481 p
->u
.prefix6
= bi
->extra
->vnc
.import
.un
.addr6
;
488 #if DEBUG_ENCAP_MONITOR
489 vnc_zlog_debug_verbose(
490 "%s: bi->extra->vnc.import.un_family is 0, no UN addr",
502 * Make a new bgp_info from gathered parameters
504 static struct bgp_info
*rfapiBgpInfoCreate(struct attr
*attr
, struct peer
*peer
,
505 void *rfd
, struct prefix_rd
*prd
,
506 uint8_t type
, uint8_t sub_type
,
509 struct bgp_info
*new;
511 new = bgp_info_new();
516 new->attr
= bgp_attr_intern(attr
);
518 bgp_info_extra_get(new);
520 new->extra
->vnc
.import
.rd
= *prd
;
521 rfapi_time(&new->extra
->vnc
.import
.create_time
);
524 encode_label(*label
, &new->extra
->label
[0]);
526 new->sub_type
= sub_type
;
534 * Frees bgp_info as used in import tables (parts are not
535 * allocated exactly the way they are in the main RIBs)
537 static void rfapiBgpInfoFree(struct bgp_info
*goner
)
543 vnc_zlog_debug_verbose("%s: calling peer_unlock(%p), #%d",
544 __func__
, goner
->peer
,
546 peer_unlock(goner
->peer
);
550 bgp_attr_unintern(&goner
->attr
);
553 assert(!goner
->extra
->damp_info
); /* Not used in import tbls */
554 XFREE(MTYPE_BGP_ROUTE_EXTRA
, goner
->extra
);
557 XFREE(MTYPE_BGP_ROUTE
, goner
);
560 struct rfapi_import_table
*rfapiMacImportTableGetNoAlloc(struct bgp
*bgp
,
564 struct rfapi_import_table
*it
= NULL
;
565 uintptr_t lni_as_ptr
= lni
;
574 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
))
580 struct rfapi_import_table
*rfapiMacImportTableGet(struct bgp
*bgp
, uint32_t lni
)
583 struct rfapi_import_table
*it
= NULL
;
584 uintptr_t lni_as_ptr
= lni
;
589 if (!h
->import_mac
) {
590 /* default cmp is good enough for LNI */
591 h
->import_mac
= skiplist_new(0, NULL
, NULL
);
594 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
)) {
596 struct ecommunity
*enew
;
597 struct ecommunity_val eval
;
600 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
601 sizeof(struct rfapi_import_table
));
602 /* set RT list of new import table based on LNI */
603 memset((char *)&eval
, 0, sizeof(eval
));
604 eval
.val
[0] = 0; /* VNC L2VPN */
605 eval
.val
[1] = 2; /* VNC L2VPN */
606 eval
.val
[5] = (lni
>> 16) & 0xff;
607 eval
.val
[6] = (lni
>> 8) & 0xff;
608 eval
.val
[7] = (lni
>> 0) & 0xff;
610 enew
= ecommunity_new();
611 ecommunity_add_val(enew
, &eval
);
612 it
->rt_import_list
= enew
;
614 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
615 it
->imported_vpn
[afi
] = route_table_init();
616 it
->imported_encap
[afi
] = route_table_init();
619 it
->l2_logical_net_id
= lni
;
621 skiplist_insert(h
->import_mac
, (void *)lni_as_ptr
, it
);
629 * Implement MONITOR_MOVE_SHORTER(original_node) from
630 * RFAPI-Import-Event-Handling.txt
632 * Returns pointer to the list of moved monitors
634 static struct rfapi_monitor_vpn
*
635 rfapiMonitorMoveShorter(struct route_node
*original_vpn_node
, int lockoffset
)
638 struct route_node
*par
;
639 struct rfapi_monitor_vpn
*m
;
640 struct rfapi_monitor_vpn
*mlast
;
641 struct rfapi_monitor_vpn
*moved
;
643 int parent_already_refcounted
= 0;
645 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
, lockoffset
);
647 #if DEBUG_MONITOR_MOVE_SHORTER
649 char buf
[PREFIX_STRLEN
];
651 prefix2str(&original_vpn_node
->p
, buf
, sizeof(buf
));
652 vnc_zlog_debug_verbose("%s: called with node pfx=%s", __func__
,
658 * 1. If there is at least one bi (either regular route or
659 * route marked as withdrawn, with a pending timer) at
660 * original_node with a valid UN address, we're done. Return.
662 for (bi
= original_vpn_node
->info
; bi
; bi
= bi
->next
) {
665 if (!rfapiGetUnAddrOfVpnBi(bi
, &pfx
)) {
666 #if DEBUG_MONITOR_MOVE_SHORTER
667 vnc_zlog_debug_verbose(
668 "%s: have valid UN at original node, no change",
676 * 2. Travel up the tree (toward less-specific prefixes) from
677 * original_node to find the first node that has at least
678 * one route (even if it is only a withdrawn route) with a
679 * valid UN address. Call this node "Node P."
681 for (par
= original_vpn_node
->parent
; par
; par
= par
->parent
) {
682 for (bi
= par
->info
; bi
; bi
= bi
->next
) {
684 if (!rfapiGetUnAddrOfVpnBi(bi
, &pfx
)) {
693 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 0);
697 * If no less-specific routes, try to use the 0/0 node
700 /* this isn't necessarily 0/0 */
701 par
= route_top(original_vpn_node
->table
);
704 * If we got the top node but it wasn't 0/0,
707 if (par
&& par
->p
.prefixlen
) {
708 route_unlock_node(par
); /* maybe free */
713 ++parent_already_refcounted
;
718 * Create 0/0 node if it isn't there
721 struct prefix pfx_default
;
723 memset(&pfx_default
, 0, sizeof(pfx_default
));
724 pfx_default
.family
= original_vpn_node
->p
.family
;
726 /* creates default node if none exists */
727 par
= route_node_get(original_vpn_node
->table
, &pfx_default
);
728 ++parent_already_refcounted
;
732 * 3. Move each of the monitors found at original_node to Node P.
733 * These are "Moved Monitors."
738 * Attach at end so that the list pointer we return points
739 * only to the moved routes
741 for (m
= RFAPI_MONITOR_VPN(par
), mlast
= NULL
; m
;
742 mlast
= m
, m
= m
->next
)
746 moved
= mlast
->next
= RFAPI_MONITOR_VPN(original_vpn_node
);
748 moved
= RFAPI_MONITOR_VPN_W_ALLOC(par
) =
749 RFAPI_MONITOR_VPN(original_vpn_node
);
751 if (RFAPI_MONITOR_VPN(
752 original_vpn_node
)) /* check agg, so not allocated */
753 RFAPI_MONITOR_VPN_W_ALLOC(original_vpn_node
) = NULL
;
756 * update the node pointers on the monitors
758 for (m
= moved
; m
; m
= m
->next
) {
763 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
,
764 parent_already_refcounted
- movecount
);
765 while (movecount
> parent_already_refcounted
) {
766 route_lock_node(par
);
767 ++parent_already_refcounted
;
769 while (movecount
< parent_already_refcounted
) {
770 /* unlikely, but code defensively */
771 route_unlock_node(par
);
772 --parent_already_refcounted
;
774 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
,
775 movecount
+ lockoffset
);
776 while (movecount
--) {
777 route_unlock_node(original_vpn_node
);
780 #if DEBUG_MONITOR_MOVE_SHORTER
782 char buf
[PREFIX_STRLEN
];
784 prefix2str(&par
->p
, buf
, sizeof(buf
));
785 vnc_zlog_debug_verbose("%s: moved to node pfx=%s", __func__
,
795 * Implement MONITOR_MOVE_LONGER(new_node) from
796 * RFAPI-Import-Event-Handling.txt
798 static void rfapiMonitorMoveLonger(struct route_node
*new_vpn_node
)
800 struct rfapi_monitor_vpn
*monitor
;
801 struct rfapi_monitor_vpn
*mlast
;
803 struct route_node
*par
;
805 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
808 * Make sure we have at least one valid route at the new node
810 for (bi
= new_vpn_node
->info
; bi
; bi
= bi
->next
) {
812 if (!rfapiGetUnAddrOfVpnBi(bi
, &pfx
))
817 vnc_zlog_debug_verbose(
818 "%s: no valid routes at node %p, so not attempting moves",
819 __func__
, new_vpn_node
);
824 * Find first parent node that has monitors
826 for (par
= new_vpn_node
->parent
; par
; par
= par
->parent
) {
827 if (RFAPI_MONITOR_VPN(par
))
832 vnc_zlog_debug_verbose(
833 "%s: no parent nodes with monitors, done", __func__
);
838 * Check each of these monitors to see of their longest-match
839 * is now the updated node. Move any such monitors to the more-
840 * specific updated node
842 for (mlast
= NULL
, monitor
= RFAPI_MONITOR_VPN(par
); monitor
;) {
845 * If new longest match for monitor prefix is the new
846 * route's prefix, move monitor to new route's prefix
848 if (prefix_match(&new_vpn_node
->p
, &monitor
->p
)) {
851 mlast
->next
= monitor
->next
;
853 RFAPI_MONITOR_VPN_W_ALLOC(par
) = monitor
->next
;
858 monitor
->next
= RFAPI_MONITOR_VPN(new_vpn_node
);
859 RFAPI_MONITOR_VPN_W_ALLOC(new_vpn_node
) = monitor
;
860 monitor
->node
= new_vpn_node
;
862 route_lock_node(new_vpn_node
); /* incr refcount */
864 monitor
= mlast
? mlast
->next
: RFAPI_MONITOR_VPN(par
);
866 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 1);
867 /* decr refcount after we're done with par as this might
869 route_unlock_node(par
);
874 monitor
= monitor
->next
;
877 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
881 static void rfapiBgpInfoChainFree(struct bgp_info
*bi
)
883 struct bgp_info
*next
;
888 * If there is a timer waiting to delete this bi, cancel
889 * the timer and delete immediately
891 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
892 && bi
->extra
->vnc
.import
.timer
) {
895 (struct thread
*)bi
->extra
->vnc
.import
.timer
;
896 struct rfapi_withdraw
*wcb
= t
->arg
;
898 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
904 rfapiBgpInfoFree(bi
);
909 static void rfapiImportTableFlush(struct rfapi_import_table
*it
)
916 ecommunity_free(&it
->rt_import_list
);
917 it
->rt_import_list
= NULL
;
919 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
921 struct route_node
*rn
;
923 for (rn
= route_top(it
->imported_vpn
[afi
]); rn
;
924 rn
= route_next(rn
)) {
926 * Each route_node has:
927 * aggregate: points to rfapi_it_extra with monitor
929 * info: points to chain of bgp_info
931 /* free bgp_info and its children */
932 rfapiBgpInfoChainFree(rn
->info
);
935 rfapiMonitorExtraFlush(SAFI_MPLS_VPN
, rn
);
938 for (rn
= route_top(it
->imported_encap
[afi
]); rn
;
939 rn
= route_next(rn
)) {
940 /* free bgp_info and its children */
941 rfapiBgpInfoChainFree(rn
->info
);
944 rfapiMonitorExtraFlush(SAFI_ENCAP
, rn
);
947 route_table_finish(it
->imported_vpn
[afi
]);
948 route_table_finish(it
->imported_encap
[afi
]);
950 if (it
->monitor_exterior_orphans
) {
951 skiplist_free(it
->monitor_exterior_orphans
);
955 void rfapiImportTableRefDelByIt(struct bgp
*bgp
,
956 struct rfapi_import_table
*it_target
)
959 struct rfapi_import_table
*it
;
960 struct rfapi_import_table
*prev
= NULL
;
967 for (it
= h
->imports
; it
; prev
= it
, it
= it
->next
) {
973 assert(it
->refcount
);
979 prev
->next
= it
->next
;
981 h
->imports
= it
->next
;
983 rfapiImportTableFlush(it
);
984 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
988 #if RFAPI_REQUIRE_ENCAP_BEEC
990 * Look for magic BGP Encapsulation Extended Community value
991 * Format in RFC 5512 Sect. 4.5
993 static int rfapiEcommunitiesMatchBeec(struct ecommunity
*ecom
,
994 bgp_encap_types type
)
1001 for (i
= 0; i
< (ecom
->size
* ECOMMUNITY_SIZE
); i
+= ECOMMUNITY_SIZE
) {
1007 if (ep
[0] == ECOMMUNITY_ENCODE_OPAQUE
1008 && ep
[1] == ECOMMUNITY_OPAQUE_SUBTYPE_ENCAP
1009 && ep
[6] == ((type
&& 0xff00) >> 8)
1010 && ep
[7] == (type
& 0xff)) {
1019 int rfapiEcommunitiesIntersect(struct ecommunity
*e1
, struct ecommunity
*e2
)
1028 s1
= ecommunity_ecom2str(e1
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
1029 s2
= ecommunity_ecom2str(e2
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
1030 vnc_zlog_debug_verbose("%s: e1[%s], e2[%s]", __func__
, s1
, s2
);
1031 XFREE(MTYPE_ECOMMUNITY_STR
, s1
);
1032 XFREE(MTYPE_ECOMMUNITY_STR
, s2
);
1035 for (i
= 0; i
< e1
->size
; ++i
) {
1036 for (j
= 0; j
< e2
->size
; ++j
) {
1037 if (!memcmp(e1
->val
+ (i
* ECOMMUNITY_SIZE
),
1038 e2
->val
+ (j
* ECOMMUNITY_SIZE
),
1048 int rfapiEcommunityGetLNI(struct ecommunity
*ecom
, uint32_t *lni
)
1052 for (i
= 0; i
< ecom
->size
; ++i
) {
1053 uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1055 if ((*(p
+ 0) == 0x00) && (*(p
+ 1) == 0x02)) {
1057 *lni
= (*(p
+ 5) << 16) | (*(p
+ 6) << 8)
1066 int rfapiEcommunityGetEthernetTag(struct ecommunity
*ecom
, uint16_t *tag_id
)
1068 struct bgp
*bgp
= bgp_get_default();
1069 *tag_id
= 0; /* default to untagged */
1072 for (i
= 0; i
< ecom
->size
; ++i
) {
1075 uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1077 /* High-order octet of type. */
1080 if (*p
++ == ECOMMUNITY_ROUTE_TARGET
) {
1081 if (encode
== ECOMMUNITY_ENCODE_AS4
) {
1082 p
= ptr_get_be32(p
, &as
);
1083 } else if (encode
== ECOMMUNITY_ENCODE_AS
) {
1086 p
+= 2; /* skip next two, tag/vid
1087 always in lowest bytes */
1089 if (as
== bgp
->as
) {
1090 *tag_id
= *p
++ << 8;
1100 static int rfapiVpnBiNhEqualsPt(struct bgp_info
*bi
, struct rfapi_ip_addr
*hpt
)
1107 family
= BGP_MP_NEXTHOP_FAMILY(bi
->attr
->mp_nexthop_len
);
1109 if (hpt
->addr_family
!= family
)
1114 if (bi
->attr
->mp_nexthop_global_in
.s_addr
1115 != hpt
->addr
.v4
.s_addr
)
1120 if (IPV6_ADDR_CMP(&bi
->attr
->mp_nexthop_global
, &hpt
->addr
.v6
))
1134 * Compare 2 VPN BIs. Return true if they have the same VN and UN addresses
1136 static int rfapiVpnBiSamePtUn(struct bgp_info
*bi1
, struct bgp_info
*bi2
)
1138 struct prefix pfx_un1
;
1139 struct prefix pfx_un2
;
1144 if (!bi1
->attr
|| !bi2
->attr
)
1148 * VN address comparisons
1151 if (BGP_MP_NEXTHOP_FAMILY(bi1
->attr
->mp_nexthop_len
)
1152 != BGP_MP_NEXTHOP_FAMILY(bi2
->attr
->mp_nexthop_len
)) {
1156 switch (BGP_MP_NEXTHOP_FAMILY(bi1
->attr
->mp_nexthop_len
)) {
1158 if (bi1
->attr
->mp_nexthop_global_in
.s_addr
1159 != bi2
->attr
->mp_nexthop_global_in
.s_addr
)
1164 if (IPV6_ADDR_CMP(&bi1
->attr
->mp_nexthop_global
,
1165 &bi2
->attr
->mp_nexthop_global
))
1175 * UN address comparisons
1177 if (rfapiGetVncTunnelUnAddr(bi1
->attr
, &pfx_un1
)) {
1179 pfx_un1
.family
= bi1
->extra
->vnc
.import
.un_family
;
1180 switch (bi1
->extra
->vnc
.import
.un_family
) {
1183 bi1
->extra
->vnc
.import
.un
.addr4
;
1187 bi1
->extra
->vnc
.import
.un
.addr6
;
1196 if (rfapiGetVncTunnelUnAddr(bi2
->attr
, &pfx_un2
)) {
1198 pfx_un2
.family
= bi2
->extra
->vnc
.import
.un_family
;
1199 switch (bi2
->extra
->vnc
.import
.un_family
) {
1202 bi2
->extra
->vnc
.import
.un
.addr4
;
1206 bi2
->extra
->vnc
.import
.un
.addr6
;
1215 if (!pfx_un1
.family
|| !pfx_un2
.family
)
1218 if (pfx_un1
.family
!= pfx_un2
.family
)
1221 switch (pfx_un1
.family
) {
1223 if (!IPV4_ADDR_SAME(&pfx_un1
.u
.prefix4
, &pfx_un2
.u
.prefix4
))
1227 if (!IPV6_ADDR_SAME(&pfx_un1
.u
.prefix6
, &pfx_un2
.u
.prefix6
))
1236 uint8_t rfapiRfpCost(struct attr
*attr
)
1238 if (attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_LOCAL_PREF
)) {
1239 if (attr
->local_pref
> 255) {
1242 return 255 - attr
->local_pref
;
1248 /*------------------------------------------
1251 * Find Layer 2 options in an option chain
1257 * l2o layer 2 options extracted
1261 * 1 no options found
1263 --------------------------------------------*/
1264 int rfapi_extract_l2o(
1265 struct bgp_tea_options
*pHop
, /* chain of options */
1266 struct rfapi_l2address_option
*l2o
) /* return extracted value */
1268 struct bgp_tea_options
*p
;
1270 for (p
= pHop
; p
; p
= p
->next
) {
1271 if ((p
->type
== RFAPI_VN_OPTION_TYPE_L2ADDR
)
1272 && (p
->length
>= 8)) {
1276 memcpy(&l2o
->macaddr
, v
, 6);
1278 l2o
->label
= ((v
[6] << 12) & 0xff000)
1279 + ((v
[7] << 4) & 0xff0)
1280 + ((v
[8] >> 4) & 0xf);
1282 l2o
->local_nve_id
= (uint8_t)v
[10];
1284 l2o
->logical_net_id
=
1285 (v
[11] << 16) + (v
[12] << 8) + (v
[13] << 0);
1293 static struct rfapi_next_hop_entry
*
1294 rfapiRouteInfo2NextHopEntry(struct rfapi_ip_prefix
*rprefix
,
1295 struct bgp_info
*bi
, /* route to encode */
1296 uint32_t lifetime
, /* use this in nhe */
1297 struct route_node
*rn
) /* req for L2 eth addr */
1299 struct rfapi_next_hop_entry
*new;
1300 int have_vnc_tunnel_un
= 0;
1302 #if DEBUG_ENCAP_MONITOR
1303 vnc_zlog_debug_verbose("%s: entry, bi %p, rn %p", __func__
, bi
, rn
);
1306 new = XCALLOC(MTYPE_RFAPI_NEXTHOP
, sizeof(struct rfapi_next_hop_entry
));
1309 new->prefix
= *rprefix
;
1312 && decode_rd_type(bi
->extra
->vnc
.import
.rd
.val
)
1313 == RD_TYPE_VNC_ETH
) {
1316 struct rfapi_vn_option
*vo
;
1318 vo
= XCALLOC(MTYPE_RFAPI_VN_OPTION
,
1319 sizeof(struct rfapi_vn_option
));
1322 vo
->type
= RFAPI_VN_OPTION_TYPE_L2ADDR
;
1324 memcpy(&vo
->v
.l2addr
.macaddr
, &rn
->p
.u
.prefix_eth
.octet
,
1326 /* only low 3 bytes of this are significant */
1328 (void)rfapiEcommunityGetLNI(
1329 bi
->attr
->ecommunity
,
1330 &vo
->v
.l2addr
.logical_net_id
);
1331 (void)rfapiEcommunityGetEthernetTag(
1332 bi
->attr
->ecommunity
, &vo
->v
.l2addr
.tag_id
);
1335 /* local_nve_id comes from lower byte of RD type */
1336 vo
->v
.l2addr
.local_nve_id
= bi
->extra
->vnc
.import
.rd
.val
[1];
1338 /* label comes from MP_REACH_NLRI label */
1339 vo
->v
.l2addr
.label
= decode_label(&bi
->extra
->label
[0]);
1341 new->vn_options
= vo
;
1344 * If there is an auxiliary prefix (i.e., host IP address),
1345 * use it as the nexthop prefix instead of the query prefix
1347 if (bi
->extra
->vnc
.import
.aux_prefix
.family
) {
1348 rfapiQprefix2Rprefix(&bi
->extra
->vnc
.import
.aux_prefix
,
1354 bgp_encap_types tun_type
;
1355 new->prefix
.cost
= rfapiRfpCost(bi
->attr
);
1357 struct bgp_attr_encap_subtlv
*pEncap
;
1359 switch (BGP_MP_NEXTHOP_FAMILY(bi
->attr
->mp_nexthop_len
)) {
1361 new->vn_address
.addr_family
= AF_INET
;
1362 new->vn_address
.addr
.v4
=
1363 bi
->attr
->mp_nexthop_global_in
;
1367 new->vn_address
.addr_family
= AF_INET6
;
1368 new->vn_address
.addr
.v6
= bi
->attr
->mp_nexthop_global
;
1372 zlog_warn("%s: invalid vpn nexthop length: %d",
1373 __func__
, bi
->attr
->mp_nexthop_len
);
1374 rfapi_free_next_hop_list(new);
1378 for (pEncap
= bi
->attr
->vnc_subtlvs
; pEncap
;
1379 pEncap
= pEncap
->next
) {
1380 switch (pEncap
->type
) {
1381 case BGP_VNC_SUBTLV_TYPE_LIFETIME
:
1382 /* use configured lifetime, not attr lifetime */
1386 zlog_warn("%s: unknown VNC option type %d",
1387 __func__
, pEncap
->type
);
1394 rfapiGetTunnelType(bi
->attr
, &tun_type
);
1395 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
1397 /* MPLS carries UN address in next hop */
1398 rfapiNexthop2Prefix(bi
->attr
, &p
);
1399 if (p
.family
!= 0) {
1400 rfapiQprefix2Raddr(&p
, &new->un_address
);
1401 have_vnc_tunnel_un
= 1;
1405 for (pEncap
= bi
->attr
->encap_subtlvs
; pEncap
;
1406 pEncap
= pEncap
->next
) {
1407 switch (pEncap
->type
) {
1408 case BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
:
1410 * Overrides ENCAP UN address, if any
1412 switch (pEncap
->length
) {
1415 new->un_address
.addr_family
= AF_INET
;
1416 memcpy(&new->un_address
.addr
.v4
,
1418 have_vnc_tunnel_un
= 1;
1422 new->un_address
.addr_family
= AF_INET6
;
1423 memcpy(&new->un_address
.addr
.v6
,
1425 have_vnc_tunnel_un
= 1;
1430 "%s: invalid tunnel subtlv UN addr length (%d) for bi %p",
1431 __func__
, pEncap
->length
, bi
);
1437 "%s: unknown Encap Attribute option type %d",
1438 __func__
, pEncap
->type
);
1445 new->un_options
= rfapi_encap_tlv_to_un_option(bi
->attr
);
1447 #if DEBUG_ENCAP_MONITOR
1448 vnc_zlog_debug_verbose("%s: line %d: have_vnc_tunnel_un=%d",
1449 __func__
, __LINE__
, have_vnc_tunnel_un
);
1452 if (!have_vnc_tunnel_un
&& bi
->extra
) {
1454 * use cached UN address from ENCAP route
1456 new->un_address
.addr_family
=
1457 bi
->extra
->vnc
.import
.un_family
;
1458 switch (new->un_address
.addr_family
) {
1460 new->un_address
.addr
.v4
=
1461 bi
->extra
->vnc
.import
.un
.addr4
;
1464 new->un_address
.addr
.v6
=
1465 bi
->extra
->vnc
.import
.un
.addr6
;
1469 "%s: invalid UN addr family (%d) for bi %p",
1470 __func__
, new->un_address
.addr_family
,
1472 rfapi_free_next_hop_list(new);
1479 new->lifetime
= lifetime
;
1483 int rfapiHasNonRemovedRoutes(struct route_node
*rn
)
1485 struct bgp_info
*bi
;
1487 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
1490 if (!CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
1491 && (bi
->extra
&& !rfapiGetUnAddrOfVpnBi(bi
, &pfx
))) {
1503 void rfapiDumpNode(struct route_node
*rn
)
1505 struct bgp_info
*bi
;
1507 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
1508 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
1510 int ctrc
= rfapiGetUnAddrOfVpnBi(bi
, &pfx
);
1513 if (!CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
1514 && (bi
->extra
&& !ctrc
)) {
1521 vnc_zlog_debug_verbose(
1522 " bi=%p, nr=%d, flags=0x%x, extra=%p, ctrc=%d", bi
, nr
,
1523 bi
->flags
, bi
->extra
, ctrc
);
1528 static int rfapiNhlAddNodeRoutes(
1529 struct route_node
*rn
, /* in */
1530 struct rfapi_ip_prefix
*rprefix
, /* in */
1531 uint32_t lifetime
, /* in */
1532 int removed
, /* in */
1533 struct rfapi_next_hop_entry
**head
, /* in/out */
1534 struct rfapi_next_hop_entry
**tail
, /* in/out */
1535 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1536 struct route_node
*rfd_rib_node
, /* preload this NVE rib node */
1537 struct prefix
*pfx_target_original
) /* query target */
1539 struct bgp_info
*bi
;
1540 struct rfapi_next_hop_entry
*new;
1541 struct prefix pfx_un
;
1542 struct skiplist
*seen_nexthops
;
1544 int is_l2
= (rn
->p
.family
== AF_ETHERNET
);
1546 if (rfd_rib_node
&& rfd_rib_node
->table
&& rfd_rib_node
->table
->info
) {
1547 struct rfapi_descriptor
*rfd
;
1549 rfd
= (struct rfapi_descriptor
*)(rfd_rib_node
->table
->info
);
1551 if (rfapiRibFTDFilterRecentPrefix(
1552 rfd
, rn
, pfx_target_original
))
1558 skiplist_new(0, vnc_prefix_cmp
, (void (*)(void *))prefix_free
);
1560 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
1562 struct prefix pfx_vn
;
1563 struct prefix
*newpfx
;
1565 if (removed
&& !CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)) {
1566 #if DEBUG_RETURNED_NHL
1567 vnc_zlog_debug_verbose(
1568 "%s: want holddown, this route not holddown, skip",
1573 if (!removed
&& CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)) {
1582 * Check for excluded VN address
1584 if (rfapiVpnBiNhEqualsPt(bi
, exclude_vnaddr
))
1588 * Check for VN address (nexthop) copied already
1591 /* L2 routes: semantic nexthop in aux_prefix; VN addr
1593 pfx_vn
= bi
->extra
->vnc
.import
.aux_prefix
;
1595 rfapiNexthop2Prefix(bi
->attr
, &pfx_vn
);
1597 if (!skiplist_search(seen_nexthops
, &pfx_vn
, NULL
)) {
1598 #if DEBUG_RETURNED_NHL
1599 char buf
[PREFIX_STRLEN
];
1601 prefix2str(&pfx_vn
, buf
, sizeof(buf
));
1602 vnc_zlog_debug_verbose(
1603 "%s: already put VN/nexthop %s, skip", __func__
,
1609 if (rfapiGetUnAddrOfVpnBi(bi
, &pfx_un
)) {
1610 #if DEBUG_ENCAP_MONITOR
1611 vnc_zlog_debug_verbose(
1612 "%s: failed to get UN address of this VPN bi",
1618 newpfx
= prefix_new();
1620 skiplist_insert(seen_nexthops
, newpfx
, newpfx
);
1622 new = rfapiRouteInfo2NextHopEntry(rprefix
, bi
, lifetime
, rn
);
1624 if (rfapiRibPreloadBi(rfd_rib_node
, &pfx_vn
, &pfx_un
,
1626 /* duplicate filtered by RIB */
1627 rfapi_free_next_hop_list(new);
1634 (*tail
)->next
= new;
1643 skiplist_free(seen_nexthops
);
1652 * omit_node is meant for the situation where we are adding a subtree
1653 * of a parent of some original requested node. The response already
1654 * contains the original requested node, and we don't want to duplicate
1655 * its routes in the list, so we skip it if the right or left node
1656 * matches (of course, we still travel down its child subtrees).
1658 static int rfapiNhlAddSubtree(
1659 struct route_node
*rn
, /* in */
1660 uint32_t lifetime
, /* in */
1661 struct rfapi_next_hop_entry
**head
, /* in/out */
1662 struct rfapi_next_hop_entry
**tail
, /* in/out */
1663 struct route_node
*omit_node
, /* in */
1664 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1665 struct route_table
*rfd_rib_table
, /* preload here */
1666 struct prefix
*pfx_target_original
) /* query target */
1668 struct rfapi_ip_prefix rprefix
;
1671 /* FIXME: need to find a better way here to work without sticking our
1672 * hands in node->link */
1673 if (rn
->l_left
&& rn
->l_left
!= omit_node
) {
1674 if (rn
->l_left
->info
) {
1676 struct route_node
*rib_rn
= NULL
;
1678 rfapiQprefix2Rprefix(&rn
->l_left
->p
, &rprefix
);
1679 if (rfd_rib_table
) {
1680 rib_rn
= route_node_get(rfd_rib_table
,
1684 count
= rfapiNhlAddNodeRoutes(
1685 rn
->l_left
, &rprefix
, lifetime
, 0, head
, tail
,
1686 exclude_vnaddr
, rib_rn
, pfx_target_original
);
1688 count
= rfapiNhlAddNodeRoutes(
1689 rn
->l_left
, &rprefix
, lifetime
, 1, head
,
1690 tail
, exclude_vnaddr
, rib_rn
,
1691 pfx_target_original
);
1695 route_unlock_node(rib_rn
);
1699 if (rn
->l_right
&& rn
->l_right
!= omit_node
) {
1700 if (rn
->l_right
->info
) {
1702 struct route_node
*rib_rn
= NULL
;
1704 rfapiQprefix2Rprefix(&rn
->l_right
->p
, &rprefix
);
1705 if (rfd_rib_table
) {
1706 rib_rn
= route_node_get(rfd_rib_table
,
1709 count
= rfapiNhlAddNodeRoutes(
1710 rn
->l_right
, &rprefix
, lifetime
, 0, head
, tail
,
1711 exclude_vnaddr
, rib_rn
, pfx_target_original
);
1713 count
= rfapiNhlAddNodeRoutes(
1714 rn
->l_right
, &rprefix
, lifetime
, 1,
1715 head
, tail
, exclude_vnaddr
, rib_rn
,
1716 pfx_target_original
);
1720 route_unlock_node(rib_rn
);
1725 rcount
+= rfapiNhlAddSubtree(
1726 rn
->l_left
, lifetime
, head
, tail
, omit_node
,
1727 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1730 rcount
+= rfapiNhlAddSubtree(
1731 rn
->l_right
, lifetime
, head
, tail
, omit_node
,
1732 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1739 * Implementation of ROUTE_LIST(node) from RFAPI-Import-Event-Handling.txt
1741 * Construct an rfapi nexthop list based on the routes attached to
1742 * the specified node.
1744 * If there are any routes that do NOT have BGP_INFO_REMOVED set,
1745 * return those only. If there are ONLY routes with BGP_INFO_REMOVED,
1746 * then return those, and also include all the non-removed routes from the
1747 * next less-specific node (i.e., this node's parent) at the end.
1749 struct rfapi_next_hop_entry
*rfapiRouteNode2NextHopList(
1750 struct route_node
*rn
, uint32_t lifetime
, /* put into nexthop entries */
1751 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1752 struct route_table
*rfd_rib_table
, /* preload here */
1753 struct prefix
*pfx_target_original
) /* query target */
1755 struct rfapi_ip_prefix rprefix
;
1756 struct rfapi_next_hop_entry
*answer
= NULL
;
1757 struct rfapi_next_hop_entry
*last
= NULL
;
1758 struct route_node
*parent
;
1760 struct route_node
*rib_rn
;
1762 #if DEBUG_RETURNED_NHL
1764 char buf
[PREFIX_STRLEN
];
1766 prefix2str(&rn
->p
, buf
, sizeof(buf
));
1767 vnc_zlog_debug_verbose("%s: called with node pfx=%s", __func__
,
1770 rfapiDebugBacktrace();
1773 rfapiQprefix2Rprefix(&rn
->p
, &rprefix
);
1775 rib_rn
= rfd_rib_table
? route_node_get(rfd_rib_table
, &rn
->p
) : NULL
;
1778 * Add non-withdrawn routes at this node
1780 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 0, &answer
, &last
,
1781 exclude_vnaddr
, rib_rn
,
1782 pfx_target_original
);
1785 * If the list has at least one entry, it's finished
1788 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
, &last
, NULL
,
1789 exclude_vnaddr
, rfd_rib_table
,
1790 pfx_target_original
);
1791 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
,
1793 #if DEBUG_RETURNED_NHL
1794 rfapiPrintNhl(NULL
, answer
);
1797 route_unlock_node(rib_rn
);
1802 * Add withdrawn routes at this node
1804 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 1, &answer
, &last
,
1805 exclude_vnaddr
, rib_rn
,
1806 pfx_target_original
);
1808 route_unlock_node(rib_rn
);
1810 // rfapiPrintNhl(NULL, answer);
1813 * walk up the tree until we find a node with non-deleted
1814 * routes, then add them
1816 for (parent
= rn
->parent
; parent
; parent
= parent
->parent
) {
1817 if (rfapiHasNonRemovedRoutes(parent
)) {
1823 * Add non-withdrawn routes from less-specific prefix
1826 rib_rn
= rfd_rib_table
1827 ? route_node_get(rfd_rib_table
, &parent
->p
)
1829 rfapiQprefix2Rprefix(&parent
->p
, &rprefix
);
1830 count
+= rfapiNhlAddNodeRoutes(parent
, &rprefix
, lifetime
, 0,
1831 &answer
, &last
, exclude_vnaddr
,
1832 rib_rn
, pfx_target_original
);
1833 count
+= rfapiNhlAddSubtree(parent
, lifetime
, &answer
, &last
,
1834 rn
, exclude_vnaddr
, rfd_rib_table
,
1835 pfx_target_original
);
1837 route_unlock_node(rib_rn
);
1840 * There is no parent with non-removed routes. Still need to
1841 * add subtree of original node if it contributed routes to the
1845 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
,
1846 &last
, rn
, exclude_vnaddr
,
1848 pfx_target_original
);
1851 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
, count
,
1853 #if DEBUG_RETURNED_NHL
1854 rfapiPrintNhl(NULL
, answer
);
1860 * Construct nexthop list of all routes in table
1862 struct rfapi_next_hop_entry
*rfapiRouteTable2NextHopList(
1863 struct route_table
*rt
,
1864 uint32_t lifetime
, /* put into nexthop entries */
1865 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1866 struct route_table
*rfd_rib_table
, /* preload this NVE rib table */
1867 struct prefix
*pfx_target_original
) /* query target */
1869 struct route_node
*rn
;
1870 struct rfapi_next_hop_entry
*biglist
= NULL
;
1871 struct rfapi_next_hop_entry
*nhl
;
1872 struct rfapi_next_hop_entry
*tail
= NULL
;
1875 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
1877 nhl
= rfapiRouteNode2NextHopList(rn
, lifetime
, exclude_vnaddr
,
1879 pfx_target_original
);
1881 tail
= biglist
= nhl
;
1888 while (tail
->next
) {
1895 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1899 struct rfapi_next_hop_entry
*rfapiEthRouteNode2NextHopList(
1900 struct route_node
*rn
, struct rfapi_ip_prefix
*rprefix
,
1901 uint32_t lifetime
, /* put into nexthop entries */
1902 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1903 struct route_table
*rfd_rib_table
, /* preload NVE rib table */
1904 struct prefix
*pfx_target_original
) /* query target */
1907 struct rfapi_next_hop_entry
*answer
= NULL
;
1908 struct rfapi_next_hop_entry
*last
= NULL
;
1909 struct route_node
*rib_rn
;
1911 rib_rn
= rfd_rib_table
? route_node_get(rfd_rib_table
, &rn
->p
) : NULL
;
1913 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 0, &answer
, &last
,
1914 NULL
, rib_rn
, pfx_target_original
);
1916 #if DEBUG_ENCAP_MONITOR
1917 vnc_zlog_debug_verbose("%s: node %p: %d non-holddown routes", __func__
,
1922 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 1, &answer
,
1923 &last
, exclude_vnaddr
, rib_rn
,
1924 pfx_target_original
);
1925 vnc_zlog_debug_verbose("%s: node %p: %d holddown routes",
1926 __func__
, rn
, count
);
1930 route_unlock_node(rib_rn
);
1932 #if DEBUG_RETURNED_NHL
1933 rfapiPrintNhl(NULL
, answer
);
1941 * Construct nexthop list of all routes in table
1943 struct rfapi_next_hop_entry
*rfapiEthRouteTable2NextHopList(
1944 uint32_t logical_net_id
, struct rfapi_ip_prefix
*rprefix
,
1945 uint32_t lifetime
, /* put into nexthop entries */
1946 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1947 struct route_table
*rfd_rib_table
, /* preload NVE rib node */
1948 struct prefix
*pfx_target_original
) /* query target */
1950 struct rfapi_import_table
*it
;
1951 struct bgp
*bgp
= bgp_get_default();
1952 struct route_table
*rt
;
1953 struct route_node
*rn
;
1954 struct rfapi_next_hop_entry
*biglist
= NULL
;
1955 struct rfapi_next_hop_entry
*nhl
;
1956 struct rfapi_next_hop_entry
*tail
= NULL
;
1960 it
= rfapiMacImportTableGet(bgp
, logical_net_id
);
1961 rt
= it
->imported_vpn
[AFI_L2VPN
];
1963 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
1965 nhl
= rfapiEthRouteNode2NextHopList(
1966 rn
, rprefix
, lifetime
, exclude_vnaddr
, rfd_rib_table
,
1967 pfx_target_original
);
1969 tail
= biglist
= nhl
;
1976 while (tail
->next
) {
1983 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1988 * Insert a new bi to the imported route table node,
1989 * keeping the list of BIs sorted best route first
1991 static void rfapiBgpInfoAttachSorted(struct route_node
*rn
,
1992 struct bgp_info
*info_new
, afi_t afi
,
1996 struct bgp_info
*prev
;
1997 struct bgp_info
*next
;
1998 char pfx_buf
[PREFIX2STR_BUFFER
];
2001 bgp
= bgp_get_default(); /* assume 1 instance for now */
2003 if (VNC_DEBUG(IMPORT_BI_ATTACH
)) {
2004 vnc_zlog_debug_verbose("%s: info_new->peer=%p", __func__
,
2006 vnc_zlog_debug_verbose("%s: info_new->peer->su_remote=%p",
2007 __func__
, info_new
->peer
->su_remote
);
2010 for (prev
= NULL
, next
= rn
->info
; next
;
2011 prev
= next
, next
= next
->next
) {
2012 if (!bgp
|| (!CHECK_FLAG(info_new
->flags
, BGP_INFO_REMOVED
)
2013 && CHECK_FLAG(next
->flags
, BGP_INFO_REMOVED
))
2014 || bgp_info_cmp_compatible(bgp
, info_new
, next
, pfx_buf
,
2016 == -1) { /* -1 if 1st is better */
2020 vnc_zlog_debug_verbose("%s: prev=%p, next=%p", __func__
, prev
, next
);
2022 prev
->next
= info_new
;
2024 rn
->info
= info_new
;
2026 info_new
->prev
= prev
;
2027 info_new
->next
= next
;
2029 next
->prev
= info_new
;
2030 bgp_attr_intern(info_new
->attr
);
2033 static void rfapiBgpInfoDetach(struct route_node
*rn
, struct bgp_info
*bi
)
2036 * Remove the route (doubly-linked)
2038 // bgp_attr_unintern (&bi->attr);
2040 bi
->next
->prev
= bi
->prev
;
2042 bi
->prev
->next
= bi
->next
;
2044 rn
->info
= bi
->next
;
2048 * For L3-indexed import tables
2050 static int rfapi_bi_peer_rd_cmp(void *b1
, void *b2
)
2052 struct bgp_info
*bi1
= b1
;
2053 struct bgp_info
*bi2
= b2
;
2058 if (bi1
->peer
< bi2
->peer
)
2060 if (bi1
->peer
> bi2
->peer
)
2066 return vnc_prefix_cmp((struct prefix
*)&bi1
->extra
->vnc
.import
.rd
,
2067 (struct prefix
*)&bi2
->extra
->vnc
.import
.rd
);
2071 * For L2-indexed import tables
2072 * The BIs in these tables should ALWAYS have an aux_prefix set because
2073 * they arrive via IPv4 or IPv6 advertisements.
2075 static int rfapi_bi_peer_rd_aux_cmp(void *b1
, void *b2
)
2077 struct bgp_info
*bi1
= b1
;
2078 struct bgp_info
*bi2
= b2
;
2084 if (bi1
->peer
< bi2
->peer
)
2086 if (bi1
->peer
> bi2
->peer
)
2092 rc
= vnc_prefix_cmp((struct prefix
*)&bi1
->extra
->vnc
.import
.rd
,
2093 (struct prefix
*)&bi2
->extra
->vnc
.import
.rd
);
2099 * L2 import tables can have multiple entries with the
2100 * same MAC address, same RD, but different L3 addresses.
2102 * Use presence of aux_prefix with AF=ethernet and prefixlen=1
2103 * as magic value to signify explicit wildcarding of the aux_prefix.
2104 * This magic value will not appear in bona fide bi entries in
2105 * the import table, but is allowed in the "fake" bi used to
2106 * probe the table when searching. (We have to test both b1 and b2
2107 * because there is no guarantee of the order the test key and
2108 * the real key will be passed)
2110 if ((bi1
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2111 && (bi1
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))
2112 || (bi2
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2113 && (bi2
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))) {
2116 * wildcard aux address specified
2121 return vnc_prefix_cmp(&bi1
->extra
->vnc
.import
.aux_prefix
,
2122 &bi2
->extra
->vnc
.import
.aux_prefix
);
2127 * Index on RD and Peer
2129 static void rfapiItBiIndexAdd(struct route_node
*rn
, /* Import table VPN node */
2130 struct bgp_info
*bi
) /* new BI */
2132 struct skiplist
*sl
;
2139 char buf
[RD_ADDRSTRLEN
];
2141 vnc_zlog_debug_verbose("%s: bi %p, peer %p, rd %s", __func__
,
2143 prefix_rd2str(&bi
->extra
->vnc
.import
.rd
,
2147 sl
= RFAPI_RDINDEX_W_ALLOC(rn
);
2149 if (AF_ETHERNET
== rn
->p
.family
) {
2150 sl
= skiplist_new(0, rfapi_bi_peer_rd_aux_cmp
, NULL
);
2152 sl
= skiplist_new(0, rfapi_bi_peer_rd_cmp
, NULL
);
2154 RFAPI_IT_EXTRA_GET(rn
)->u
.vpn
.idx_rd
= sl
;
2155 route_lock_node(rn
); /* for skiplist */
2157 assert(!skiplist_insert(sl
, (void *)bi
, (void *)bi
));
2158 route_lock_node(rn
); /* for skiplist entry */
2160 /* NB: BIs in import tables are not refcounted */
2163 static void rfapiItBiIndexDump(struct route_node
*rn
)
2165 struct skiplist
*sl
;
2166 void *cursor
= NULL
;
2171 sl
= RFAPI_RDINDEX(rn
);
2175 for (rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
); !rc
;
2176 rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
)) {
2178 char buf
[RD_ADDRSTRLEN
];
2179 char buf_aux_pfx
[PREFIX_STRLEN
];
2181 prefix_rd2str(&k
->extra
->vnc
.import
.rd
, buf
, sizeof(buf
));
2182 if (k
->extra
->vnc
.import
.aux_prefix
.family
) {
2183 prefix2str(&k
->extra
->vnc
.import
.aux_prefix
,
2184 buf_aux_pfx
, sizeof(buf_aux_pfx
));
2186 strncpy(buf_aux_pfx
, "(none)", PREFIX_STRLEN
);
2188 vnc_zlog_debug_verbose("bi %p, peer %p, rd %s, aux_prefix %s",
2189 k
, k
->peer
, buf
, buf_aux_pfx
);
2193 static struct bgp_info
*rfapiItBiIndexSearch(
2194 struct route_node
*rn
, /* Import table VPN node */
2195 struct prefix_rd
*prd
, struct peer
*peer
,
2196 struct prefix
*aux_prefix
) /* optional L3 addr for L2 ITs */
2198 struct skiplist
*sl
;
2200 struct bgp_info bi_fake
;
2201 struct bgp_info_extra bi_extra
;
2202 struct bgp_info
*bi_result
;
2204 sl
= RFAPI_RDINDEX(rn
);
2210 char buf
[RD_ADDRSTRLEN
];
2211 char buf_aux_pfx
[PREFIX_STRLEN
];
2214 prefix2str(aux_prefix
, buf_aux_pfx
,
2215 sizeof(buf_aux_pfx
));
2217 strncpy(buf_aux_pfx
, "(nil)", sizeof(buf_aux_pfx
));
2219 vnc_zlog_debug_verbose("%s want prd=%s, peer=%p, aux_prefix=%s",
2221 prefix_rd2str(prd
, buf
, sizeof(buf
)),
2223 rfapiItBiIndexDump(rn
);
2227 /* threshold is a WAG */
2228 if (sl
->count
< 3) {
2230 vnc_zlog_debug_verbose("%s: short list algorithm", __func__
);
2232 /* if short list, linear search might be faster */
2233 for (bi_result
= rn
->info
; bi_result
;
2234 bi_result
= bi_result
->next
) {
2237 char buf
[RD_ADDRSTRLEN
];
2239 vnc_zlog_debug_verbose(
2240 "%s: bi has prd=%s, peer=%p", __func__
,
2241 prefix_rd2str(&bi_result
->extra
->vnc
2247 if (peer
== bi_result
->peer
2248 && !prefix_cmp((struct prefix
*)&bi_result
->extra
2250 (struct prefix
*)prd
)) {
2253 vnc_zlog_debug_verbose(
2254 "%s: peer and RD same, doing aux_prefix check",
2258 || !prefix_cmp(aux_prefix
,
2259 &bi_result
->extra
->vnc
.import
2263 vnc_zlog_debug_verbose("%s: match",
2273 bi_fake
.peer
= peer
;
2274 bi_fake
.extra
= &bi_extra
;
2275 bi_fake
.extra
->vnc
.import
.rd
= *(struct prefix_rd
*)prd
;
2277 bi_fake
.extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
2280 bi_fake
.extra
->vnc
.import
.aux_prefix
.family
= AF_ETHERNET
;
2281 bi_fake
.extra
->vnc
.import
.aux_prefix
.prefixlen
= 1;
2284 rc
= skiplist_search(sl
, (void *)&bi_fake
, (void *)&bi_result
);
2288 vnc_zlog_debug_verbose("%s: no match", __func__
);
2294 vnc_zlog_debug_verbose("%s: matched bi=%p", __func__
, bi_result
);
2300 static void rfapiItBiIndexDel(struct route_node
*rn
, /* Import table VPN node */
2301 struct bgp_info
*bi
) /* old BI */
2303 struct skiplist
*sl
;
2307 char buf
[RD_ADDRSTRLEN
];
2309 vnc_zlog_debug_verbose("%s: bi %p, peer %p, rd %s", __func__
,
2311 prefix_rd2str(&bi
->extra
->vnc
.import
.rd
,
2315 sl
= RFAPI_RDINDEX(rn
);
2318 rc
= skiplist_delete(sl
, (void *)(bi
), (void *)bi
);
2320 rfapiItBiIndexDump(rn
);
2324 route_unlock_node(rn
); /* for skiplist entry */
2326 /* NB: BIs in import tables are not refcounted */
2330 * Add a backreference at the ENCAP node to the VPN route that
2333 static void rfapiMonitorEncapAdd(struct rfapi_import_table
*import_table
,
2334 struct prefix
*p
, /* VN address */
2335 struct route_node
*vpn_rn
, /* VPN node */
2336 struct bgp_info
*vpn_bi
) /* VPN bi/route */
2338 afi_t afi
= family2afi(p
->family
);
2339 struct route_node
*rn
;
2340 struct rfapi_monitor_encap
*m
;
2343 rn
= route_node_get(import_table
->imported_encap
[afi
],
2347 m
= XCALLOC(MTYPE_RFAPI_MONITOR_ENCAP
,
2348 sizeof(struct rfapi_monitor_encap
));
2355 /* insert to encap node's list */
2356 m
->next
= RFAPI_MONITOR_ENCAP(rn
);
2359 RFAPI_MONITOR_ENCAP_W_ALLOC(rn
) = m
;
2361 /* for easy lookup when deleting vpn route */
2362 vpn_bi
->extra
->vnc
.import
.hme
= m
;
2364 vnc_zlog_debug_verbose(
2365 "%s: it=%p, vpn_bi=%p, afi=%d, encap rn=%p, setting vpn_bi->extra->vnc.import.hme=%p",
2366 __func__
, import_table
, vpn_bi
, afi
, rn
, m
);
2368 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
2369 bgp_attr_intern(vpn_bi
->attr
);
2372 static void rfapiMonitorEncapDelete(struct bgp_info
*vpn_bi
)
2375 * Remove encap monitor
2377 vnc_zlog_debug_verbose("%s: vpn_bi=%p", __func__
, vpn_bi
);
2378 if (vpn_bi
->extra
) {
2379 struct rfapi_monitor_encap
*hme
= vpn_bi
->extra
->vnc
.import
.hme
;
2383 vnc_zlog_debug_verbose("%s: hme=%p", __func__
, hme
);
2385 /* Refcount checking takes too long here */
2386 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 0);
2388 hme
->next
->prev
= hme
->prev
;
2390 hme
->prev
->next
= hme
->next
;
2392 RFAPI_MONITOR_ENCAP_W_ALLOC(hme
->rn
) =
2394 /* Refcount checking takes too long here */
2395 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 1);
2397 /* see if the struct rfapi_it_extra is empty and can be
2399 rfapiMonitorExtraPrune(SAFI_ENCAP
, hme
->rn
);
2401 route_unlock_node(hme
->rn
); /* decr ref count */
2402 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, hme
);
2403 vpn_bi
->extra
->vnc
.import
.hme
= NULL
;
2409 * quagga lib/thread.h says this must return int even though
2410 * it doesn't do anything with the return value
2412 static int rfapiWithdrawTimerVPN(struct thread
*t
)
2414 struct rfapi_withdraw
*wcb
= t
->arg
;
2415 struct bgp_info
*bi
= wcb
->info
;
2416 struct bgp
*bgp
= bgp_get_default();
2418 struct rfapi_monitor_vpn
*moved
;
2423 assert(wcb
->import_table
);
2426 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, wcb
->lockoffset
);
2431 vnc_zlog_debug_verbose(
2432 "%s: removing bi %p at prefix %s/%d", __func__
, bi
,
2433 rfapi_ntop(wcb
->node
->p
.family
, &wcb
->node
->p
.u
.prefix
,
2435 wcb
->node
->p
.prefixlen
);
2439 * Remove the route (doubly-linked)
2441 if (CHECK_FLAG(bi
->flags
, BGP_INFO_VALID
)
2442 && VALID_INTERIOR_TYPE(bi
->type
))
2443 RFAPI_MONITOR_EXTERIOR(wcb
->node
)->valid_interior_count
--;
2445 afi
= family2afi(wcb
->node
->p
.family
);
2446 wcb
->import_table
->holddown_count
[afi
] -= 1; /* keep count consistent */
2447 rfapiItBiIndexDel(wcb
->node
, bi
);
2448 rfapiBgpInfoDetach(wcb
->node
, bi
); /* with removed bi */
2450 vnc_import_bgp_exterior_del_route_interior(bgp
, wcb
->import_table
,
2455 * If VNC is configured to send response remove messages, AND
2456 * if the removed route had a UN address, do response removal
2459 if (!(bgp
->rfapi_cfg
->flags
2460 & BGP_VNC_CONFIG_RESPONSE_REMOVAL_DISABLE
)) {
2462 int has_valid_duplicate
= 0;
2463 struct bgp_info
*bii
;
2466 * First check if there are any OTHER routes at this node
2467 * that have the same nexthop and a valid UN address. If
2468 * there are (e.g., from other peers), then the route isn't
2469 * really gone, so skip sending a response removal message.
2471 for (bii
= wcb
->node
->info
; bii
; bii
= bii
->next
) {
2472 if (rfapiVpnBiSamePtUn(bi
, bii
)) {
2473 has_valid_duplicate
= 1;
2478 vnc_zlog_debug_verbose("%s: has_valid_duplicate=%d", __func__
,
2479 has_valid_duplicate
);
2481 if (!has_valid_duplicate
) {
2482 rfapiRibPendingDeleteRoute(bgp
, wcb
->import_table
, afi
,
2487 rfapiMonitorEncapDelete(bi
);
2490 * If there are no VPN monitors at this VPN Node A,
2493 if (!RFAPI_MONITOR_VPN(wcb
->node
)) {
2494 vnc_zlog_debug_verbose("%s: no VPN monitors at this node",
2500 * rfapiMonitorMoveShorter only moves monitors if there are
2501 * no remaining valid routes at the current node
2503 moved
= rfapiMonitorMoveShorter(wcb
->node
, 1);
2506 rfapiMonitorMovedUp(wcb
->import_table
, wcb
->node
, moved
->node
,
2514 rfapiBgpInfoFree(bi
);
2518 * If route count at this node has gone to 0, withdraw exported prefix
2520 if (!wcb
->node
->info
) {
2521 /* see if the struct rfapi_it_extra is empty and can be freed */
2522 rfapiMonitorExtraPrune(SAFI_MPLS_VPN
, wcb
->node
);
2523 vnc_direct_bgp_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2524 vnc_zebra_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2527 * nexthop change event
2528 * vnc_direct_bgp_add_prefix() will recompute the VN addr
2531 vnc_direct_bgp_add_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2534 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, 1 + wcb
->lockoffset
);
2535 route_unlock_node(wcb
->node
); /* decr ref count */
2536 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2541 * This works for multiprotocol extension, but not for plain ol'
2542 * unicast IPv4 because that nexthop is stored in attr->nexthop
2544 void rfapiNexthop2Prefix(struct attr
*attr
, struct prefix
*p
)
2549 memset(p
, 0, sizeof(struct prefix
));
2551 switch (p
->family
= BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2553 p
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2558 p
->u
.prefix6
= attr
->mp_nexthop_global
;
2563 vnc_zlog_debug_verbose("%s: Family is unknown = %d", __func__
,
2568 void rfapiUnicastNexthop2Prefix(afi_t afi
, struct attr
*attr
, struct prefix
*p
)
2570 if (afi
== AFI_IP
) {
2571 p
->family
= AF_INET
;
2573 p
->u
.prefix4
= attr
->nexthop
;
2575 rfapiNexthop2Prefix(attr
, p
);
2579 static int rfapiAttrNexthopAddrDifferent(struct prefix
*p1
, struct prefix
*p2
)
2582 vnc_zlog_debug_verbose("%s: p1 or p2 is NULL", __func__
);
2587 * Are address families the same?
2589 if (p1
->family
!= p2
->family
) {
2593 switch (p1
->family
) {
2595 if (IPV4_ADDR_SAME(&p1
->u
.prefix4
, &p2
->u
.prefix4
))
2600 if (IPV6_ADDR_SAME(&p1
->u
.prefix6
, &p2
->u
.prefix6
))
2611 static void rfapiCopyUnEncap2VPN(struct bgp_info
*encap_bi
,
2612 struct bgp_info
*vpn_bi
)
2614 if (!encap_bi
->attr
) {
2615 zlog_warn("%s: no encap bi attr/extra, can't copy UN address",
2620 if (!vpn_bi
|| !vpn_bi
->extra
) {
2621 zlog_warn("%s: no vpn bi attr/extra, can't copy UN address",
2626 switch (BGP_MP_NEXTHOP_FAMILY(encap_bi
->attr
->mp_nexthop_len
)) {
2630 * instrumentation to debug segfault of 091127
2632 vnc_zlog_debug_verbose("%s: vpn_bi=%p", __func__
, vpn_bi
);
2634 vnc_zlog_debug_verbose("%s: vpn_bi->extra=%p", __func__
,
2638 vpn_bi
->extra
->vnc
.import
.un_family
= AF_INET
;
2639 vpn_bi
->extra
->vnc
.import
.un
.addr4
=
2640 encap_bi
->attr
->mp_nexthop_global_in
;
2644 vpn_bi
->extra
->vnc
.import
.un_family
= AF_INET6
;
2645 vpn_bi
->extra
->vnc
.import
.un
.addr6
=
2646 encap_bi
->attr
->mp_nexthop_global
;
2650 zlog_warn("%s: invalid encap nexthop length: %d", __func__
,
2651 encap_bi
->attr
->mp_nexthop_len
);
2652 vpn_bi
->extra
->vnc
.import
.un_family
= 0;
2658 * returns 0 on success, nonzero on error
2660 static int rfapiWithdrawEncapUpdateCachedUn(
2661 struct rfapi_import_table
*import_table
, struct bgp_info
*encap_bi
,
2662 struct route_node
*vpn_rn
, struct bgp_info
*vpn_bi
)
2667 * clear cached UN address
2669 if (!vpn_bi
|| !vpn_bi
->extra
) {
2671 "%s: missing VPN bi/extra, can't clear UN addr",
2675 vpn_bi
->extra
->vnc
.import
.un_family
= 0;
2676 memset(&vpn_bi
->extra
->vnc
.import
.un
, 0,
2677 sizeof(vpn_bi
->extra
->vnc
.import
.un
));
2678 if (CHECK_FLAG(vpn_bi
->flags
, BGP_INFO_VALID
)) {
2679 if (rfapiGetVncTunnelUnAddr(vpn_bi
->attr
, NULL
)) {
2680 UNSET_FLAG(vpn_bi
->flags
, BGP_INFO_VALID
);
2681 if (VALID_INTERIOR_TYPE(vpn_bi
->type
))
2682 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2683 ->valid_interior_count
--;
2684 /* signal interior route withdrawal to
2685 * import-exterior */
2686 vnc_import_bgp_exterior_del_route_interior(
2687 bgp_get_default(), import_table
, vpn_rn
,
2694 zlog_warn("%s: missing VPN bi, can't clear UN addr",
2698 rfapiCopyUnEncap2VPN(encap_bi
, vpn_bi
);
2699 if (!CHECK_FLAG(vpn_bi
->flags
, BGP_INFO_VALID
)) {
2700 SET_FLAG(vpn_bi
->flags
, BGP_INFO_VALID
);
2701 if (VALID_INTERIOR_TYPE(vpn_bi
->type
))
2702 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2703 ->valid_interior_count
++;
2704 /* signal interior route withdrawal to import-exterior
2706 vnc_import_bgp_exterior_add_route_interior(
2707 bgp_get_default(), import_table
, vpn_rn
,
2714 static int rfapiWithdrawTimerEncap(struct thread
*t
)
2716 struct rfapi_withdraw
*wcb
= t
->arg
;
2717 struct bgp_info
*bi
= wcb
->info
;
2718 int was_first_route
= 0;
2719 struct rfapi_monitor_encap
*em
;
2720 struct skiplist
*vpn_node_sl
= skiplist_new(0, NULL
, NULL
);
2724 assert(wcb
->import_table
);
2726 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 0);
2728 if (wcb
->node
->info
== bi
)
2729 was_first_route
= 1;
2732 * Remove the route/bi and free it
2734 rfapiBgpInfoDetach(wcb
->node
, bi
);
2735 rfapiBgpInfoFree(bi
);
2737 if (!was_first_route
)
2740 for (em
= RFAPI_MONITOR_ENCAP(wcb
->node
); em
; em
= em
->next
) {
2743 * Update monitoring VPN BIs with new encap info at the
2744 * head of the encap bi chain (which could be NULL after
2745 * removing the expiring bi above)
2747 if (rfapiWithdrawEncapUpdateCachedUn(wcb
->import_table
,
2748 wcb
->node
->info
, em
->node
,
2753 * Build a list of unique VPN nodes referenced by these
2755 * Use a skiplist for speed.
2757 skiplist_insert(vpn_node_sl
, em
->node
, em
->node
);
2762 * for each VPN node referenced in the ENCAP monitors:
2764 struct route_node
*rn
;
2765 while (!skiplist_first(vpn_node_sl
, (void **)&rn
, NULL
)) {
2766 if (!wcb
->node
->info
) {
2767 struct rfapi_monitor_vpn
*moved
;
2769 moved
= rfapiMonitorMoveShorter(rn
, 0);
2771 // rfapiDoRouteCallback(wcb->import_table,
2772 // moved->node, moved);
2773 rfapiMonitorMovedUp(wcb
->import_table
, rn
,
2774 moved
->node
, moved
);
2777 // rfapiDoRouteCallback(wcb->import_table, rn, NULL);
2778 rfapiMonitorItNodeChanged(wcb
->import_table
, rn
, NULL
);
2780 skiplist_delete_first(vpn_node_sl
);
2784 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 1);
2785 route_unlock_node(wcb
->node
); /* decr ref count */
2786 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2787 skiplist_free(vpn_node_sl
);
2793 * Works for both VPN and ENCAP routes; timer_service_func is different
2797 rfapiBiStartWithdrawTimer(struct rfapi_import_table
*import_table
,
2798 struct route_node
*rn
, struct bgp_info
*bi
, afi_t afi
,
2800 int (*timer_service_func
)(struct thread
*))
2803 struct rfapi_withdraw
*wcb
;
2805 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)) {
2807 * Already on the path to being withdrawn,
2808 * should already have a timer set up to
2811 vnc_zlog_debug_verbose(
2812 "%s: already being withdrawn, do nothing", __func__
);
2816 rfapiGetVncLifetime(bi
->attr
, &lifetime
);
2817 vnc_zlog_debug_verbose("%s: VNC lifetime is %u", __func__
, lifetime
);
2820 * withdrawn routes get to hang around for a while
2822 SET_FLAG(bi
->flags
, BGP_INFO_REMOVED
);
2824 /* set timer to remove the route later */
2825 lifetime
= rfapiGetHolddownFromLifetime(lifetime
);
2826 vnc_zlog_debug_verbose("%s: using timeout %u", __func__
, lifetime
);
2829 * Stash import_table, node, and info for use by timer
2830 * service routine, which is supposed to free the wcb.
2832 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2836 wcb
->import_table
= import_table
;
2837 bgp_attr_intern(bi
->attr
);
2839 if (VNC_DEBUG(VERBOSE
)) {
2840 vnc_zlog_debug_verbose(
2841 "%s: wcb values: node=%p, info=%p, import_table=%p (bi follows)",
2842 __func__
, wcb
->node
, wcb
->info
, wcb
->import_table
);
2843 rfapiPrintBi(NULL
, bi
);
2848 if (lifetime
> UINT32_MAX
/ 1001) {
2849 /* sub-optimal case, but will probably never happen */
2850 bi
->extra
->vnc
.import
.timer
= NULL
;
2851 thread_add_timer(bm
->master
, timer_service_func
, wcb
, lifetime
,
2852 &bi
->extra
->vnc
.import
.timer
);
2854 static uint32_t jitter
;
2855 uint32_t lifetime_msec
;
2858 * the goal here is to spread out the timers so they are
2859 * sortable in the skip list
2861 if (++jitter
>= 1000)
2864 lifetime_msec
= (lifetime
* 1000) + jitter
;
2866 bi
->extra
->vnc
.import
.timer
= NULL
;
2867 thread_add_timer_msec(bm
->master
, timer_service_func
, wcb
,
2869 &bi
->extra
->vnc
.import
.timer
);
2872 /* re-sort route list (BGP_INFO_REMOVED routes are last) */
2873 if (((struct bgp_info
*)rn
->info
)->next
) {
2874 rfapiBgpInfoDetach(rn
, bi
);
2875 rfapiBgpInfoAttachSorted(rn
, bi
, afi
, safi
);
2880 typedef void(rfapi_bi_filtered_import_f
)(struct rfapi_import_table
*, int,
2881 struct peer
*, void *, struct prefix
*,
2882 struct prefix
*, afi_t
,
2883 struct prefix_rd
*, struct attr
*,
2884 uint8_t, uint8_t, uint32_t *);
2887 static void rfapiExpireEncapNow(struct rfapi_import_table
*it
,
2888 struct route_node
*rn
, struct bgp_info
*bi
)
2890 struct rfapi_withdraw
*wcb
;
2894 * pretend we're an expiring timer
2896 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2899 wcb
->import_table
= it
;
2900 memset(&t
, 0, sizeof(t
));
2902 rfapiWithdrawTimerEncap(&t
); /* frees wcb */
2905 static int rfapiGetNexthop(struct attr
*attr
, struct prefix
*prefix
)
2907 switch (BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2909 prefix
->family
= AF_INET
;
2910 prefix
->prefixlen
= 32;
2911 prefix
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2914 prefix
->family
= AF_INET6
;
2915 prefix
->prefixlen
= 128;
2916 prefix
->u
.prefix6
= attr
->mp_nexthop_global
;
2919 vnc_zlog_debug_verbose("%s: unknown attr->mp_nexthop_len %d",
2920 __func__
, attr
->mp_nexthop_len
);
2927 * import a bgp_info if its route target list intersects with the
2928 * import table's route target list
2930 static void rfapiBgpInfoFilteredImportEncap(
2931 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
2932 void *rfd
, /* set for looped back routes */
2934 struct prefix
*aux_prefix
, /* Unused for encap routes */
2935 afi_t afi
, struct prefix_rd
*prd
,
2936 struct attr
*attr
, /* part of bgp_info */
2937 uint8_t type
, /* part of bgp_info */
2938 uint8_t sub_type
, /* part of bgp_info */
2939 uint32_t *label
) /* part of bgp_info */
2941 struct route_table
*rt
= NULL
;
2942 struct route_node
*rn
;
2943 struct bgp_info
*info_new
;
2944 struct bgp_info
*bi
;
2945 struct bgp_info
*next
;
2948 struct prefix p_firstbi_old
;
2949 struct prefix p_firstbi_new
;
2951 const char *action_str
= NULL
;
2952 struct prefix un_prefix
;
2955 bgp
= bgp_get_default(); /* assume 1 instance for now */
2958 case FIF_ACTION_UPDATE
:
2959 action_str
= "update";
2961 case FIF_ACTION_WITHDRAW
:
2962 action_str
= "withdraw";
2964 case FIF_ACTION_KILL
:
2965 action_str
= "kill";
2972 vnc_zlog_debug_verbose(
2973 "%s: entry: %s: prefix %s/%d", __func__
, action_str
,
2974 inet_ntop(p
->family
, &p
->u
.prefix
, buf
, BUFSIZ
), p
->prefixlen
);
2976 memset(&p_firstbi_old
, 0, sizeof(p_firstbi_old
));
2977 memset(&p_firstbi_new
, 0, sizeof(p_firstbi_new
));
2979 if (action
== FIF_ACTION_UPDATE
) {
2981 * Compare rt lists. If no intersection, don't import this route
2982 * On a withdraw, peer and RD are sufficient to determine if
2985 if (!attr
|| !attr
->ecommunity
) {
2987 vnc_zlog_debug_verbose(
2988 "%s: attr, extra, or ecommunity missing, not importing",
2992 #if RFAPI_REQUIRE_ENCAP_BEEC
2993 if (!rfapiEcommunitiesMatchBeec(attr
->ecommunity
)) {
2994 vnc_zlog_debug_verbose(
2995 "%s: it=%p: no match for BGP Encapsulation ecommunity",
2996 __func__
, import_table
);
3000 if (!rfapiEcommunitiesIntersect(import_table
->rt_import_list
,
3001 attr
->ecommunity
)) {
3003 vnc_zlog_debug_verbose(
3004 "%s: it=%p: no ecommunity intersection",
3005 __func__
, import_table
);
3010 * Updates must also have a nexthop address
3012 memset(&un_prefix
, 0,
3013 sizeof(un_prefix
)); /* keep valgrind happy */
3014 if (rfapiGetNexthop(attr
, &un_prefix
)) {
3015 vnc_zlog_debug_verbose("%s: missing nexthop address",
3022 * Figure out which radix tree the route would go into
3027 rt
= import_table
->imported_encap
[afi
];
3031 zlog_err("%s: bad afi %d", __func__
, afi
);
3036 * route_node_lookup returns a node only if there is at least
3037 * one route attached.
3039 rn
= route_node_lookup(rt
, p
);
3041 #if DEBUG_ENCAP_MONITOR
3042 vnc_zlog_debug_verbose("%s: initial encap lookup(it=%p) rn=%p",
3043 __func__
, import_table
, rn
);
3048 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 1);
3049 route_unlock_node(rn
); /* undo lock in route_node_lookup */
3053 * capture nexthop of first bi
3056 rfapiNexthop2Prefix(
3057 ((struct bgp_info
*)(rn
->info
))->attr
,
3061 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
3064 * Does this bgp_info refer to the same route
3065 * as we are trying to add?
3067 vnc_zlog_debug_verbose("%s: comparing BI %p", __func__
,
3074 * RD of import table bi is in bi->extra->vnc.import.rd
3075 * RD of info_orig is in prd
3078 vnc_zlog_debug_verbose("%s: no bi->extra",
3083 (struct prefix
*)&bi
->extra
->vnc
.import
.rd
,
3084 (struct prefix
*)prd
)) {
3086 vnc_zlog_debug_verbose("%s: prd does not match",
3094 if (bi
->peer
!= peer
) {
3095 vnc_zlog_debug_verbose(
3096 "%s: peer does not match", __func__
);
3100 vnc_zlog_debug_verbose("%s: found matching bi",
3103 /* Same route. Delete this bi, replace with new one */
3105 if (action
== FIF_ACTION_WITHDRAW
) {
3107 vnc_zlog_debug_verbose(
3108 "%s: withdrawing at prefix %s/%d",
3110 inet_ntop(rn
->p
.family
, &rn
->p
.u
.prefix
,
3114 rfapiBiStartWithdrawTimer(
3115 import_table
, rn
, bi
, afi
, SAFI_ENCAP
,
3116 rfapiWithdrawTimerEncap
);
3119 vnc_zlog_debug_verbose(
3120 "%s: %s at prefix %s/%d", __func__
,
3121 ((action
== FIF_ACTION_KILL
)
3124 inet_ntop(rn
->p
.family
, &rn
->p
.u
.prefix
,
3129 * If this route is waiting to be deleted
3131 * a previous withdraw, we must cancel its
3134 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
3135 && bi
->extra
->vnc
.import
.timer
) {
3138 (struct thread
*)bi
->extra
->vnc
3140 struct rfapi_withdraw
*wcb
= t
->arg
;
3142 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3146 if (action
== FIF_ACTION_UPDATE
) {
3147 rfapiBgpInfoDetach(rn
, bi
);
3148 rfapiBgpInfoFree(bi
);
3152 * Kill: do export stuff when removing
3155 struct rfapi_withdraw
*wcb
;
3159 * pretend we're an expiring timer
3162 MTYPE_RFAPI_WITHDRAW
,
3163 sizeof(struct rfapi_withdraw
));
3166 wcb
->import_table
= import_table
;
3167 memset(&t
, 0, sizeof(t
));
3169 rfapiWithdrawTimerEncap(
3170 &t
); /* frees wcb */
3179 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, replacing
? 1 : 0);
3181 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
)
3185 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, NULL
);
3189 route_lock_node(rn
); /* incr ref count for new BI */
3191 rn
= route_node_get(rt
, p
);
3194 vnc_zlog_debug_verbose(
3195 "%s: (afi=%d, rn=%p) inserting at prefix %s/%d", __func__
, afi
,
3196 rn
, inet_ntop(rn
->p
.family
, &rn
->p
.u
.prefix
, buf
, BUFSIZ
),
3199 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_ENCAP
);
3202 * Delete holddown routes from same NVE. See details in
3203 * rfapiBgpInfoFilteredImportVPN()
3205 for (bi
= info_new
->next
; bi
; bi
= next
) {
3207 struct prefix pfx_un
;
3211 if (!CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
))
3215 * We already match the VN address (it is the prefix
3216 * of the route node)
3219 if (!rfapiGetNexthop(bi
->attr
, &pfx_un
)
3220 && prefix_same(&pfx_un
, &un_prefix
)) {
3228 vnc_zlog_debug_verbose(
3229 "%s: removing holddown bi matching NVE of new route",
3231 if (bi
->extra
->vnc
.import
.timer
) {
3233 (struct thread
*)bi
->extra
->vnc
.import
.timer
;
3234 struct rfapi_withdraw
*wcb
= t
->arg
;
3236 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3239 rfapiExpireEncapNow(import_table
, rn
, bi
);
3242 rfapiNexthop2Prefix(((struct bgp_info
*)(rn
->info
))->attr
,
3246 * If the nexthop address of the selected Encap route (i.e.,
3247 * the UN address) has changed, then we must update the VPN
3248 * routes that refer to this Encap route and possibly force
3251 if (rfapiAttrNexthopAddrDifferent(&p_firstbi_old
, &p_firstbi_new
)) {
3253 struct rfapi_monitor_encap
*m
;
3254 struct rfapi_monitor_encap
*mnext
;
3256 struct route_node
*referenced_vpn_prefix
;
3259 * Optimized approach: build radix tree on the fly to
3260 * hold list of VPN nodes referenced by the ENCAP monitors
3262 * The nodes in this table correspond to prefixes of VPN routes.
3263 * The "info" pointer of the node points to a chain of
3264 * struct rfapi_monitor_encap, each of which refers to a
3265 * specific VPN node.
3267 struct route_table
*referenced_vpn_table
;
3269 referenced_vpn_table
= route_table_init();
3270 assert(referenced_vpn_table
);
3273 * iterate over the set of monitors at this ENCAP node.
3275 #if DEBUG_ENCAP_MONITOR
3276 vnc_zlog_debug_verbose("%s: examining monitors at rn=%p",
3279 for (m
= RFAPI_MONITOR_ENCAP(rn
); m
; m
= m
->next
) {
3282 * For each referenced bi/route, copy the ENCAP route's
3283 * nexthop to the VPN route's cached UN address field
3285 * the address family of the cached UN address field.
3287 rfapiCopyUnEncap2VPN(info_new
, m
->bi
);
3288 if (!CHECK_FLAG(m
->bi
->flags
, BGP_INFO_VALID
)) {
3289 SET_FLAG(m
->bi
->flags
, BGP_INFO_VALID
);
3290 if (VALID_INTERIOR_TYPE(m
->bi
->type
))
3291 RFAPI_MONITOR_EXTERIOR(m
->node
)
3292 ->valid_interior_count
++;
3293 vnc_import_bgp_exterior_add_route_interior(
3294 bgp
, import_table
, m
->node
, m
->bi
);
3298 * Build a list of unique VPN nodes referenced by these
3301 * There could be more than one VPN node here with a
3303 * prefix. Those are currently in an unsorted linear
3308 referenced_vpn_prefix
= route_node_get(
3309 referenced_vpn_table
, &m
->node
->p
);
3310 assert(referenced_vpn_prefix
);
3311 for (mnext
= referenced_vpn_prefix
->info
; mnext
;
3312 mnext
= mnext
->next
) {
3314 if (mnext
->node
== m
->node
)
3320 * already have an entry for this VPN node
3322 route_unlock_node(referenced_vpn_prefix
);
3325 MTYPE_RFAPI_MONITOR_ENCAP
,
3326 sizeof(struct rfapi_monitor_encap
));
3328 mnext
->node
= m
->node
;
3329 mnext
->next
= referenced_vpn_prefix
->info
;
3330 referenced_vpn_prefix
->info
= mnext
;
3335 * for each VPN node referenced in the ENCAP monitors:
3337 for (referenced_vpn_prefix
= route_top(referenced_vpn_table
);
3338 referenced_vpn_prefix
; referenced_vpn_prefix
= route_next(
3339 referenced_vpn_prefix
)) {
3341 while ((m
= referenced_vpn_prefix
->info
)) {
3343 struct route_node
*n
;
3345 rfapiMonitorMoveLonger(m
->node
);
3346 for (n
= m
->node
; n
; n
= n
->parent
) {
3347 // rfapiDoRouteCallback(import_table, n,
3350 rfapiMonitorItNodeChanged(import_table
, m
->node
,
3353 referenced_vpn_prefix
->info
= m
->next
;
3354 route_unlock_node(referenced_vpn_prefix
);
3355 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, m
);
3358 route_table_finish(referenced_vpn_table
);
3361 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
3364 static void rfapiExpireVpnNow(struct rfapi_import_table
*it
,
3365 struct route_node
*rn
, struct bgp_info
*bi
,
3368 struct rfapi_withdraw
*wcb
;
3372 * pretend we're an expiring timer
3374 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
3377 wcb
->import_table
= it
;
3378 wcb
->lockoffset
= lockoffset
;
3379 memset(&t
, 0, sizeof(t
));
3381 rfapiWithdrawTimerVPN(&t
); /* frees wcb */
3386 * import a bgp_info if its route target list intersects with the
3387 * import table's route target list
3389 void rfapiBgpInfoFilteredImportVPN(
3390 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3391 void *rfd
, /* set for looped back routes */
3393 struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3394 afi_t afi
, struct prefix_rd
*prd
,
3395 struct attr
*attr
, /* part of bgp_info */
3396 uint8_t type
, /* part of bgp_info */
3397 uint8_t sub_type
, /* part of bgp_info */
3398 uint32_t *label
) /* part of bgp_info */
3400 struct route_table
*rt
= NULL
;
3401 struct route_node
*rn
;
3402 struct route_node
*n
;
3403 struct bgp_info
*info_new
;
3404 struct bgp_info
*bi
;
3405 struct bgp_info
*next
;
3407 struct prefix vn_prefix
;
3408 struct prefix un_prefix
;
3409 int un_prefix_valid
= 0;
3410 struct route_node
*ern
;
3412 int original_had_routes
= 0;
3413 struct prefix original_nexthop
;
3414 const char *action_str
= NULL
;
3418 bgp
= bgp_get_default(); /* assume 1 instance for now */
3421 case FIF_ACTION_UPDATE
:
3422 action_str
= "update";
3424 case FIF_ACTION_WITHDRAW
:
3425 action_str
= "withdraw";
3427 case FIF_ACTION_KILL
:
3428 action_str
= "kill";
3435 if (import_table
== bgp
->rfapi
->it_ce
)
3438 vnc_zlog_debug_verbose("%s: entry: %s%s: prefix %s/%d: it %p, afi %s",
3439 __func__
, (is_it_ce
? "CE-IT " : ""), action_str
,
3440 rfapi_ntop(p
->family
, &p
->u
.prefix
, buf
, BUFSIZ
),
3441 p
->prefixlen
, import_table
, afi2str(afi
));
3446 * Compare rt lists. If no intersection, don't import this route
3447 * On a withdraw, peer and RD are sufficient to determine if
3450 if (action
== FIF_ACTION_UPDATE
) {
3451 if (!attr
|| !attr
->ecommunity
) {
3453 vnc_zlog_debug_verbose(
3454 "%s: attr, extra, or ecommunity missing, not importing",
3458 if ((import_table
!= bgp
->rfapi
->it_ce
)
3459 && !rfapiEcommunitiesIntersect(import_table
->rt_import_list
,
3460 attr
->ecommunity
)) {
3462 vnc_zlog_debug_verbose(
3463 "%s: it=%p: no ecommunity intersection",
3464 __func__
, import_table
);
3468 memset(&vn_prefix
, 0,
3469 sizeof(vn_prefix
)); /* keep valgrind happy */
3470 if (rfapiGetNexthop(attr
, &vn_prefix
)) {
3471 /* missing nexthop address would be a bad, bad thing */
3472 vnc_zlog_debug_verbose("%s: missing nexthop", __func__
);
3478 * Figure out which radix tree the route would go into
3484 rt
= import_table
->imported_vpn
[afi
];
3488 zlog_err("%s: bad afi %d", __func__
, afi
);
3493 memset(&original_nexthop
, 0, sizeof(original_nexthop
));
3496 * route_node_lookup returns a node only if there is at least
3497 * one route attached.
3499 rn
= route_node_lookup(rt
, p
);
3501 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
3505 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
3506 route_unlock_node(rn
); /* undo lock in route_node_lookup */
3509 original_had_routes
= 1;
3511 if (VNC_DEBUG(VERBOSE
)) {
3512 vnc_zlog_debug_verbose("%s: showing IT node on entry",
3514 rfapiShowItNode(NULL
, rn
); /* debug */
3518 * Look for same route (will have same RD and peer)
3520 bi
= rfapiItBiIndexSearch(rn
, prd
, peer
, aux_prefix
);
3525 * This was an old test when we iterated over the
3526 * BIs linearly. Since we're now looking up with
3527 * RD and peer, comparing types should not be
3528 * needed. Changed to assertion.
3530 * Compare types. Doing so prevents a RFP-originated
3531 * route from matching an imported route, for example.
3533 if (VNC_DEBUG(VERBOSE
) && bi
->type
!= type
)
3534 /* should be handled by RDs, but warn for now */
3535 zlog_warn("%s: type mismatch! (bi=%d, arg=%d)",
3536 __func__
, bi
->type
, type
);
3538 vnc_zlog_debug_verbose("%s: found matching bi",
3542 * In the special CE table, withdrawals occur without
3545 if (import_table
== bgp
->rfapi
->it_ce
) {
3546 vnc_direct_bgp_del_route_ce(bgp
, rn
, bi
);
3547 if (action
== FIF_ACTION_WITHDRAW
)
3548 action
= FIF_ACTION_KILL
;
3551 if (action
== FIF_ACTION_WITHDRAW
) {
3554 CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
);
3556 vnc_zlog_debug_verbose(
3557 "%s: withdrawing at prefix %s/%d%s",
3558 __func__
, rfapi_ntop(rn
->p
.family
,
3563 ? " (already being withdrawn)"
3568 rfapiBiStartWithdrawTimer(
3569 import_table
, rn
, bi
, afi
,
3571 rfapiWithdrawTimerVPN
);
3573 RFAPI_UPDATE_ITABLE_COUNT(
3574 bi
, import_table
, afi
, -1);
3575 import_table
->holddown_count
[afi
] += 1;
3579 vnc_zlog_debug_verbose(
3580 "%s: %s at prefix %s/%d", __func__
,
3581 ((action
== FIF_ACTION_KILL
)
3584 rfapi_ntop(rn
->p
.family
,
3585 &rn
->p
.u
.prefix
, buf
,
3590 * If this route is waiting to be deleted
3592 * a previous withdraw, we must cancel its
3595 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
3596 && bi
->extra
->vnc
.import
.timer
) {
3599 (struct thread
*)bi
->extra
->vnc
3601 struct rfapi_withdraw
*wcb
= t
->arg
;
3603 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3606 import_table
->holddown_count
[afi
] -= 1;
3607 RFAPI_UPDATE_ITABLE_COUNT(
3608 bi
, import_table
, afi
, 1);
3611 * decrement remote count (if route is remote)
3613 * we are going to remove it below
3615 RFAPI_UPDATE_ITABLE_COUNT(bi
, import_table
, afi
,
3617 if (action
== FIF_ACTION_UPDATE
) {
3621 * make copy of original nexthop so we
3622 * can see if it changed
3624 rfapiGetNexthop(bi
->attr
,
3628 * remove bi without doing any export
3631 if (CHECK_FLAG(bi
->flags
,
3633 && VALID_INTERIOR_TYPE(bi
->type
))
3634 RFAPI_MONITOR_EXTERIOR(rn
)
3635 ->valid_interior_count
--;
3636 rfapiItBiIndexDel(rn
, bi
);
3637 rfapiBgpInfoDetach(rn
, bi
);
3638 rfapiMonitorEncapDelete(bi
);
3639 vnc_import_bgp_exterior_del_route_interior(
3640 bgp
, import_table
, rn
, bi
);
3641 rfapiBgpInfoFree(bi
);
3645 * remove bi and do export processing
3647 import_table
->holddown_count
[afi
] += 1;
3648 rfapiExpireVpnNow(import_table
, rn
, bi
,
3656 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, replacing
? 1 : 0);
3658 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
) {
3664 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, label
);
3667 * lookup un address in encap table
3669 ern
= route_node_match(import_table
->imported_encap
[afi
], &vn_prefix
);
3671 rfapiCopyUnEncap2VPN(ern
->info
, info_new
);
3672 route_unlock_node(ern
); /* undo lock in route_note_match */
3674 char buf
[PREFIX_STRLEN
];
3676 prefix2str(&vn_prefix
, buf
, sizeof(buf
));
3677 /* Not a big deal, just means VPN route got here first */
3678 vnc_zlog_debug_verbose("%s: no encap route for vn addr %s",
3680 info_new
->extra
->vnc
.import
.un_family
= 0;
3685 route_lock_node(rn
);
3688 * No need to increment reference count, so only "get"
3689 * if the node is not there already
3691 rn
= route_node_get(rt
, p
);
3695 * For ethernet routes, if there is an accompanying IP address,
3698 if ((AFI_L2VPN
== afi
) && aux_prefix
) {
3700 vnc_zlog_debug_verbose("%s: setting BI's aux_prefix", __func__
);
3701 info_new
->extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
3704 vnc_zlog_debug_verbose(
3705 "%s: inserting bi %p at prefix %s/%d #%d", __func__
, info_new
,
3706 rfapi_ntop(rn
->p
.family
, &rn
->p
.u
.prefix
, buf
, BUFSIZ
),
3707 rn
->p
.prefixlen
, rn
->lock
);
3709 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_MPLS_VPN
);
3710 rfapiItBiIndexAdd(rn
, info_new
);
3711 if (!rfapiGetUnAddrOfVpnBi(info_new
, NULL
)) {
3712 if (VALID_INTERIOR_TYPE(info_new
->type
))
3713 RFAPI_MONITOR_EXTERIOR(rn
)->valid_interior_count
++;
3714 SET_FLAG(info_new
->flags
, BGP_INFO_VALID
);
3716 RFAPI_UPDATE_ITABLE_COUNT(info_new
, import_table
, afi
, 1);
3717 vnc_import_bgp_exterior_add_route_interior(bgp
, import_table
, rn
,
3720 if (import_table
== bgp
->rfapi
->it_ce
)
3721 vnc_direct_bgp_add_route_ce(bgp
, rn
, info_new
);
3723 if (VNC_DEBUG(VERBOSE
)) {
3724 vnc_zlog_debug_verbose("%s: showing IT node", __func__
);
3725 rfapiShowItNode(NULL
, rn
); /* debug */
3728 rfapiMonitorEncapAdd(import_table
, &vn_prefix
, rn
, info_new
);
3730 if (!rfapiGetUnAddrOfVpnBi(info_new
, &un_prefix
)) {
3733 * if we have a valid UN address (either via Encap route
3734 * or via tunnel attribute), then we should attempt
3735 * to move any monitors at less-specific nodes to this node
3737 rfapiMonitorMoveLonger(rn
);
3739 un_prefix_valid
= 1;
3743 * 101129 Enhancement: if we add a route (implication: it is not
3744 * in holddown), delete all other routes from this nve at this
3745 * node that are in holddown, regardless of peer.
3747 * Reasons it's OK to do that:
3749 * - if the holddown route being deleted originally came from BGP VPN,
3750 * it is already gone from BGP (implication of holddown), so there
3751 * won't be any added inconsistency with the BGP RIB.
3753 * - once a fresh route is added at a prefix, any routes in holddown
3754 * at that prefix will not show up in RFP responses, so deleting
3755 * the holddown routes won't affect the contents of responses.
3757 * - lifetimes are supposed to be consistent, so there should not
3758 * be a case where the fresh route has a shorter lifetime than
3759 * the holddown route, so we don't expect the fresh route to
3760 * disappear and complete its holddown time before the existing
3761 * holddown routes time out. Therefore, we won't have a situation
3762 * where we expect the existing holddown routes to be hidden and
3763 * then to reappear sometime later (as holddown routes) in a
3766 * Among other things, this would enable us to skirt the problem
3767 * of local holddown routes that refer to NVE descriptors that
3768 * have already been closed (if the same NVE triggers a subsequent
3769 * rfapi_open(), the new peer is different and doesn't match the
3770 * peer of the holddown route, so the stale holddown route still
3771 * hangs around until it times out instead of just being replaced
3772 * by the fresh route).
3775 * We know that the new bi will have been inserted before any routes
3776 * in holddown, so we can skip any that came before it
3778 for (bi
= info_new
->next
; bi
; bi
= next
) {
3780 struct prefix pfx_vn
;
3781 struct prefix pfx_un
;
3783 int remote_peer_match
= 0;
3790 if (!CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
))
3794 * Must match VN address (nexthop of VPN route)
3796 if (rfapiGetNexthop(bi
->attr
, &pfx_vn
))
3798 if (!prefix_same(&pfx_vn
, &vn_prefix
))
3801 if (un_prefix_valid
&& /* new route UN addr */
3802 !rfapiGetUnAddrOfVpnBi(bi
, &pfx_un
)
3803 && /* old route UN addr */
3804 prefix_same(&pfx_un
, &un_prefix
)) { /* compare */
3807 if (!RFAPI_LOCAL_BI(bi
) && !RFAPI_LOCAL_BI(info_new
)
3808 && sockunion_same(&bi
->peer
->su
, &info_new
->peer
->su
)) {
3809 /* old & new are both remote, same peer */
3810 remote_peer_match
= 1;
3813 if (!un_match
& !remote_peer_match
)
3816 vnc_zlog_debug_verbose(
3817 "%s: removing holddown bi matching NVE of new route",
3819 if (bi
->extra
->vnc
.import
.timer
) {
3821 (struct thread
*)bi
->extra
->vnc
.import
.timer
;
3822 struct rfapi_withdraw
*wcb
= t
->arg
;
3824 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3827 rfapiExpireVpnNow(import_table
, rn
, bi
, 0);
3830 if (!original_had_routes
) {
3832 * We went from 0 usable routes to 1 usable route. Perform the
3833 * "Adding a Route" export process.
3835 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3836 vnc_zebra_add_prefix(bgp
, import_table
, rn
);
3839 * Check for nexthop change event
3840 * Note: the prefix_same() test below detects two situations:
3841 * 1. route is replaced, new route has different nexthop
3842 * 2. new route is added (original_nexthop is 0)
3844 struct prefix new_nexthop
;
3846 rfapiGetNexthop(attr
, &new_nexthop
);
3847 if (!prefix_same(&original_nexthop
, &new_nexthop
)) {
3849 * nexthop change event
3850 * vnc_direct_bgp_add_prefix() will recompute VN addr
3853 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3857 if (!(bgp
->rfapi_cfg
->flags
& BGP_VNC_CONFIG_CALLBACK_DISABLE
)) {
3858 for (n
= rn
; n
; n
= n
->parent
) {
3859 // rfapiDoRouteCallback(import_table, n, NULL);
3861 rfapiMonitorItNodeChanged(import_table
, rn
, NULL
);
3863 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 0);
3867 static void rfapiBgpInfoFilteredImportBadSafi(
3868 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3869 void *rfd
, /* set for looped back routes */
3871 struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3872 afi_t afi
, struct prefix_rd
*prd
,
3873 struct attr
*attr
, /* part of bgp_info */
3874 uint8_t type
, /* part of bgp_info */
3875 uint8_t sub_type
, /* part of bgp_info */
3876 uint32_t *label
) /* part of bgp_info */
3878 vnc_zlog_debug_verbose("%s: Error, bad safi", __func__
);
3881 static rfapi_bi_filtered_import_f
*
3882 rfapiBgpInfoFilteredImportFunction(safi_t safi
)
3886 return rfapiBgpInfoFilteredImportVPN
;
3889 return rfapiBgpInfoFilteredImportEncap
;
3893 zlog_err("%s: bad safi %d", __func__
, safi
);
3894 return rfapiBgpInfoFilteredImportBadSafi
;
3898 void rfapiProcessUpdate(struct peer
*peer
,
3899 void *rfd
, /* set when looped from RFP/RFAPI */
3900 struct prefix
*p
, struct prefix_rd
*prd
,
3901 struct attr
*attr
, afi_t afi
, safi_t safi
, uint8_t type
,
3902 uint8_t sub_type
, uint32_t *label
)
3906 struct rfapi_import_table
*it
;
3907 int has_ip_route
= 1;
3910 bgp
= bgp_get_default(); /* assume 1 instance for now */
3917 * look at high-order byte of RD. FF means MAC
3918 * address is present (VNC L2VPN)
3920 if ((safi
== SAFI_MPLS_VPN
)
3921 && (decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
)) {
3922 struct prefix pfx_mac_buf
;
3923 struct prefix pfx_nexthop_buf
;
3927 * Set flag if prefix and nexthop are the same - don't
3928 * add the route to normal IP-based import tables
3930 if (!rfapiGetNexthop(attr
, &pfx_nexthop_buf
)) {
3931 if (!prefix_cmp(&pfx_nexthop_buf
, p
)) {
3936 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
3937 pfx_mac_buf
.family
= AF_ETHERNET
;
3938 pfx_mac_buf
.prefixlen
= 48;
3939 memcpy(&pfx_mac_buf
.u
.prefix_eth
.octet
, prd
->val
+ 2, 6);
3942 * Find rt containing LNI (Logical Network ID), which
3943 * _should_ always be present when mac address is present
3945 rc
= rfapiEcommunityGetLNI(attr
->ecommunity
, &lni
);
3947 vnc_zlog_debug_verbose(
3948 "%s: rfapiEcommunityGetLNI returned %d, lni=%d, attr=%p",
3949 __func__
, rc
, lni
, attr
);
3951 it
= rfapiMacImportTableGet(bgp
, lni
);
3953 rfapiBgpInfoFilteredImportVPN(
3954 it
, FIF_ACTION_UPDATE
, peer
, rfd
,
3955 &pfx_mac_buf
, /* prefix */
3956 p
, /* aux prefix: IP addr */
3957 AFI_L2VPN
, prd
, attr
, type
, sub_type
, label
);
3965 * Iterate over all import tables; do a filtered import
3966 * for the afi/safi combination
3968 for (it
= h
->imports
; it
; it
= it
->next
) {
3969 (*rfapiBgpInfoFilteredImportFunction(safi
))(
3970 it
, FIF_ACTION_UPDATE
, peer
, rfd
, p
, /* prefix */
3971 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3974 if (safi
== SAFI_MPLS_VPN
) {
3975 vnc_direct_bgp_rh_add_route(bgp
, afi
, p
, peer
, attr
);
3976 rfapiBgpInfoFilteredImportVPN(
3977 bgp
->rfapi
->it_ce
, FIF_ACTION_UPDATE
, peer
, rfd
,
3979 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3984 void rfapiProcessWithdraw(struct peer
*peer
, void *rfd
, struct prefix
*p
,
3985 struct prefix_rd
*prd
, struct attr
*attr
, afi_t afi
,
3986 safi_t safi
, uint8_t type
, int kill
)
3990 struct rfapi_import_table
*it
;
3992 bgp
= bgp_get_default(); /* assume 1 instance for now */
3999 * look at high-order byte of RD. FF means MAC
4000 * address is present (VNC L2VPN)
4002 if (h
->import_mac
!= NULL
&& safi
== SAFI_MPLS_VPN
4003 && decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
) {
4004 struct prefix pfx_mac_buf
;
4005 void *cursor
= NULL
;
4008 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
4009 pfx_mac_buf
.family
= AF_ETHERNET
;
4010 pfx_mac_buf
.prefixlen
= 48;
4011 memcpy(&pfx_mac_buf
.u
.prefix_eth
, prd
->val
+ 2, 6);
4014 * withdraw does not contain attrs, so we don't have
4015 * access to the route's LNI, which would ordinarily
4016 * select the specific mac-based import table. Instead,
4017 * we must iterate over all mac-based tables and rely
4018 * on the RD to match.
4020 * If this approach is too slow, add an index where
4021 * key is {RD, peer} and value is the import table
4023 for (rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4025 rc
== 0; rc
= skiplist_next(h
->import_mac
, NULL
,
4026 (void **)&it
, &cursor
)) {
4029 vnc_zlog_debug_verbose(
4030 "%s: calling rfapiBgpInfoFilteredImportVPN(it=%p, afi=AFI_L2VPN)",
4034 rfapiBgpInfoFilteredImportVPN(
4036 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
4037 peer
, rfd
, &pfx_mac_buf
, /* prefix */
4038 p
, /* aux_prefix: IP */
4039 AFI_L2VPN
, prd
, attr
, type
, 0,
4040 NULL
); /* sub_type & label unused for withdraw
4046 * XXX For the case where the withdraw involves an L2
4047 * route with no IP information, we rely on the lack
4048 * of RT-list intersection to filter out the withdraw
4049 * from the IP-based import tables below
4053 * Iterate over all import tables; do a filtered import
4054 * for the afi/safi combination
4057 for (it
= h
->imports
; it
; it
= it
->next
) {
4058 (*rfapiBgpInfoFilteredImportFunction(safi
))(
4059 it
, (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
4060 peer
, rfd
, p
, /* prefix */
4061 NULL
, afi
, prd
, attr
, type
, 0,
4062 NULL
); /* sub_type & label unused for withdraw */
4065 /* TBD the deletion should happen after the lifetime expires */
4066 if (safi
== SAFI_MPLS_VPN
)
4067 vnc_direct_bgp_rh_del_route(bgp
, afi
, p
, peer
);
4069 if (safi
== SAFI_MPLS_VPN
) {
4070 rfapiBgpInfoFilteredImportVPN(
4072 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
), peer
,
4073 rfd
, p
, /* prefix */
4074 NULL
, afi
, prd
, attr
, type
, 0,
4075 NULL
); /* sub_type & label unused for withdraw */
4080 * TBD optimized withdraw timer algorithm for case of many
4081 * routes expiring at the same time due to peer drop.
4084 * 1. Visit all BIs in all ENCAP import tables.
4086 * a. If a bi's peer is the failed peer, remove the bi.
4087 * b. If the removed ENCAP bi was first in the list of
4088 * BIs at this ENCAP node, loop over all monitors
4091 * (1) for each ENCAP monitor, loop over all its
4092 * VPN node monitors and set their RFAPI_MON_FLAG_NEEDCALLBACK
4095 * 2. Visit all BIs in all VPN import tables.
4096 * a. If a bi's peer is the failed peer, remove the bi.
4097 * b. loop over all the VPN node monitors and set their
4098 * RFAPI_MON_FLAG_NEEDCALLBACK flags
4099 * c. If there are no BIs left at this VPN node,
4104 /* surprise, this gets called from peer_delete(), from rfapi_close() */
4105 static void rfapiProcessPeerDownRt(struct peer
*peer
,
4106 struct rfapi_import_table
*import_table
,
4107 afi_t afi
, safi_t safi
)
4109 struct route_node
*rn
;
4110 struct bgp_info
*bi
;
4111 struct route_table
*rt
;
4112 int (*timer_service_func
)(struct thread
*);
4114 assert(afi
== AFI_IP
|| afi
== AFI_IP6
);
4120 rt
= import_table
->imported_vpn
[afi
];
4121 timer_service_func
= rfapiWithdrawTimerVPN
;
4124 rt
= import_table
->imported_encap
[afi
];
4125 timer_service_func
= rfapiWithdrawTimerEncap
;
4132 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
4133 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
4134 if (bi
->peer
== peer
) {
4136 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)) {
4137 /* already in holddown, skip */
4141 if (safi
== SAFI_MPLS_VPN
) {
4142 RFAPI_UPDATE_ITABLE_COUNT(
4143 bi
, import_table
, afi
, -1);
4144 import_table
->holddown_count
[afi
] += 1;
4146 rfapiBiStartWithdrawTimer(import_table
, rn
, bi
,
4148 timer_service_func
);
4156 * This gets called when a peer connection drops. We have to remove
4157 * all the routes from this peer.
4159 * Current approach is crude. TBD Optimize by setting fewer timers and
4160 * grouping withdrawn routes so we can generate callbacks more
4163 void rfapiProcessPeerDown(struct peer
*peer
)
4167 struct rfapi_import_table
*it
;
4170 * If this peer is a "dummy" peer structure atached to a RFAPI
4171 * nve_descriptor, we don't need to walk the import tables
4172 * because the routes are already withdrawn by rfapi_close()
4174 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_IS_RFAPI_HD
))
4178 * 1. Visit all BIs in all ENCAP import tables.
4179 * Start withdraw timer on the BIs that match peer.
4181 * 2. Visit All BIs in all VPN import tables.
4182 * Start withdraw timer on the BIs that match peer.
4185 bgp
= bgp_get_default(); /* assume 1 instance for now */
4192 for (it
= h
->imports
; it
; it
= it
->next
) {
4193 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_ENCAP
);
4194 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_ENCAP
);
4195 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_MPLS_VPN
);
4196 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_MPLS_VPN
);
4200 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4201 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4206 * Import an entire RIB (for an afi/safi) to an import table RIB,
4207 * filtered according to the import table's RT list
4209 * TBD: does this function need additions to match rfapiProcessUpdate()
4210 * for, e.g., L2 handling?
4212 static void rfapiBgpTableFilteredImport(struct bgp
*bgp
,
4213 struct rfapi_import_table
*it
,
4214 afi_t afi
, safi_t safi
)
4216 struct bgp_node
*rn1
;
4217 struct bgp_node
*rn2
;
4219 /* Only these SAFIs have 2-level RIBS */
4220 assert(safi
== SAFI_MPLS_VPN
|| safi
== SAFI_ENCAP
);
4223 * Now visit all the rd nodes and the nodes of all the
4224 * route tables attached to them, and import the routes
4225 * if they have matching route targets
4227 for (rn1
= bgp_table_top(bgp
->rib
[afi
][safi
]); rn1
;
4228 rn1
= bgp_route_next(rn1
)) {
4231 for (rn2
= bgp_table_top(rn1
->info
); rn2
;
4232 rn2
= bgp_route_next(rn2
)) {
4234 struct bgp_info
*bi
;
4236 for (bi
= rn2
->info
; bi
; bi
= bi
->next
) {
4239 if (CHECK_FLAG(bi
->flags
,
4244 label
= decode_label(
4245 &bi
->extra
->label
[0]);
4246 (*rfapiBgpInfoFilteredImportFunction(
4248 it
, /* which import table */
4249 FIF_ACTION_UPDATE
, bi
->peer
,
4250 NULL
, &rn2
->p
, /* prefix */
4252 (struct prefix_rd
*)&rn1
->p
,
4254 bi
->sub_type
, &label
);
4262 /* per-bgp-instance rfapi data */
4263 struct rfapi
*bgp_rfapi_new(struct bgp
*bgp
)
4267 struct rfapi_rfp_cfg
*cfg
= NULL
;
4268 struct rfapi_rfp_cb_methods
*cbm
= NULL
;
4270 assert(bgp
->rfapi_cfg
== NULL
);
4272 h
= (struct rfapi
*)XCALLOC(MTYPE_RFAPI
, sizeof(struct rfapi
));
4274 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4275 h
->un
[afi
] = route_table_init();
4279 * initialize the ce import table
4281 h
->it_ce
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4282 sizeof(struct rfapi_import_table
));
4283 h
->it_ce
->imported_vpn
[AFI_IP
] = route_table_init();
4284 h
->it_ce
->imported_vpn
[AFI_IP6
] = route_table_init();
4285 h
->it_ce
->imported_encap
[AFI_IP
] = route_table_init();
4286 h
->it_ce
->imported_encap
[AFI_IP6
] = route_table_init();
4287 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4288 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4291 * Set up work queue for deferred rfapi_close operations
4293 h
->deferred_close_q
=
4294 work_queue_new(bm
->master
, "rfapi deferred close");
4295 h
->deferred_close_q
->spec
.workfunc
= rfapi_deferred_close_workfunc
;
4296 h
->deferred_close_q
->spec
.data
= h
;
4298 h
->rfp
= rfp_start(bm
->master
, &cfg
, &cbm
);
4299 bgp
->rfapi_cfg
= bgp_rfapi_cfg_new(cfg
);
4301 h
->rfp_methods
= *cbm
;
4306 void bgp_rfapi_destroy(struct bgp
*bgp
, struct rfapi
*h
)
4310 if (bgp
== NULL
|| h
== NULL
)
4313 if (h
->resolve_nve_nexthop
) {
4314 skiplist_free(h
->resolve_nve_nexthop
);
4315 h
->resolve_nve_nexthop
= NULL
;
4318 route_table_finish(h
->it_ce
->imported_vpn
[AFI_IP
]);
4319 route_table_finish(h
->it_ce
->imported_vpn
[AFI_IP6
]);
4320 route_table_finish(h
->it_ce
->imported_encap
[AFI_IP
]);
4321 route_table_finish(h
->it_ce
->imported_encap
[AFI_IP6
]);
4323 if (h
->import_mac
) {
4324 struct rfapi_import_table
*it
;
4329 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4331 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4334 rfapiImportTableFlush(it
);
4335 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
4337 skiplist_free(h
->import_mac
);
4338 h
->import_mac
= NULL
;
4341 work_queue_free_and_null(&h
->deferred_close_q
);
4346 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4347 route_table_finish(h
->un
[afi
]);
4350 XFREE(MTYPE_RFAPI_IMPORTTABLE
, h
->it_ce
);
4351 XFREE(MTYPE_RFAPI
, h
);
4354 struct rfapi_import_table
*
4355 rfapiImportTableRefAdd(struct bgp
*bgp
, struct ecommunity
*rt_import_list
,
4356 struct rfapi_nve_group_cfg
*rfg
)
4359 struct rfapi_import_table
*it
;
4365 for (it
= h
->imports
; it
; it
= it
->next
) {
4366 if (ecommunity_cmp(it
->rt_import_list
, rt_import_list
))
4370 vnc_zlog_debug_verbose("%s: matched it=%p", __func__
, it
);
4373 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4374 sizeof(struct rfapi_import_table
));
4376 it
->next
= h
->imports
;
4379 it
->rt_import_list
= ecommunity_dup(rt_import_list
);
4381 it
->monitor_exterior_orphans
=
4382 skiplist_new(0, NULL
, (void (*)(void *))prefix_free
);
4385 * fill import route tables from RIBs
4387 * Potential area for optimization. If this occurs when
4388 * tables are large (e.g., the operator adds a nve group
4389 * with a new RT list to a running system), it could take
4393 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4395 it
->imported_vpn
[afi
] = route_table_init();
4396 it
->imported_encap
[afi
] = route_table_init();
4398 rfapiBgpTableFilteredImport(bgp
, it
, afi
,
4400 rfapiBgpTableFilteredImport(bgp
, it
, afi
, SAFI_ENCAP
);
4402 vnc_import_bgp_exterior_redist_enable_it(bgp
, afi
, it
);
4412 * skiplist element free function
4414 static void delete_rem_pfx_na_free(void *na
)
4416 uint32_t *pCounter
= ((struct rfapi_nve_addr
*)na
)->info
;
4419 XFREE(MTYPE_RFAPI_NVE_ADDR
, na
);
4423 * Common deleter for IP and MAC import tables
4425 static void rfapiDeleteRemotePrefixesIt(
4426 struct bgp
*bgp
, struct rfapi_import_table
*it
, struct prefix
*un
,
4427 struct prefix
*vn
, struct prefix
*p
, int delete_active
,
4428 int delete_holddown
, uint32_t *pARcount
, uint32_t *pAHcount
,
4429 uint32_t *pHRcount
, uint32_t *pHHcount
,
4430 struct skiplist
*uniq_active_nves
, struct skiplist
*uniq_holddown_nves
)
4436 char buf_pfx
[PREFIX_STRLEN
];
4439 prefix2str(p
, buf_pfx
, sizeof(buf_pfx
));
4445 vnc_zlog_debug_verbose(
4446 "%s: entry, p=%s, delete_active=%d, delete_holddown=%d",
4447 __func__
, buf_pfx
, delete_active
, delete_holddown
);
4451 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4453 struct route_table
*rt
;
4454 struct route_node
*rn
;
4456 if (p
&& (family2afi(p
->family
) != afi
)) {
4460 rt
= it
->imported_vpn
[afi
];
4464 vnc_zlog_debug_verbose("%s: scanning rt for afi=%d", __func__
,
4467 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
4468 struct bgp_info
*bi
;
4469 struct bgp_info
*next
;
4471 if (VNC_DEBUG(IMPORT_DEL_REMOTE
)) {
4472 char p1line
[PREFIX_STRLEN
];
4473 char p2line
[PREFIX_STRLEN
];
4475 prefix2str(p
, p1line
, sizeof(p1line
));
4476 prefix2str(&rn
->p
, p2line
, sizeof(p2line
));
4477 vnc_zlog_debug_any("%s: want %s, have %s",
4478 __func__
, p1line
, p2line
);
4481 if (p
&& prefix_cmp(p
, &rn
->p
))
4485 char buf_pfx
[PREFIX_STRLEN
];
4487 prefix2str(&rn
->p
, buf_pfx
, sizeof(buf_pfx
));
4488 vnc_zlog_debug_verbose("%s: rn pfx=%s",
4492 /* TBD is this valid for afi == AFI_L2VPN? */
4493 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
4495 for (bi
= rn
->info
; bi
; bi
= next
) {
4504 vnc_zlog_debug_verbose("%s: examining bi %p",
4508 if (!rfapiGetNexthop(bi
->attr
, &qpt
))
4513 || !prefix_match(vn
, &qpt
)) {
4515 vnc_zlog_debug_verbose(
4516 "%s: continue at vn && !qpt_valid || !prefix_match(vn, &qpt)",
4523 if (!rfapiGetUnAddrOfVpnBi(bi
, &qct
))
4528 || !prefix_match(un
, &qct
)) {
4530 vnc_zlog_debug_verbose(
4531 "%s: continue at un && !qct_valid || !prefix_match(un, &qct)",
4543 * If this route is waiting to be deleted
4545 * a previous withdraw, we must cancel its
4548 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)) {
4549 if (!delete_holddown
)
4551 if (bi
->extra
->vnc
.import
.timer
) {
4557 struct rfapi_withdraw
*wcb
=
4561 ->holddown_count
[afi
] -=
4563 RFAPI_UPDATE_ITABLE_COUNT(
4564 bi
, wcb
->import_table
,
4566 XFREE(MTYPE_RFAPI_WITHDRAW
,
4576 vnc_zlog_debug_verbose(
4577 "%s: deleting bi %p (qct_valid=%d, qpt_valid=%d, delete_holddown=%d, delete_active=%d)",
4578 __func__
, bi
, qct_valid
, qpt_valid
,
4579 delete_holddown
, delete_active
);
4585 if (qct_valid
&& qpt_valid
) {
4587 struct rfapi_nve_addr na
;
4588 struct rfapi_nve_addr
*nap
;
4590 memset(&na
, 0, sizeof(na
));
4591 assert(!rfapiQprefix2Raddr(&qct
,
4593 assert(!rfapiQprefix2Raddr(&qpt
,
4596 if (skiplist_search(
4599 : uniq_holddown_nves
),
4600 &na
, (void **)&nap
)) {
4604 MTYPE_RFAPI_NVE_ADDR
,
4609 nap
->info
= is_active
4615 : uniq_holddown_nves
),
4618 rfapiNveAddr2Str(nap
, line
,
4623 vnc_direct_bgp_rh_del_route(bgp
, afi
, &rn
->p
,
4626 RFAPI_UPDATE_ITABLE_COUNT(bi
, it
, afi
, -1);
4627 it
->holddown_count
[afi
] += 1;
4628 rfapiExpireVpnNow(it
, rn
, bi
, 1);
4630 vnc_zlog_debug_verbose(
4631 "%s: incrementing count (is_active=%d)",
4632 __func__
, is_active
);
4645 * For use by the "clear vnc prefixes" command
4647 /*------------------------------------------
4648 * rfapiDeleteRemotePrefixes
4650 * UI helper: For use by the "clear vnc prefixes" command
4653 * un if set, tunnel must match this prefix
4654 * vn if set, nexthop prefix must match this prefix
4655 * p if set, prefix must match this prefix
4656 * it if set, only look in this import table
4659 * pARcount number of active routes deleted
4660 * pAHcount number of active nves deleted
4661 * pHRcount number of holddown routes deleted
4662 * pHHcount number of holddown nves deleted
4666 --------------------------------------------*/
4667 void rfapiDeleteRemotePrefixes(struct prefix
*un
, struct prefix
*vn
,
4669 struct rfapi_import_table
*arg_it
,
4670 int delete_active
, int delete_holddown
,
4671 uint32_t *pARcount
, uint32_t *pAHcount
,
4672 uint32_t *pHRcount
, uint32_t *pHHcount
)
4676 struct rfapi_import_table
*it
;
4677 uint32_t deleted_holddown_route_count
= 0;
4678 uint32_t deleted_active_route_count
= 0;
4679 uint32_t deleted_holddown_nve_count
= 0;
4680 uint32_t deleted_active_nve_count
= 0;
4681 struct skiplist
*uniq_holddown_nves
;
4682 struct skiplist
*uniq_active_nves
;
4686 bgp
= bgp_get_default(); /* assume 1 instance for now */
4687 /* If no bgp instantiated yet, no vnc prefixes exist */
4694 uniq_holddown_nves
=
4695 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4697 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4700 * Iterate over all import tables; do a filtered import
4701 * for the afi/safi combination
4710 vnc_zlog_debug_verbose(
4711 "%s: calling rfapiDeleteRemotePrefixesIt() on (IP) import %p",
4714 rfapiDeleteRemotePrefixesIt(
4715 bgp
, it
, un
, vn
, p
, delete_active
, delete_holddown
,
4716 &deleted_active_route_count
, &deleted_active_nve_count
,
4717 &deleted_holddown_route_count
,
4718 &deleted_holddown_nve_count
, uniq_active_nves
,
4719 uniq_holddown_nves
);
4728 * Now iterate over L2 import tables
4730 if (h
->import_mac
&& !(p
&& (p
->family
!= AF_ETHERNET
))) {
4732 void *cursor
= NULL
;
4736 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4738 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4741 vnc_zlog_debug_verbose(
4742 "%s: calling rfapiDeleteRemotePrefixesIt() on import_mac %p",
4745 rfapiDeleteRemotePrefixesIt(
4746 bgp
, it
, un
, vn
, p
, delete_active
,
4747 delete_holddown
, &deleted_active_route_count
,
4748 &deleted_active_nve_count
,
4749 &deleted_holddown_route_count
,
4750 &deleted_holddown_nve_count
, uniq_active_nves
,
4751 uniq_holddown_nves
);
4756 * our custom element freeing function above counts as it deletes
4758 skiplist_free(uniq_holddown_nves
);
4759 skiplist_free(uniq_active_nves
);
4762 *pARcount
= deleted_active_route_count
;
4764 *pAHcount
= deleted_active_nve_count
;
4766 *pHRcount
= deleted_holddown_route_count
;
4768 *pHHcount
= deleted_holddown_nve_count
;
4773 /*------------------------------------------
4774 * rfapiCountRemoteRoutes
4776 * UI helper: count VRF routes from BGP side
4781 * pALRcount count of active local routes
4782 * pARRcount count of active remote routes
4783 * pHRcount count of holddown routes
4784 * pIRcount count of direct imported routes
4788 --------------------------------------------*/
4789 void rfapiCountAllItRoutes(int *pALRcount
, /* active local routes */
4790 int *pARRcount
, /* active remote routes */
4791 int *pHRcount
, /* holddown routes */
4792 int *pIRcount
) /* imported routes */
4796 struct rfapi_import_table
*it
;
4799 int total_active_local
= 0;
4800 int total_active_remote
= 0;
4801 int total_holddown
= 0;
4802 int total_imported
= 0;
4804 bgp
= bgp_get_default(); /* assume 1 instance for now */
4811 * Iterate over all import tables; do a filtered import
4812 * for the afi/safi combination
4815 for (it
= h
->imports
; it
; it
= it
->next
) {
4817 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4819 total_active_local
+= it
->local_count
[afi
];
4820 total_active_remote
+= it
->remote_count
[afi
];
4821 total_holddown
+= it
->holddown_count
[afi
];
4822 total_imported
+= it
->imported_count
[afi
];
4829 if (h
->import_mac
) {
4831 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4833 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4836 total_active_local
+= it
->local_count
[AFI_L2VPN
];
4837 total_active_remote
+= it
->remote_count
[AFI_L2VPN
];
4838 total_holddown
+= it
->holddown_count
[AFI_L2VPN
];
4839 total_imported
+= it
->imported_count
[AFI_L2VPN
];
4845 *pALRcount
= total_active_local
;
4848 *pARRcount
= total_active_remote
;
4851 *pHRcount
= total_holddown
;
4854 *pIRcount
= total_imported
;
4858 /*------------------------------------------
4859 * rfapiGetHolddownFromLifetime
4861 * calculate holddown value based on lifetime
4867 * Holddown value based on lifetime, holddown_factor,
4868 * and RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
4870 --------------------------------------------*/
4871 /* hold down time maxes out at RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY */
4872 uint32_t rfapiGetHolddownFromLifetime(uint32_t lifetime
)
4877 bgp
= bgp_get_default();
4878 if (bgp
&& bgp
->rfapi_cfg
)
4879 factor
= bgp
->rfapi_cfg
->rfp_cfg
.holddown_factor
;
4881 factor
= RFAPI_RFP_CFG_DEFAULT_HOLDDOWN_FACTOR
;
4883 if (factor
< 100 || lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4884 lifetime
= lifetime
* factor
/ 100;
4885 if (lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4888 return RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
;