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
26 #include "lib/zebra.h"
27 #include "lib/prefix.h"
28 #include "lib/agg_table.h"
30 #include "lib/memory.h"
32 #include "lib/skiplist.h"
33 #include "lib/thread.h"
34 #include "lib/stream.h"
35 #include "lib/lib_errors.h"
37 #include "bgpd/bgpd.h"
38 #include "bgpd/bgp_ecommunity.h"
39 #include "bgpd/bgp_attr.h"
40 #include "bgpd/bgp_route.h"
41 #include "bgpd/bgp_mplsvpn.h" /* prefix_rd2str() */
42 #include "bgpd/bgp_vnc_types.h"
43 #include "bgpd/bgp_rd.h"
45 #include "bgpd/rfapi/rfapi.h"
46 #include "bgpd/rfapi/bgp_rfapi_cfg.h"
47 #include "bgpd/rfapi/rfapi_backend.h"
48 #include "bgpd/rfapi/rfapi_import.h"
49 #include "bgpd/rfapi/rfapi_private.h"
50 #include "bgpd/rfapi/rfapi_monitor.h"
51 #include "bgpd/rfapi/rfapi_nve_addr.h"
52 #include "bgpd/rfapi/rfapi_vty.h"
53 #include "bgpd/rfapi/vnc_export_bgp.h"
54 #include "bgpd/rfapi/vnc_export_bgp_p.h"
55 #include "bgpd/rfapi/vnc_zebra.h"
56 #include "bgpd/rfapi/vnc_import_bgp.h"
57 #include "bgpd/rfapi/vnc_import_bgp_p.h"
58 #include "bgpd/rfapi/rfapi_rib.h"
59 #include "bgpd/rfapi/rfapi_encap_tlv.h"
60 #include "bgpd/rfapi/vnc_debug.h"
62 #ifdef HAVE_GLIBC_BACKTRACE
63 /* for backtrace and friends */
65 #endif /* HAVE_GLIBC_BACKTRACE */
67 #undef DEBUG_MONITOR_MOVE_SHORTER
68 #undef DEBUG_RETURNED_NHL
69 #undef DEBUG_ROUTE_COUNTERS
70 #undef DEBUG_ENCAP_MONITOR
73 #undef DEBUG_BI_SEARCH
76 * Allocated for each withdraw timer instance; freed when the timer
77 * expires or is canceled
79 struct rfapi_withdraw
{
80 struct rfapi_import_table
*import_table
;
81 struct agg_node
*node
;
82 struct bgp_path_info
*info
;
83 safi_t safi
; /* used only for bulk operations */
85 * For import table node reference count checking (i.e., debugging).
86 * Normally when a timer expires, lockoffset should be 0. However, if
87 * the timer expiration function is called directly (e.g.,
88 * rfapiExpireVpnNow), the node could be locked by a preceding
89 * agg_route_top() or agg_route_next() in a loop, so we need to pass
97 * It's evil and fiendish. It's compiler-dependent.
98 * ? Might need LDFLAGS -rdynamic to produce all function names
100 void rfapiDebugBacktrace(void)
102 #ifdef HAVE_GLIBC_BACKTRACE
103 #define RFAPI_DEBUG_BACKTRACE_NENTRIES 200
104 void *buf
[RFAPI_DEBUG_BACKTRACE_NENTRIES
];
109 size
= backtrace(buf
, RFAPI_DEBUG_BACKTRACE_NENTRIES
);
110 syms
= backtrace_symbols(buf
, size
);
112 for (i
= 0; i
< size
&& i
< RFAPI_DEBUG_BACKTRACE_NENTRIES
; ++i
) {
113 vnc_zlog_debug_verbose("backtrace[%2zu]: %s", i
, syms
[i
]);
123 * Count remote routes and compare with actively-maintained values.
124 * Abort if they disagree.
126 void rfapiCheckRouteCount(void)
128 struct bgp
*bgp
= bgp_get_default();
130 struct rfapi_import_table
*it
;
138 for (it
= h
->imports
; it
; it
= it
->next
) {
139 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
141 struct agg_table
*rt
;
144 int holddown_count
= 0;
146 int imported_count
= 0;
147 int remote_count
= 0;
149 rt
= it
->imported_vpn
[afi
];
151 for (rn
= agg_route_top(rt
); rn
;
152 rn
= agg_route_next(rn
)) {
153 struct bgp_path_info
*bpi
;
154 struct bgp_path_info
*next
;
156 for (bpi
= rn
->info
; bpi
; bpi
= next
) {
159 if (CHECK_FLAG(bpi
->flags
,
164 if (RFAPI_LOCAL_BI(bpi
)) {
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 #ifdef 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 * agg_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 agg_node
*rn
, safi_t safi
, int lockoffset
)
219 unsigned int count_bpi
= 0;
220 unsigned int count_monitor
= 0;
221 struct bgp_path_info
*bpi
;
222 struct rfapi_monitor_encap
*hme
;
223 struct rfapi_monitor_vpn
*hmv
;
225 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->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 */
267 case SAFI_LABELED_UNICAST
:
270 assert(!"Passed in safi should be impossible");
274 if (count_bpi
+ count_monitor
+ lockoffset
275 != agg_node_get_lock_count(rn
)) {
276 vnc_zlog_debug_verbose(
277 "%s: count_bpi=%d, count_monitor=%d, lockoffset=%d, rn->lock=%d",
278 __func__
, count_bpi
, count_monitor
, lockoffset
,
279 agg_node_get_lock_count(rn
));
285 * Perform deferred rfapi_close operations that were queued
288 static wq_item_status
rfapi_deferred_close_workfunc(struct work_queue
*q
,
291 struct rfapi_descriptor
*rfd
= data
;
292 struct rfapi
*h
= q
->spec
.data
;
294 assert(!(h
->flags
& RFAPI_INCALLBACK
));
296 vnc_zlog_debug_verbose("%s: completed deferred close on handle %p",
302 * Extract layer 2 option from Encap TLVS in BGP attrs
304 int rfapiGetL2o(struct attr
*attr
, struct rfapi_l2address_option
*l2o
)
307 struct bgp_attr_encap_subtlv
*pEncap
;
309 for (pEncap
= bgp_attr_get_vnc_subtlvs(attr
); pEncap
;
310 pEncap
= pEncap
->next
) {
312 if (pEncap
->type
== BGP_VNC_SUBTLV_TYPE_RFPOPTION
) {
314 == RFAPI_VN_OPTION_TYPE_L2ADDR
) {
316 if (pEncap
->value
[1] == 14) {
317 memcpy(l2o
->macaddr
.octet
,
334 l2o
->logical_net_id
=
337 + ((pEncap
->value
[14]
340 + ((pEncap
->value
[13]
355 * Extract the lifetime from the Tunnel Encap attribute of a route in
358 int rfapiGetVncLifetime(struct attr
*attr
, uint32_t *lifetime
)
360 struct bgp_attr_encap_subtlv
*pEncap
;
362 *lifetime
= RFAPI_INFINITE_LIFETIME
; /* default to infinite */
366 for (pEncap
= bgp_attr_get_vnc_subtlvs(attr
); pEncap
;
367 pEncap
= pEncap
->next
) {
370 == BGP_VNC_SUBTLV_TYPE_LIFETIME
) { /* lifetime */
371 if (pEncap
->length
== 4) {
372 memcpy(lifetime
, pEncap
->value
, 4);
373 *lifetime
= ntohl(*lifetime
);
384 * Look for UN address in Encap attribute
386 int rfapiGetVncTunnelUnAddr(struct attr
*attr
, struct prefix
*p
)
388 struct bgp_attr_encap_subtlv
*pEncap
;
389 bgp_encap_types tun_type
= BGP_ENCAP_TYPE_MPLS
;/*Default tunnel type*/
391 bgp_attr_extcom_tunnel_type(attr
, &tun_type
);
392 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
395 /* MPLS carries UN address in next hop */
396 rfapiNexthop2Prefix(attr
, p
);
397 if (p
->family
!= AF_UNSPEC
)
403 for (pEncap
= attr
->encap_subtlvs
; pEncap
;
404 pEncap
= pEncap
->next
) {
407 == BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
) { /* un
410 switch (pEncap
->length
) {
414 p
->prefixlen
= IPV4_MAX_BITLEN
;
415 memcpy(p
->u
.val
, pEncap
->value
,
422 p
->family
= AF_INET6
;
423 p
->prefixlen
= IPV6_MAX_BITLEN
;
424 memcpy(p
->u
.val
, pEncap
->value
,
437 * Get UN address wherever it might be
439 int rfapiGetUnAddrOfVpnBi(struct bgp_path_info
*bpi
, struct prefix
*p
)
441 /* If it's in this route's VNC attribute, we're done */
442 if (!rfapiGetVncTunnelUnAddr(bpi
->attr
, p
))
445 * Otherwise, see if it's cached from a corresponding ENCAP SAFI
449 switch (bpi
->extra
->vnc
.import
.un_family
) {
452 p
->family
= bpi
->extra
->vnc
.import
.un_family
;
453 p
->u
.prefix4
= bpi
->extra
->vnc
.import
.un
.addr4
;
454 p
->prefixlen
= IPV4_MAX_BITLEN
;
459 p
->family
= bpi
->extra
->vnc
.import
.un_family
;
460 p
->u
.prefix6
= bpi
->extra
->vnc
.import
.un
.addr6
;
461 p
->prefixlen
= IPV6_MAX_BITLEN
;
466 p
->family
= AF_UNSPEC
;
467 #ifdef DEBUG_ENCAP_MONITOR
468 vnc_zlog_debug_verbose(
469 "%s: bpi->extra->vnc.import.un_family is 0, no UN addr",
481 * Make a new bgp_path_info from gathered parameters
483 static struct bgp_path_info
*rfapiBgpInfoCreate(struct attr
*attr
,
484 struct peer
*peer
, void *rfd
,
485 struct prefix_rd
*prd
,
486 uint8_t type
, uint8_t sub_type
,
489 struct bgp_path_info
*new;
491 new = info_make(type
, sub_type
, 0, peer
, attr
, NULL
);
493 new->attr
= bgp_attr_intern(attr
);
495 bgp_path_info_extra_get(new);
497 new->extra
->vnc
.import
.rd
= *prd
;
498 new->extra
->vnc
.import
.create_time
= monotime(NULL
);
501 encode_label(*label
, &new->extra
->label
[0]);
509 * Frees bgp_path_info as used in import tables (parts are not
510 * allocated exactly the way they are in the main RIBs)
512 static void rfapiBgpInfoFree(struct bgp_path_info
*goner
)
518 vnc_zlog_debug_verbose("%s: calling peer_unlock(%p), #%d",
519 __func__
, goner
->peer
,
521 peer_unlock(goner
->peer
);
524 bgp_attr_unintern(&goner
->attr
);
527 bgp_path_info_extra_free(&goner
->extra
);
528 XFREE(MTYPE_BGP_ROUTE
, goner
);
531 struct rfapi_import_table
*rfapiMacImportTableGetNoAlloc(struct bgp
*bgp
,
535 struct rfapi_import_table
*it
= NULL
;
536 uintptr_t lni_as_ptr
= lni
;
545 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
))
551 struct rfapi_import_table
*rfapiMacImportTableGet(struct bgp
*bgp
, uint32_t lni
)
554 struct rfapi_import_table
*it
= NULL
;
555 uintptr_t lni_as_ptr
= lni
;
560 if (!h
->import_mac
) {
561 /* default cmp is good enough for LNI */
562 h
->import_mac
= skiplist_new(0, NULL
, NULL
);
565 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
)) {
567 struct ecommunity
*enew
;
568 struct ecommunity_val eval
;
571 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
572 sizeof(struct rfapi_import_table
));
573 /* set RT list of new import table based on LNI */
574 memset((char *)&eval
, 0, sizeof(eval
));
575 eval
.val
[0] = 0; /* VNC L2VPN */
576 eval
.val
[1] = 2; /* VNC L2VPN */
577 eval
.val
[5] = (lni
>> 16) & 0xff;
578 eval
.val
[6] = (lni
>> 8) & 0xff;
579 eval
.val
[7] = (lni
>> 0) & 0xff;
581 enew
= ecommunity_new();
582 ecommunity_add_val(enew
, &eval
, false, false);
583 it
->rt_import_list
= enew
;
585 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
586 it
->imported_vpn
[afi
] = agg_table_init();
587 it
->imported_encap
[afi
] = agg_table_init();
590 it
->l2_logical_net_id
= lni
;
592 skiplist_insert(h
->import_mac
, (void *)lni_as_ptr
, it
);
600 * Implement MONITOR_MOVE_SHORTER(original_node) from
601 * RFAPI-Import-Event-Handling.txt
603 * Returns pointer to the list of moved monitors
605 static struct rfapi_monitor_vpn
*
606 rfapiMonitorMoveShorter(struct agg_node
*original_vpn_node
, int lockoffset
)
608 struct bgp_path_info
*bpi
;
609 struct agg_node
*par
;
610 struct rfapi_monitor_vpn
*m
;
611 struct rfapi_monitor_vpn
*mlast
;
612 struct rfapi_monitor_vpn
*moved
;
614 int parent_already_refcounted
= 0;
616 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
, lockoffset
);
618 #ifdef DEBUG_MONITOR_MOVE_SHORTER
620 vnc_zlog_debug_verbose("%s: called with node pfx=%pFX",
621 __func__
, &original_vpn_node
->p
);
626 * 1. If there is at least one bpi (either regular route or
627 * route marked as withdrawn, with a pending timer) at
628 * original_node with a valid UN address, we're done. Return.
630 for (bpi
= original_vpn_node
->info
; bpi
; bpi
= bpi
->next
) {
633 if (!rfapiGetUnAddrOfVpnBi(bpi
, &pfx
)) {
634 #ifdef DEBUG_MONITOR_MOVE_SHORTER
635 vnc_zlog_debug_verbose(
636 "%s: have valid UN at original node, no change",
644 * 2. Travel up the tree (toward less-specific prefixes) from
645 * original_node to find the first node that has at least
646 * one route (even if it is only a withdrawn route) with a
647 * valid UN address. Call this node "Node P."
649 for (par
= agg_node_parent(original_vpn_node
); par
;
650 par
= agg_node_parent(par
)) {
651 for (bpi
= par
->info
; bpi
; bpi
= bpi
->next
) {
653 if (!rfapiGetUnAddrOfVpnBi(bpi
, &pfx
)) {
662 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 0);
666 * If no less-specific routes, try to use the 0/0 node
669 const struct prefix
*p
;
670 /* this isn't necessarily 0/0 */
671 par
= agg_route_table_top(original_vpn_node
);
674 p
= agg_node_get_prefix(par
);
676 * If we got the top node but it wasn't 0/0,
679 if (par
&& p
->prefixlen
) {
680 agg_unlock_node(par
); /* maybe free */
685 ++parent_already_refcounted
;
690 * Create 0/0 node if it isn't there
693 struct prefix pfx_default
;
694 const struct prefix
*p
= agg_node_get_prefix(original_vpn_node
);
696 memset(&pfx_default
, 0, sizeof(pfx_default
));
697 pfx_default
.family
= p
->family
;
699 /* creates default node if none exists */
700 par
= agg_node_get(agg_get_table(original_vpn_node
),
702 ++parent_already_refcounted
;
706 * 3. Move each of the monitors found at original_node to Node P.
707 * These are "Moved Monitors."
712 * Attach at end so that the list pointer we return points
713 * only to the moved routes
715 for (m
= RFAPI_MONITOR_VPN(par
), mlast
= NULL
; m
;
716 mlast
= m
, m
= m
->next
)
720 moved
= mlast
->next
= RFAPI_MONITOR_VPN(original_vpn_node
);
722 moved
= RFAPI_MONITOR_VPN_W_ALLOC(par
) =
723 RFAPI_MONITOR_VPN(original_vpn_node
);
725 if (RFAPI_MONITOR_VPN(
726 original_vpn_node
)) /* check agg, so not allocated */
727 RFAPI_MONITOR_VPN_W_ALLOC(original_vpn_node
) = NULL
;
730 * update the node pointers on the monitors
732 for (m
= moved
; m
; m
= m
->next
) {
737 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
,
738 parent_already_refcounted
- movecount
);
739 while (movecount
> parent_already_refcounted
) {
741 ++parent_already_refcounted
;
743 while (movecount
< parent_already_refcounted
) {
744 /* unlikely, but code defensively */
745 agg_unlock_node(par
);
746 --parent_already_refcounted
;
748 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
,
749 movecount
+ lockoffset
);
750 while (movecount
--) {
751 agg_unlock_node(original_vpn_node
);
754 #ifdef DEBUG_MONITOR_MOVE_SHORTER
756 vnc_zlog_debug_verbose("%s: moved to node pfx=%pFX", __func__
,
766 * Implement MONITOR_MOVE_LONGER(new_node) from
767 * RFAPI-Import-Event-Handling.txt
769 static void rfapiMonitorMoveLonger(struct agg_node
*new_vpn_node
)
771 struct rfapi_monitor_vpn
*monitor
;
772 struct rfapi_monitor_vpn
*mlast
;
773 struct bgp_path_info
*bpi
;
774 struct agg_node
*par
;
775 const struct prefix
*new_vpn_node_p
= agg_node_get_prefix(new_vpn_node
);
777 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
780 * Make sure we have at least one valid route at the new node
782 for (bpi
= new_vpn_node
->info
; bpi
; bpi
= bpi
->next
) {
784 if (!rfapiGetUnAddrOfVpnBi(bpi
, &pfx
))
789 vnc_zlog_debug_verbose(
790 "%s: no valid routes at node %p, so not attempting moves",
791 __func__
, new_vpn_node
);
796 * Find first parent node that has monitors
798 for (par
= agg_node_parent(new_vpn_node
); par
;
799 par
= agg_node_parent(par
)) {
800 if (RFAPI_MONITOR_VPN(par
))
805 vnc_zlog_debug_verbose(
806 "%s: no parent nodes with monitors, done", __func__
);
811 * Check each of these monitors to see of their longest-match
812 * is now the updated node. Move any such monitors to the more-
813 * specific updated node
815 for (mlast
= NULL
, monitor
= RFAPI_MONITOR_VPN(par
); monitor
;) {
817 * If new longest match for monitor prefix is the new
818 * route's prefix, move monitor to new route's prefix
820 if (prefix_match(new_vpn_node_p
, &monitor
->p
)) {
823 mlast
->next
= monitor
->next
;
825 RFAPI_MONITOR_VPN_W_ALLOC(par
) = monitor
->next
;
830 monitor
->next
= RFAPI_MONITOR_VPN(new_vpn_node
);
831 RFAPI_MONITOR_VPN_W_ALLOC(new_vpn_node
) = monitor
;
832 monitor
->node
= new_vpn_node
;
834 agg_lock_node(new_vpn_node
); /* incr refcount */
836 monitor
= mlast
? mlast
->next
: RFAPI_MONITOR_VPN(par
);
838 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 1);
839 /* decr refcount after we're done with par as this might
841 agg_unlock_node(par
);
846 monitor
= monitor
->next
;
849 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
853 static void rfapiBgpInfoChainFree(struct bgp_path_info
*bpi
)
855 struct bgp_path_info
*next
;
860 * If there is a timer waiting to delete this bpi, cancel
861 * the timer and delete immediately
863 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
864 && bpi
->extra
->vnc
.import
.timer
) {
865 struct rfapi_withdraw
*wcb
=
866 THREAD_ARG(bpi
->extra
->vnc
.import
.timer
);
868 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
869 THREAD_OFF(bpi
->extra
->vnc
.import
.timer
);
874 rfapiBgpInfoFree(bpi
);
879 static void rfapiImportTableFlush(struct rfapi_import_table
*it
)
886 ecommunity_free(&it
->rt_import_list
);
887 it
->rt_import_list
= NULL
;
889 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
893 for (rn
= agg_route_top(it
->imported_vpn
[afi
]); rn
;
894 rn
= agg_route_next(rn
)) {
896 * Each route_node has:
897 * aggregate: points to rfapi_it_extra with monitor
899 * info: points to chain of bgp_path_info
901 /* free bgp_path_info and its children */
902 rfapiBgpInfoChainFree(rn
->info
);
905 rfapiMonitorExtraFlush(SAFI_MPLS_VPN
, rn
);
908 for (rn
= agg_route_top(it
->imported_encap
[afi
]); rn
;
909 rn
= agg_route_next(rn
)) {
910 /* free bgp_path_info and its children */
911 rfapiBgpInfoChainFree(rn
->info
);
914 rfapiMonitorExtraFlush(SAFI_ENCAP
, rn
);
917 agg_table_finish(it
->imported_vpn
[afi
]);
918 agg_table_finish(it
->imported_encap
[afi
]);
920 if (it
->monitor_exterior_orphans
) {
921 skiplist_free(it
->monitor_exterior_orphans
);
925 void rfapiImportTableRefDelByIt(struct bgp
*bgp
,
926 struct rfapi_import_table
*it_target
)
929 struct rfapi_import_table
*it
;
930 struct rfapi_import_table
*prev
= NULL
;
937 for (it
= h
->imports
; it
; prev
= it
, it
= it
->next
) {
943 assert(it
->refcount
);
949 prev
->next
= it
->next
;
951 h
->imports
= it
->next
;
953 rfapiImportTableFlush(it
);
954 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
958 #ifdef RFAPI_REQUIRE_ENCAP_BEEC
960 * Look for magic BGP Encapsulation Extended Community value
961 * Format in RFC 5512 Sect. 4.5
963 static int rfapiEcommunitiesMatchBeec(struct ecommunity
*ecom
,
964 bgp_encap_types type
)
971 for (i
= 0; i
< (ecom
->size
* ECOMMUNITY_SIZE
); i
+= ECOMMUNITY_SIZE
) {
977 if (ep
[0] == ECOMMUNITY_ENCODE_OPAQUE
978 && ep
[1] == ECOMMUNITY_OPAQUE_SUBTYPE_ENCAP
979 && ep
[6] == ((type
&& 0xff00) >> 8)
980 && ep
[7] == (type
& 0xff)) {
989 int rfapiEcommunitiesIntersect(struct ecommunity
*e1
, struct ecommunity
*e2
)
998 s1
= ecommunity_ecom2str(e1
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
999 s2
= ecommunity_ecom2str(e2
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
1000 vnc_zlog_debug_verbose("%s: e1[%s], e2[%s]", __func__
, s1
, s2
);
1001 XFREE(MTYPE_ECOMMUNITY_STR
, s1
);
1002 XFREE(MTYPE_ECOMMUNITY_STR
, s2
);
1005 for (i
= 0; i
< e1
->size
; ++i
) {
1006 for (j
= 0; j
< e2
->size
; ++j
) {
1007 if (!memcmp(e1
->val
+ (i
* ECOMMUNITY_SIZE
),
1008 e2
->val
+ (j
* ECOMMUNITY_SIZE
),
1018 int rfapiEcommunityGetLNI(struct ecommunity
*ecom
, uint32_t *lni
)
1023 for (i
= 0; i
< ecom
->size
; ++i
) {
1024 uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1026 if ((*(p
+ 0) == 0x00) && (*(p
+ 1) == 0x02)) {
1028 *lni
= (*(p
+ 5) << 16) | (*(p
+ 6) << 8)
1037 int rfapiEcommunityGetEthernetTag(struct ecommunity
*ecom
, uint16_t *tag_id
)
1039 struct bgp
*bgp
= bgp_get_default();
1040 *tag_id
= 0; /* default to untagged */
1044 for (i
= 0; i
< ecom
->size
; ++i
) {
1047 const uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1049 /* High-order octet of type. */
1052 if (*p
++ == ECOMMUNITY_ROUTE_TARGET
) {
1053 if (encode
== ECOMMUNITY_ENCODE_AS4
) {
1054 p
= ptr_get_be32(p
, &as
);
1055 } else if (encode
== ECOMMUNITY_ENCODE_AS
) {
1058 p
+= 2; /* skip next two, tag/vid
1059 always in lowest bytes */
1061 if (as
== bgp
->as
) {
1062 *tag_id
= *p
++ << 8;
1072 static int rfapiVpnBiNhEqualsPt(struct bgp_path_info
*bpi
,
1073 struct rfapi_ip_addr
*hpt
)
1080 family
= BGP_MP_NEXTHOP_FAMILY(bpi
->attr
->mp_nexthop_len
);
1082 if (hpt
->addr_family
!= family
)
1087 if (bpi
->attr
->mp_nexthop_global_in
.s_addr
1088 != hpt
->addr
.v4
.s_addr
)
1093 if (IPV6_ADDR_CMP(&bpi
->attr
->mp_nexthop_global
, &hpt
->addr
.v6
))
1106 * Compare 2 VPN BIs. Return true if they have the same VN and UN addresses
1108 static int rfapiVpnBiSamePtUn(struct bgp_path_info
*bpi1
,
1109 struct bgp_path_info
*bpi2
)
1111 struct prefix pfx_un1
;
1112 struct prefix pfx_un2
;
1118 * VN address comparisons
1121 if (BGP_MP_NEXTHOP_FAMILY(bpi1
->attr
->mp_nexthop_len
)
1122 != BGP_MP_NEXTHOP_FAMILY(bpi2
->attr
->mp_nexthop_len
)) {
1126 switch (BGP_MP_NEXTHOP_FAMILY(bpi1
->attr
->mp_nexthop_len
)) {
1128 if (bpi1
->attr
->mp_nexthop_global_in
.s_addr
1129 != bpi2
->attr
->mp_nexthop_global_in
.s_addr
)
1134 if (IPV6_ADDR_CMP(&bpi1
->attr
->mp_nexthop_global
,
1135 &bpi2
->attr
->mp_nexthop_global
))
1143 memset(&pfx_un1
, 0, sizeof(pfx_un1
));
1144 memset(&pfx_un2
, 0, sizeof(pfx_un2
));
1147 * UN address comparisons
1149 if (rfapiGetVncTunnelUnAddr(bpi1
->attr
, &pfx_un1
)) {
1151 pfx_un1
.family
= bpi1
->extra
->vnc
.import
.un_family
;
1152 switch (bpi1
->extra
->vnc
.import
.un_family
) {
1155 bpi1
->extra
->vnc
.import
.un
.addr4
;
1159 bpi1
->extra
->vnc
.import
.un
.addr6
;
1162 pfx_un1
.family
= AF_UNSPEC
;
1168 if (rfapiGetVncTunnelUnAddr(bpi2
->attr
, &pfx_un2
)) {
1170 pfx_un2
.family
= bpi2
->extra
->vnc
.import
.un_family
;
1171 switch (bpi2
->extra
->vnc
.import
.un_family
) {
1174 bpi2
->extra
->vnc
.import
.un
.addr4
;
1178 bpi2
->extra
->vnc
.import
.un
.addr6
;
1181 pfx_un2
.family
= AF_UNSPEC
;
1187 if (pfx_un1
.family
== AF_UNSPEC
|| pfx_un2
.family
== AF_UNSPEC
)
1190 if (pfx_un1
.family
!= pfx_un2
.family
)
1193 switch (pfx_un1
.family
) {
1195 if (!IPV4_ADDR_SAME(&pfx_un1
.u
.prefix4
, &pfx_un2
.u
.prefix4
))
1199 if (!IPV6_ADDR_SAME(&pfx_un1
.u
.prefix6
, &pfx_un2
.u
.prefix6
))
1208 uint8_t rfapiRfpCost(struct attr
*attr
)
1210 if (attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_LOCAL_PREF
)) {
1211 if (attr
->local_pref
> 255) {
1214 return 255 - attr
->local_pref
;
1220 /*------------------------------------------
1223 * Find Layer 2 options in an option chain
1229 * l2o layer 2 options extracted
1233 * 1 no options found
1235 --------------------------------------------*/
1236 int rfapi_extract_l2o(
1237 struct bgp_tea_options
*pHop
, /* chain of options */
1238 struct rfapi_l2address_option
*l2o
) /* return extracted value */
1240 struct bgp_tea_options
*p
;
1242 for (p
= pHop
; p
; p
= p
->next
) {
1243 if ((p
->type
== RFAPI_VN_OPTION_TYPE_L2ADDR
)
1244 && (p
->length
>= 8)) {
1248 memcpy(&l2o
->macaddr
, v
, 6);
1250 l2o
->label
= ((v
[6] << 12) & 0xff000)
1251 + ((v
[7] << 4) & 0xff0)
1252 + ((v
[8] >> 4) & 0xf);
1254 l2o
->local_nve_id
= (uint8_t)v
[10];
1256 l2o
->logical_net_id
=
1257 (v
[11] << 16) + (v
[12] << 8) + (v
[13] << 0);
1265 static struct rfapi_next_hop_entry
*
1266 rfapiRouteInfo2NextHopEntry(struct rfapi_ip_prefix
*rprefix
,
1267 struct bgp_path_info
*bpi
, /* route to encode */
1268 uint32_t lifetime
, /* use this in nhe */
1269 struct agg_node
*rn
) /* req for L2 eth addr */
1271 struct rfapi_next_hop_entry
*new;
1272 int have_vnc_tunnel_un
= 0;
1273 const struct prefix
*p
= agg_node_get_prefix(rn
);
1275 #ifdef DEBUG_ENCAP_MONITOR
1276 vnc_zlog_debug_verbose("%s: entry, bpi %p, rn %p", __func__
, bpi
, rn
);
1279 new = XCALLOC(MTYPE_RFAPI_NEXTHOP
, sizeof(struct rfapi_next_hop_entry
));
1281 new->prefix
= *rprefix
;
1284 && decode_rd_type(bpi
->extra
->vnc
.import
.rd
.val
)
1285 == RD_TYPE_VNC_ETH
) {
1288 struct rfapi_vn_option
*vo
;
1290 vo
= XCALLOC(MTYPE_RFAPI_VN_OPTION
,
1291 sizeof(struct rfapi_vn_option
));
1293 vo
->type
= RFAPI_VN_OPTION_TYPE_L2ADDR
;
1295 memcpy(&vo
->v
.l2addr
.macaddr
, &p
->u
.prefix_eth
.octet
, ETH_ALEN
);
1296 /* only low 3 bytes of this are significant */
1297 (void)rfapiEcommunityGetLNI(bgp_attr_get_ecommunity(bpi
->attr
),
1298 &vo
->v
.l2addr
.logical_net_id
);
1299 (void)rfapiEcommunityGetEthernetTag(
1300 bgp_attr_get_ecommunity(bpi
->attr
),
1301 &vo
->v
.l2addr
.tag_id
);
1303 /* local_nve_id comes from lower byte of RD type */
1304 vo
->v
.l2addr
.local_nve_id
= bpi
->extra
->vnc
.import
.rd
.val
[1];
1306 /* label comes from MP_REACH_NLRI label */
1307 vo
->v
.l2addr
.label
= decode_label(&bpi
->extra
->label
[0]);
1309 new->vn_options
= vo
;
1312 * If there is an auxiliary prefix (i.e., host IP address),
1313 * use it as the nexthop prefix instead of the query prefix
1315 if (bpi
->extra
->vnc
.import
.aux_prefix
.family
) {
1316 rfapiQprefix2Rprefix(&bpi
->extra
->vnc
.import
.aux_prefix
,
1321 bgp_encap_types tun_type
= BGP_ENCAP_TYPE_MPLS
; /*Default*/
1322 new->prefix
.cost
= rfapiRfpCost(bpi
->attr
);
1324 struct bgp_attr_encap_subtlv
*pEncap
;
1326 switch (BGP_MP_NEXTHOP_FAMILY(bpi
->attr
->mp_nexthop_len
)) {
1328 new->vn_address
.addr_family
= AF_INET
;
1329 new->vn_address
.addr
.v4
= bpi
->attr
->mp_nexthop_global_in
;
1333 new->vn_address
.addr_family
= AF_INET6
;
1334 new->vn_address
.addr
.v6
= bpi
->attr
->mp_nexthop_global
;
1338 zlog_warn("%s: invalid vpn nexthop length: %d", __func__
,
1339 bpi
->attr
->mp_nexthop_len
);
1340 rfapi_free_next_hop_list(new);
1344 for (pEncap
= bgp_attr_get_vnc_subtlvs(bpi
->attr
); pEncap
;
1345 pEncap
= pEncap
->next
) {
1346 switch (pEncap
->type
) {
1347 case BGP_VNC_SUBTLV_TYPE_LIFETIME
:
1348 /* use configured lifetime, not attr lifetime */
1352 zlog_warn("%s: unknown VNC option type %d", __func__
,
1359 bgp_attr_extcom_tunnel_type(bpi
->attr
, &tun_type
);
1360 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
1362 /* MPLS carries UN address in next hop */
1363 rfapiNexthop2Prefix(bpi
->attr
, &p
);
1364 if (p
.family
!= AF_UNSPEC
) {
1365 rfapiQprefix2Raddr(&p
, &new->un_address
);
1366 have_vnc_tunnel_un
= 1;
1370 for (pEncap
= bpi
->attr
->encap_subtlvs
; pEncap
; pEncap
= pEncap
->next
) {
1371 switch (pEncap
->type
) {
1372 case BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
:
1374 * Overrides ENCAP UN address, if any
1376 switch (pEncap
->length
) {
1379 new->un_address
.addr_family
= AF_INET
;
1380 memcpy(&new->un_address
.addr
.v4
, pEncap
->value
,
1382 have_vnc_tunnel_un
= 1;
1386 new->un_address
.addr_family
= AF_INET6
;
1387 memcpy(&new->un_address
.addr
.v6
, pEncap
->value
,
1389 have_vnc_tunnel_un
= 1;
1394 "%s: invalid tunnel subtlv UN addr length (%d) for bpi %p",
1395 __func__
, pEncap
->length
, bpi
);
1400 zlog_warn("%s: unknown Encap Attribute option type %d",
1401 __func__
, pEncap
->type
);
1406 new->un_options
= rfapi_encap_tlv_to_un_option(bpi
->attr
);
1408 #ifdef DEBUG_ENCAP_MONITOR
1409 vnc_zlog_debug_verbose("%s: line %d: have_vnc_tunnel_un=%d", __func__
,
1410 __LINE__
, have_vnc_tunnel_un
);
1413 if (!have_vnc_tunnel_un
&& bpi
->extra
) {
1415 * use cached UN address from ENCAP route
1417 new->un_address
.addr_family
= bpi
->extra
->vnc
.import
.un_family
;
1418 switch (new->un_address
.addr_family
) {
1420 new->un_address
.addr
.v4
=
1421 bpi
->extra
->vnc
.import
.un
.addr4
;
1424 new->un_address
.addr
.v6
=
1425 bpi
->extra
->vnc
.import
.un
.addr6
;
1428 zlog_warn("%s: invalid UN addr family (%d) for bpi %p",
1429 __func__
, new->un_address
.addr_family
, bpi
);
1430 rfapi_free_next_hop_list(new);
1435 new->lifetime
= lifetime
;
1439 int rfapiHasNonRemovedRoutes(struct agg_node
*rn
)
1441 struct bgp_path_info
*bpi
;
1443 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
1446 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
1447 && (bpi
->extra
&& !rfapiGetUnAddrOfVpnBi(bpi
, &pfx
))) {
1455 #ifdef DEBUG_IT_NODES
1459 void rfapiDumpNode(struct agg_node
*rn
)
1461 struct bgp_path_info
*bpi
;
1463 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
1464 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
1466 int ctrc
= rfapiGetUnAddrOfVpnBi(bpi
, &pfx
);
1469 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
1470 && (bpi
->extra
&& !ctrc
)) {
1477 vnc_zlog_debug_verbose(
1478 " bpi=%p, nr=%d, flags=0x%x, extra=%p, ctrc=%d", bpi
,
1479 nr
, bpi
->flags
, bpi
->extra
, ctrc
);
1484 static int rfapiNhlAddNodeRoutes(
1485 struct agg_node
*rn
, /* in */
1486 struct rfapi_ip_prefix
*rprefix
, /* in */
1487 uint32_t lifetime
, /* in */
1488 int removed
, /* in */
1489 struct rfapi_next_hop_entry
**head
, /* in/out */
1490 struct rfapi_next_hop_entry
**tail
, /* in/out */
1491 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1492 struct agg_node
*rfd_rib_node
, /* preload this NVE rib node */
1493 struct prefix
*pfx_target_original
) /* query target */
1495 struct bgp_path_info
*bpi
;
1496 struct rfapi_next_hop_entry
*new;
1497 struct prefix pfx_un
;
1498 struct skiplist
*seen_nexthops
;
1500 const struct prefix
*p
= agg_node_get_prefix(rn
);
1501 int is_l2
= (p
->family
== AF_ETHERNET
);
1504 struct agg_table
*atable
= agg_get_table(rfd_rib_node
);
1505 struct rfapi_descriptor
*rfd
;
1508 rfd
= agg_get_table_info(atable
);
1510 if (rfapiRibFTDFilterRecentPrefix(rfd
, rn
,
1511 pfx_target_original
))
1517 skiplist_new(0, vnc_prefix_cmp
, prefix_free_lists
);
1519 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
1521 struct prefix pfx_vn
;
1522 struct prefix
*newpfx
;
1524 if (removed
&& !CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
1525 #ifdef DEBUG_RETURNED_NHL
1526 vnc_zlog_debug_verbose(
1527 "%s: want holddown, this route not holddown, skip",
1532 if (!removed
&& CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
1541 * Check for excluded VN address
1543 if (rfapiVpnBiNhEqualsPt(bpi
, exclude_vnaddr
))
1547 * Check for VN address (nexthop) copied already
1550 /* L2 routes: semantic nexthop in aux_prefix; VN addr
1552 pfx_vn
= bpi
->extra
->vnc
.import
.aux_prefix
;
1554 rfapiNexthop2Prefix(bpi
->attr
, &pfx_vn
);
1556 if (!skiplist_search(seen_nexthops
, &pfx_vn
, NULL
)) {
1557 #ifdef DEBUG_RETURNED_NHL
1558 vnc_zlog_debug_verbose(
1559 "%s: already put VN/nexthop %pFX, skip",
1565 if (rfapiGetUnAddrOfVpnBi(bpi
, &pfx_un
)) {
1566 #ifdef DEBUG_ENCAP_MONITOR
1567 vnc_zlog_debug_verbose(
1568 "%s: failed to get UN address of this VPN bpi",
1574 newpfx
= prefix_new();
1576 skiplist_insert(seen_nexthops
, newpfx
, newpfx
);
1578 new = rfapiRouteInfo2NextHopEntry(rprefix
, bpi
, lifetime
, rn
);
1580 if (rfapiRibPreloadBi(rfd_rib_node
, &pfx_vn
, &pfx_un
,
1582 /* duplicate filtered by RIB */
1583 rfapi_free_next_hop_list(new);
1590 (*tail
)->next
= new;
1599 skiplist_free(seen_nexthops
);
1608 * omit_node is meant for the situation where we are adding a subtree
1609 * of a parent of some original requested node. The response already
1610 * contains the original requested node, and we don't want to duplicate
1611 * its routes in the list, so we skip it if the right or left node
1612 * matches (of course, we still travel down its child subtrees).
1614 static int rfapiNhlAddSubtree(
1615 struct agg_node
*rn
, /* in */
1616 uint32_t lifetime
, /* in */
1617 struct rfapi_next_hop_entry
**head
, /* in/out */
1618 struct rfapi_next_hop_entry
**tail
, /* in/out */
1619 struct agg_node
*omit_node
, /* in */
1620 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1621 struct agg_table
*rfd_rib_table
, /* preload here */
1622 struct prefix
*pfx_target_original
) /* query target */
1624 struct rfapi_ip_prefix rprefix
;
1627 /* FIXME: need to find a better way here to work without sticking our
1628 * hands in node->link */
1629 if (agg_node_left(rn
) && agg_node_left(rn
) != omit_node
) {
1630 if (agg_node_left(rn
)->info
) {
1631 const struct prefix
*p
=
1632 agg_node_get_prefix(agg_node_left(rn
));
1634 struct agg_node
*rib_rn
= NULL
;
1636 rfapiQprefix2Rprefix(p
, &rprefix
);
1638 rib_rn
= agg_node_get(rfd_rib_table
, p
);
1640 count
= rfapiNhlAddNodeRoutes(
1641 agg_node_left(rn
), &rprefix
, lifetime
, 0, head
,
1642 tail
, exclude_vnaddr
, rib_rn
,
1643 pfx_target_original
);
1645 count
= rfapiNhlAddNodeRoutes(
1646 agg_node_left(rn
), &rprefix
, lifetime
,
1647 1, head
, tail
, exclude_vnaddr
, rib_rn
,
1648 pfx_target_original
);
1652 agg_unlock_node(rib_rn
);
1656 if (agg_node_right(rn
) && agg_node_right(rn
) != omit_node
) {
1657 if (agg_node_right(rn
)->info
) {
1658 const struct prefix
*p
=
1659 agg_node_get_prefix(agg_node_right(rn
));
1661 struct agg_node
*rib_rn
= NULL
;
1663 rfapiQprefix2Rprefix(p
, &rprefix
);
1665 rib_rn
= agg_node_get(rfd_rib_table
, p
);
1667 count
= rfapiNhlAddNodeRoutes(
1668 agg_node_right(rn
), &rprefix
, lifetime
, 0, head
,
1669 tail
, exclude_vnaddr
, rib_rn
,
1670 pfx_target_original
);
1672 count
= rfapiNhlAddNodeRoutes(
1673 agg_node_right(rn
), &rprefix
, lifetime
,
1674 1, head
, tail
, exclude_vnaddr
, rib_rn
,
1675 pfx_target_original
);
1679 agg_unlock_node(rib_rn
);
1683 if (agg_node_left(rn
)) {
1684 rcount
+= rfapiNhlAddSubtree(
1685 agg_node_left(rn
), lifetime
, head
, tail
, omit_node
,
1686 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1688 if (agg_node_right(rn
)) {
1689 rcount
+= rfapiNhlAddSubtree(
1690 agg_node_right(rn
), lifetime
, head
, tail
, omit_node
,
1691 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1698 * Implementation of ROUTE_LIST(node) from RFAPI-Import-Event-Handling.txt
1700 * Construct an rfapi nexthop list based on the routes attached to
1701 * the specified node.
1703 * If there are any routes that do NOT have BGP_PATH_REMOVED set,
1704 * return those only. If there are ONLY routes with BGP_PATH_REMOVED,
1705 * then return those, and also include all the non-removed routes from the
1706 * next less-specific node (i.e., this node's parent) at the end.
1708 struct rfapi_next_hop_entry
*rfapiRouteNode2NextHopList(
1709 struct agg_node
*rn
, uint32_t lifetime
, /* put into nexthop entries */
1710 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1711 struct agg_table
*rfd_rib_table
, /* preload here */
1712 struct prefix
*pfx_target_original
) /* query target */
1714 struct rfapi_ip_prefix rprefix
;
1715 struct rfapi_next_hop_entry
*answer
= NULL
;
1716 struct rfapi_next_hop_entry
*last
= NULL
;
1717 struct agg_node
*parent
;
1718 const struct prefix
*p
= agg_node_get_prefix(rn
);
1720 struct agg_node
*rib_rn
;
1722 #ifdef DEBUG_RETURNED_NHL
1723 vnc_zlog_debug_verbose("%s: called with node pfx=%rRN", __func__
, rn
);
1724 rfapiDebugBacktrace();
1727 rfapiQprefix2Rprefix(p
, &rprefix
);
1729 rib_rn
= rfd_rib_table
? agg_node_get(rfd_rib_table
, p
) : NULL
;
1732 * Add non-withdrawn routes at this node
1734 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 0, &answer
, &last
,
1735 exclude_vnaddr
, rib_rn
,
1736 pfx_target_original
);
1739 * If the list has at least one entry, it's finished
1742 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
, &last
, NULL
,
1743 exclude_vnaddr
, rfd_rib_table
,
1744 pfx_target_original
);
1745 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
,
1747 #ifdef DEBUG_RETURNED_NHL
1748 rfapiPrintNhl(NULL
, answer
);
1751 agg_unlock_node(rib_rn
);
1756 * Add withdrawn routes at this node
1758 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 1, &answer
, &last
,
1759 exclude_vnaddr
, rib_rn
,
1760 pfx_target_original
);
1762 agg_unlock_node(rib_rn
);
1764 // rfapiPrintNhl(NULL, answer);
1767 * walk up the tree until we find a node with non-deleted
1768 * routes, then add them
1770 for (parent
= agg_node_parent(rn
); parent
;
1771 parent
= agg_node_parent(parent
)) {
1772 if (rfapiHasNonRemovedRoutes(parent
)) {
1778 * Add non-withdrawn routes from less-specific prefix
1781 const struct prefix
*p
= agg_node_get_prefix(parent
);
1783 rib_rn
= rfd_rib_table
? agg_node_get(rfd_rib_table
, p
) : NULL
;
1784 rfapiQprefix2Rprefix(p
, &rprefix
);
1785 count
+= rfapiNhlAddNodeRoutes(parent
, &rprefix
, lifetime
, 0,
1786 &answer
, &last
, exclude_vnaddr
,
1787 rib_rn
, pfx_target_original
);
1788 count
+= rfapiNhlAddSubtree(parent
, lifetime
, &answer
, &last
,
1789 rn
, exclude_vnaddr
, rfd_rib_table
,
1790 pfx_target_original
);
1792 agg_unlock_node(rib_rn
);
1795 * There is no parent with non-removed routes. Still need to
1796 * add subtree of original node if it contributed routes to the
1800 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
,
1801 &last
, rn
, exclude_vnaddr
,
1803 pfx_target_original
);
1806 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
, count
,
1808 #ifdef DEBUG_RETURNED_NHL
1809 rfapiPrintNhl(NULL
, answer
);
1815 * Construct nexthop list of all routes in table
1817 struct rfapi_next_hop_entry
*rfapiRouteTable2NextHopList(
1818 struct agg_table
*rt
, uint32_t lifetime
, /* put into nexthop entries */
1819 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1820 struct agg_table
*rfd_rib_table
, /* preload this NVE rib table */
1821 struct prefix
*pfx_target_original
) /* query target */
1823 struct agg_node
*rn
;
1824 struct rfapi_next_hop_entry
*biglist
= NULL
;
1825 struct rfapi_next_hop_entry
*nhl
;
1826 struct rfapi_next_hop_entry
*tail
= NULL
;
1829 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
1831 nhl
= rfapiRouteNode2NextHopList(rn
, lifetime
, exclude_vnaddr
,
1833 pfx_target_original
);
1835 tail
= biglist
= nhl
;
1842 while (tail
->next
) {
1849 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1853 struct rfapi_next_hop_entry
*rfapiEthRouteNode2NextHopList(
1854 struct agg_node
*rn
, struct rfapi_ip_prefix
*rprefix
,
1855 uint32_t lifetime
, /* put into nexthop entries */
1856 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1857 struct agg_table
*rfd_rib_table
, /* preload NVE rib table */
1858 struct prefix
*pfx_target_original
) /* query target */
1861 struct rfapi_next_hop_entry
*answer
= NULL
;
1862 struct rfapi_next_hop_entry
*last
= NULL
;
1863 struct agg_node
*rib_rn
;
1865 rib_rn
= rfd_rib_table
1866 ? agg_node_get(rfd_rib_table
, agg_node_get_prefix(rn
))
1869 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 0, &answer
, &last
,
1870 NULL
, rib_rn
, pfx_target_original
);
1872 #ifdef DEBUG_ENCAP_MONITOR
1873 vnc_zlog_debug_verbose("%s: node %p: %d non-holddown routes", __func__
,
1878 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 1, &answer
,
1879 &last
, exclude_vnaddr
, rib_rn
,
1880 pfx_target_original
);
1881 vnc_zlog_debug_verbose("%s: node %p: %d holddown routes",
1882 __func__
, rn
, count
);
1886 agg_unlock_node(rib_rn
);
1888 #ifdef DEBUG_RETURNED_NHL
1889 rfapiPrintNhl(NULL
, answer
);
1897 * Construct nexthop list of all routes in table
1899 struct rfapi_next_hop_entry
*rfapiEthRouteTable2NextHopList(
1900 uint32_t logical_net_id
, 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 agg_table
*rfd_rib_table
, /* preload NVE rib node */
1904 struct prefix
*pfx_target_original
) /* query target */
1906 struct rfapi_import_table
*it
;
1907 struct bgp
*bgp
= bgp_get_default();
1908 struct agg_table
*rt
;
1909 struct agg_node
*rn
;
1910 struct rfapi_next_hop_entry
*biglist
= NULL
;
1911 struct rfapi_next_hop_entry
*nhl
;
1912 struct rfapi_next_hop_entry
*tail
= NULL
;
1916 it
= rfapiMacImportTableGet(bgp
, logical_net_id
);
1917 rt
= it
->imported_vpn
[AFI_L2VPN
];
1919 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
1921 nhl
= rfapiEthRouteNode2NextHopList(
1922 rn
, rprefix
, lifetime
, exclude_vnaddr
, rfd_rib_table
,
1923 pfx_target_original
);
1925 tail
= biglist
= nhl
;
1932 while (tail
->next
) {
1939 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1944 * Insert a new bpi to the imported route table node,
1945 * keeping the list of BPIs sorted best route first
1947 static void rfapiBgpInfoAttachSorted(struct agg_node
*rn
,
1948 struct bgp_path_info
*info_new
, afi_t afi
,
1952 struct bgp_path_info
*prev
;
1953 struct bgp_path_info
*next
;
1954 char pfx_buf
[PREFIX2STR_BUFFER
];
1957 bgp
= bgp_get_default(); /* assume 1 instance for now */
1959 if (VNC_DEBUG(IMPORT_BI_ATTACH
)) {
1960 vnc_zlog_debug_verbose("%s: info_new->peer=%p", __func__
,
1962 vnc_zlog_debug_verbose("%s: info_new->peer->su_remote=%p",
1963 __func__
, info_new
->peer
->su_remote
);
1966 for (prev
= NULL
, next
= rn
->info
; next
;
1967 prev
= next
, next
= next
->next
) {
1968 enum bgp_path_selection_reason reason
;
1971 || (!CHECK_FLAG(info_new
->flags
, BGP_PATH_REMOVED
)
1972 && CHECK_FLAG(next
->flags
, BGP_PATH_REMOVED
))
1973 || bgp_path_info_cmp_compatible(bgp
, info_new
, next
,
1976 == -1) { /* -1 if 1st is better */
1980 vnc_zlog_debug_verbose("%s: prev=%p, next=%p", __func__
, prev
, next
);
1982 prev
->next
= info_new
;
1984 rn
->info
= info_new
;
1986 info_new
->prev
= prev
;
1987 info_new
->next
= next
;
1989 next
->prev
= info_new
;
1990 bgp_attr_intern(info_new
->attr
);
1993 static void rfapiBgpInfoDetach(struct agg_node
*rn
, struct bgp_path_info
*bpi
)
1996 * Remove the route (doubly-linked)
1998 // bgp_attr_unintern (&bpi->attr);
2000 bpi
->next
->prev
= bpi
->prev
;
2002 bpi
->prev
->next
= bpi
->next
;
2004 rn
->info
= bpi
->next
;
2008 * For L3-indexed import tables
2010 static int rfapi_bi_peer_rd_cmp(const void *b1
, const void *b2
)
2012 const struct bgp_path_info
*bpi1
= b1
;
2013 const struct bgp_path_info
*bpi2
= b2
;
2018 if (bpi1
->peer
< bpi2
->peer
)
2020 if (bpi1
->peer
> bpi2
->peer
)
2026 return vnc_prefix_cmp(
2027 (const struct prefix
*)&bpi1
->extra
->vnc
.import
.rd
,
2028 (const struct prefix
*)&bpi2
->extra
->vnc
.import
.rd
);
2032 * For L2-indexed import tables
2033 * The BPIs in these tables should ALWAYS have an aux_prefix set because
2034 * they arrive via IPv4 or IPv6 advertisements.
2036 static int rfapi_bi_peer_rd_aux_cmp(const void *b1
, const void *b2
)
2038 const struct bgp_path_info
*bpi1
= b1
;
2039 const struct bgp_path_info
*bpi2
= b2
;
2045 if (bpi1
->peer
< bpi2
->peer
)
2047 if (bpi1
->peer
> bpi2
->peer
)
2053 rc
= vnc_prefix_cmp((struct prefix
*)&bpi1
->extra
->vnc
.import
.rd
,
2054 (struct prefix
*)&bpi2
->extra
->vnc
.import
.rd
);
2060 * L2 import tables can have multiple entries with the
2061 * same MAC address, same RD, but different L3 addresses.
2063 * Use presence of aux_prefix with AF=ethernet and prefixlen=1
2064 * as magic value to signify explicit wildcarding of the aux_prefix.
2065 * This magic value will not appear in bona fide bpi entries in
2066 * the import table, but is allowed in the "fake" bpi used to
2067 * probe the table when searching. (We have to test both b1 and b2
2068 * because there is no guarantee of the order the test key and
2069 * the real key will be passed)
2071 if ((bpi1
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2072 && (bpi1
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))
2073 || (bpi2
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2074 && (bpi2
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))) {
2077 * wildcard aux address specified
2082 return vnc_prefix_cmp(&bpi1
->extra
->vnc
.import
.aux_prefix
,
2083 &bpi2
->extra
->vnc
.import
.aux_prefix
);
2088 * Index on RD and Peer
2090 static void rfapiItBiIndexAdd(struct agg_node
*rn
, /* Import table VPN node */
2091 struct bgp_path_info
*bpi
) /* new BPI */
2093 struct skiplist
*sl
;
2094 const struct prefix
*p
;
2100 vnc_zlog_debug_verbose("%s: bpi %p, peer %p, rd %pRD", __func__
, bpi
,
2101 bpi
->peer
, &bpi
->extra
->vnc
.import
.rd
);
2103 sl
= RFAPI_RDINDEX_W_ALLOC(rn
);
2105 p
= agg_node_get_prefix(rn
);
2106 if (AF_ETHERNET
== p
->family
) {
2107 sl
= skiplist_new(0, rfapi_bi_peer_rd_aux_cmp
, NULL
);
2109 sl
= skiplist_new(0, rfapi_bi_peer_rd_cmp
, NULL
);
2111 RFAPI_IT_EXTRA_GET(rn
)->u
.vpn
.idx_rd
= sl
;
2112 agg_lock_node(rn
); /* for skiplist */
2114 assert(!skiplist_insert(sl
, (void *)bpi
, (void *)bpi
));
2115 agg_lock_node(rn
); /* for skiplist entry */
2117 /* NB: BPIs in import tables are not refcounted */
2120 static void rfapiItBiIndexDump(struct agg_node
*rn
)
2122 struct skiplist
*sl
;
2123 void *cursor
= NULL
;
2124 struct bgp_path_info
*k
;
2125 struct bgp_path_info
*v
;
2128 sl
= RFAPI_RDINDEX(rn
);
2132 for (rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
); !rc
;
2133 rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
)) {
2135 char buf
[RD_ADDRSTRLEN
];
2136 char buf_aux_pfx
[PREFIX_STRLEN
];
2138 prefix_rd2str(&k
->extra
->vnc
.import
.rd
, buf
, sizeof(buf
));
2139 if (k
->extra
->vnc
.import
.aux_prefix
.family
) {
2140 prefix2str(&k
->extra
->vnc
.import
.aux_prefix
,
2141 buf_aux_pfx
, sizeof(buf_aux_pfx
));
2143 strlcpy(buf_aux_pfx
, "(none)", sizeof(buf_aux_pfx
));
2145 vnc_zlog_debug_verbose("bpi %p, peer %p, rd %s, aux_prefix %s",
2146 k
, k
->peer
, buf
, buf_aux_pfx
);
2150 static struct bgp_path_info
*rfapiItBiIndexSearch(
2151 struct agg_node
*rn
, /* Import table VPN node */
2152 struct prefix_rd
*prd
, struct peer
*peer
,
2153 const struct prefix
*aux_prefix
) /* optional L3 addr for L2 ITs */
2155 struct skiplist
*sl
;
2157 struct bgp_path_info bpi_fake
= {0};
2158 struct bgp_path_info_extra bpi_extra
= {0};
2159 struct bgp_path_info
*bpi_result
;
2161 sl
= RFAPI_RDINDEX(rn
);
2165 #ifdef DEBUG_BI_SEARCH
2167 char buf_aux_pfx
[PREFIX_STRLEN
];
2170 prefix2str(aux_prefix
, buf_aux_pfx
,
2171 sizeof(buf_aux_pfx
));
2173 strlcpy(buf_aux_pfx
, "(nil)", sizeof(buf_aux_pfx
));
2175 vnc_zlog_debug_verbose(
2176 "%s want prd=%pRD, peer=%p, aux_prefix=%s", __func__
,
2177 prd
, peer
, buf_aux_pfx
);
2178 rfapiItBiIndexDump(rn
);
2182 /* threshold is a WAG */
2183 if (sl
->count
< 3) {
2184 #ifdef DEBUG_BI_SEARCH
2185 vnc_zlog_debug_verbose("%s: short list algorithm", __func__
);
2187 /* if short list, linear search might be faster */
2188 for (bpi_result
= rn
->info
; bpi_result
;
2189 bpi_result
= bpi_result
->next
) {
2190 #ifdef DEBUG_BI_SEARCH
2191 vnc_zlog_debug_verbose(
2192 "%s: bpi has prd=%pRD, peer=%p", __func__
,
2193 &bpi_result
->extra
->vnc
.import
.rd
,
2196 if (peer
== bpi_result
->peer
2197 && !prefix_cmp((struct prefix
*)&bpi_result
->extra
2199 (struct prefix
*)prd
)) {
2201 #ifdef DEBUG_BI_SEARCH
2202 vnc_zlog_debug_verbose(
2203 "%s: peer and RD same, doing aux_prefix check",
2209 &bpi_result
->extra
->vnc
.import
2212 #ifdef DEBUG_BI_SEARCH
2213 vnc_zlog_debug_verbose("%s: match",
2223 bpi_fake
.peer
= peer
;
2224 bpi_fake
.extra
= &bpi_extra
;
2225 bpi_fake
.extra
->vnc
.import
.rd
= *prd
;
2227 bpi_fake
.extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
2230 bpi_fake
.extra
->vnc
.import
.aux_prefix
.family
= AF_ETHERNET
;
2231 bpi_fake
.extra
->vnc
.import
.aux_prefix
.prefixlen
= 1;
2234 rc
= skiplist_search(sl
, (void *)&bpi_fake
, (void *)&bpi_result
);
2237 #ifdef DEBUG_BI_SEARCH
2238 vnc_zlog_debug_verbose("%s: no match", __func__
);
2243 #ifdef DEBUG_BI_SEARCH
2244 vnc_zlog_debug_verbose("%s: matched bpi=%p", __func__
, bpi_result
);
2250 static void rfapiItBiIndexDel(struct agg_node
*rn
, /* Import table VPN node */
2251 struct bgp_path_info
*bpi
) /* old BPI */
2253 struct skiplist
*sl
;
2256 vnc_zlog_debug_verbose("%s: bpi %p, peer %p, rd %pRD", __func__
, bpi
,
2257 bpi
->peer
, &bpi
->extra
->vnc
.import
.rd
);
2259 sl
= RFAPI_RDINDEX(rn
);
2262 rc
= skiplist_delete(sl
, (void *)(bpi
), (void *)bpi
);
2264 rfapiItBiIndexDump(rn
);
2268 agg_unlock_node(rn
); /* for skiplist entry */
2270 /* NB: BPIs in import tables are not refcounted */
2274 * Add a backreference at the ENCAP node to the VPN route that
2278 rfapiMonitorEncapAdd(struct rfapi_import_table
*import_table
,
2279 struct prefix
*p
, /* VN address */
2280 struct agg_node
*vpn_rn
, /* VPN node */
2281 struct bgp_path_info
*vpn_bpi
) /* VPN bpi/route */
2283 afi_t afi
= family2afi(p
->family
);
2284 struct agg_node
*rn
;
2285 struct rfapi_monitor_encap
*m
;
2288 rn
= agg_node_get(import_table
->imported_encap
[afi
], p
); /* locks rn */
2291 m
= XCALLOC(MTYPE_RFAPI_MONITOR_ENCAP
,
2292 sizeof(struct rfapi_monitor_encap
));
2298 /* insert to encap node's list */
2299 m
->next
= RFAPI_MONITOR_ENCAP(rn
);
2302 RFAPI_MONITOR_ENCAP_W_ALLOC(rn
) = m
;
2304 /* for easy lookup when deleting vpn route */
2305 vpn_bpi
->extra
->vnc
.import
.hme
= m
;
2307 vnc_zlog_debug_verbose(
2308 "%s: it=%p, vpn_bpi=%p, afi=%d, encap rn=%p, setting vpn_bpi->extra->vnc.import.hme=%p",
2309 __func__
, import_table
, vpn_bpi
, afi
, rn
, m
);
2311 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
2312 bgp_attr_intern(vpn_bpi
->attr
);
2315 static void rfapiMonitorEncapDelete(struct bgp_path_info
*vpn_bpi
)
2318 * Remove encap monitor
2320 vnc_zlog_debug_verbose("%s: vpn_bpi=%p", __func__
, vpn_bpi
);
2321 if (vpn_bpi
->extra
) {
2322 struct rfapi_monitor_encap
*hme
=
2323 vpn_bpi
->extra
->vnc
.import
.hme
;
2327 vnc_zlog_debug_verbose("%s: hme=%p", __func__
, hme
);
2329 /* Refcount checking takes too long here */
2330 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 0);
2332 hme
->next
->prev
= hme
->prev
;
2334 hme
->prev
->next
= hme
->next
;
2336 RFAPI_MONITOR_ENCAP_W_ALLOC(hme
->rn
) =
2338 /* Refcount checking takes too long here */
2339 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 1);
2341 /* see if the struct rfapi_it_extra is empty and can be
2343 rfapiMonitorExtraPrune(SAFI_ENCAP
, hme
->rn
);
2345 agg_unlock_node(hme
->rn
); /* decr ref count */
2346 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, hme
);
2347 vpn_bpi
->extra
->vnc
.import
.hme
= NULL
;
2353 * quagga lib/thread.h says this must return int even though
2354 * it doesn't do anything with the return value
2356 static void rfapiWithdrawTimerVPN(struct thread
*t
)
2358 struct rfapi_withdraw
*wcb
= THREAD_ARG(t
);
2359 struct bgp_path_info
*bpi
= wcb
->info
;
2360 struct bgp
*bgp
= bgp_get_default();
2361 const struct prefix
*p
;
2362 struct rfapi_monitor_vpn
*moved
;
2366 vnc_zlog_debug_verbose(
2367 "%s: NULL BGP pointer, assume shutdown race condition!!!",
2371 if (CHECK_FLAG(bgp
->flags
, BGP_FLAG_DELETE_IN_PROGRESS
)) {
2372 vnc_zlog_debug_verbose(
2373 "%s: BGP delete in progress, assume shutdown race condition!!!",
2379 assert(wcb
->import_table
);
2382 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, wcb
->lockoffset
);
2384 vnc_zlog_debug_verbose("%s: removing bpi %p at prefix %pRN", __func__
,
2388 * Remove the route (doubly-linked)
2390 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_VALID
)
2391 && VALID_INTERIOR_TYPE(bpi
->type
))
2392 RFAPI_MONITOR_EXTERIOR(wcb
->node
)->valid_interior_count
--;
2394 p
= agg_node_get_prefix(wcb
->node
);
2395 afi
= family2afi(p
->family
);
2396 wcb
->import_table
->holddown_count
[afi
] -= 1; /* keep count consistent */
2397 rfapiItBiIndexDel(wcb
->node
, bpi
);
2398 rfapiBgpInfoDetach(wcb
->node
, bpi
); /* with removed bpi */
2400 vnc_import_bgp_exterior_del_route_interior(bgp
, wcb
->import_table
,
2405 * If VNC is configured to send response remove messages, AND
2406 * if the removed route had a UN address, do response removal
2409 if (!(bgp
->rfapi_cfg
->flags
2410 & BGP_VNC_CONFIG_RESPONSE_REMOVAL_DISABLE
)) {
2412 int has_valid_duplicate
= 0;
2413 struct bgp_path_info
*bpii
;
2416 * First check if there are any OTHER routes at this node
2417 * that have the same nexthop and a valid UN address. If
2418 * there are (e.g., from other peers), then the route isn't
2419 * really gone, so skip sending a response removal message.
2421 for (bpii
= wcb
->node
->info
; bpii
; bpii
= bpii
->next
) {
2422 if (rfapiVpnBiSamePtUn(bpi
, bpii
)) {
2423 has_valid_duplicate
= 1;
2428 vnc_zlog_debug_verbose("%s: has_valid_duplicate=%d", __func__
,
2429 has_valid_duplicate
);
2431 if (!has_valid_duplicate
) {
2432 rfapiRibPendingDeleteRoute(bgp
, wcb
->import_table
, afi
,
2437 rfapiMonitorEncapDelete(bpi
);
2440 * If there are no VPN monitors at this VPN Node A,
2443 if (!RFAPI_MONITOR_VPN(wcb
->node
)) {
2444 vnc_zlog_debug_verbose("%s: no VPN monitors at this node",
2450 * rfapiMonitorMoveShorter only moves monitors if there are
2451 * no remaining valid routes at the current node
2453 moved
= rfapiMonitorMoveShorter(wcb
->node
, 1);
2456 rfapiMonitorMovedUp(wcb
->import_table
, wcb
->node
, moved
->node
,
2464 rfapiBgpInfoFree(bpi
);
2468 * If route count at this node has gone to 0, withdraw exported prefix
2470 if (!wcb
->node
->info
) {
2471 /* see if the struct rfapi_it_extra is empty and can be freed */
2472 rfapiMonitorExtraPrune(SAFI_MPLS_VPN
, wcb
->node
);
2473 vnc_direct_bgp_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2474 vnc_zebra_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2477 * nexthop change event
2478 * vnc_direct_bgp_add_prefix() will recompute the VN addr
2481 vnc_direct_bgp_add_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2484 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, 1 + wcb
->lockoffset
);
2485 agg_unlock_node(wcb
->node
); /* decr ref count */
2486 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2490 * This works for multiprotocol extension, but not for plain ol'
2491 * unicast IPv4 because that nexthop is stored in attr->nexthop
2493 void rfapiNexthop2Prefix(struct attr
*attr
, struct prefix
*p
)
2498 memset(p
, 0, sizeof(struct prefix
));
2500 switch (p
->family
= BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2502 p
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2503 p
->prefixlen
= IPV4_MAX_BITLEN
;
2507 p
->u
.prefix6
= attr
->mp_nexthop_global
;
2508 p
->prefixlen
= IPV6_MAX_BITLEN
;
2512 vnc_zlog_debug_verbose("%s: Family is unknown = %d", __func__
,
2517 void rfapiUnicastNexthop2Prefix(afi_t afi
, struct attr
*attr
, struct prefix
*p
)
2519 if (afi
== AFI_IP
) {
2520 p
->family
= AF_INET
;
2521 p
->prefixlen
= IPV4_MAX_BITLEN
;
2522 p
->u
.prefix4
= attr
->nexthop
;
2524 rfapiNexthop2Prefix(attr
, p
);
2528 static int rfapiAttrNexthopAddrDifferent(struct prefix
*p1
, struct prefix
*p2
)
2531 vnc_zlog_debug_verbose("%s: p1 or p2 is NULL", __func__
);
2536 * Are address families the same?
2538 if (p1
->family
!= p2
->family
) {
2542 switch (p1
->family
) {
2544 if (IPV4_ADDR_SAME(&p1
->u
.prefix4
, &p2
->u
.prefix4
))
2549 if (IPV6_ADDR_SAME(&p1
->u
.prefix6
, &p2
->u
.prefix6
))
2560 static void rfapiCopyUnEncap2VPN(struct bgp_path_info
*encap_bpi
,
2561 struct bgp_path_info
*vpn_bpi
)
2563 if (!vpn_bpi
|| !vpn_bpi
->extra
) {
2564 zlog_warn("%s: no vpn bpi attr/extra, can't copy UN address",
2569 switch (BGP_MP_NEXTHOP_FAMILY(encap_bpi
->attr
->mp_nexthop_len
)) {
2573 * instrumentation to debug segfault of 091127
2575 vnc_zlog_debug_verbose("%s: vpn_bpi=%p", __func__
, vpn_bpi
);
2576 vnc_zlog_debug_verbose("%s: vpn_bpi->extra=%p", __func__
,
2579 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_INET
;
2580 vpn_bpi
->extra
->vnc
.import
.un
.addr4
=
2581 encap_bpi
->attr
->mp_nexthop_global_in
;
2585 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_INET6
;
2586 vpn_bpi
->extra
->vnc
.import
.un
.addr6
=
2587 encap_bpi
->attr
->mp_nexthop_global
;
2591 zlog_warn("%s: invalid encap nexthop length: %d", __func__
,
2592 encap_bpi
->attr
->mp_nexthop_len
);
2593 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_UNSPEC
;
2599 * returns 0 on success, nonzero on error
2602 rfapiWithdrawEncapUpdateCachedUn(struct rfapi_import_table
*import_table
,
2603 struct bgp_path_info
*encap_bpi
,
2604 struct agg_node
*vpn_rn
,
2605 struct bgp_path_info
*vpn_bpi
)
2610 * clear cached UN address
2612 if (!vpn_bpi
|| !vpn_bpi
->extra
) {
2614 "%s: missing VPN bpi/extra, can't clear UN addr",
2618 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_UNSPEC
;
2619 memset(&vpn_bpi
->extra
->vnc
.import
.un
, 0,
2620 sizeof(vpn_bpi
->extra
->vnc
.import
.un
));
2621 if (CHECK_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
)) {
2622 if (rfapiGetVncTunnelUnAddr(vpn_bpi
->attr
, NULL
)) {
2623 UNSET_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
);
2624 if (VALID_INTERIOR_TYPE(vpn_bpi
->type
))
2625 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2626 ->valid_interior_count
--;
2627 /* signal interior route withdrawal to
2628 * import-exterior */
2629 vnc_import_bgp_exterior_del_route_interior(
2630 bgp_get_default(), import_table
, vpn_rn
,
2637 zlog_warn("%s: missing VPN bpi, can't clear UN addr",
2641 rfapiCopyUnEncap2VPN(encap_bpi
, vpn_bpi
);
2642 if (!CHECK_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
)) {
2643 SET_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
);
2644 if (VALID_INTERIOR_TYPE(vpn_bpi
->type
))
2645 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2646 ->valid_interior_count
++;
2647 /* signal interior route withdrawal to import-exterior
2649 vnc_import_bgp_exterior_add_route_interior(
2650 bgp_get_default(), import_table
, vpn_rn
,
2657 static void rfapiWithdrawTimerEncap(struct thread
*t
)
2659 struct rfapi_withdraw
*wcb
= THREAD_ARG(t
);
2660 struct bgp_path_info
*bpi
= wcb
->info
;
2661 int was_first_route
= 0;
2662 struct rfapi_monitor_encap
*em
;
2663 struct skiplist
*vpn_node_sl
= skiplist_new(0, NULL
, NULL
);
2667 assert(wcb
->import_table
);
2669 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 0);
2671 if (wcb
->node
->info
== bpi
)
2672 was_first_route
= 1;
2675 * Remove the route/bpi and free it
2677 rfapiBgpInfoDetach(wcb
->node
, bpi
);
2678 rfapiBgpInfoFree(bpi
);
2680 if (!was_first_route
)
2683 for (em
= RFAPI_MONITOR_ENCAP(wcb
->node
); em
; em
= em
->next
) {
2686 * Update monitoring VPN BPIs with new encap info at the
2687 * head of the encap bpi chain (which could be NULL after
2688 * removing the expiring bpi above)
2690 if (rfapiWithdrawEncapUpdateCachedUn(wcb
->import_table
,
2691 wcb
->node
->info
, em
->node
,
2696 * Build a list of unique VPN nodes referenced by these
2698 * Use a skiplist for speed.
2700 skiplist_insert(vpn_node_sl
, em
->node
, em
->node
);
2705 * for each VPN node referenced in the ENCAP monitors:
2707 struct agg_node
*rn
;
2708 while (!skiplist_first(vpn_node_sl
, (void **)&rn
, NULL
)) {
2709 if (!wcb
->node
->info
) {
2710 struct rfapi_monitor_vpn
*moved
;
2712 moved
= rfapiMonitorMoveShorter(rn
, 0);
2714 // rfapiDoRouteCallback(wcb->import_table,
2715 // moved->node, moved);
2716 rfapiMonitorMovedUp(wcb
->import_table
, rn
,
2717 moved
->node
, moved
);
2720 // rfapiDoRouteCallback(wcb->import_table, rn, NULL);
2721 rfapiMonitorItNodeChanged(wcb
->import_table
, rn
, NULL
);
2723 skiplist_delete_first(vpn_node_sl
);
2727 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 1);
2728 agg_unlock_node(wcb
->node
); /* decr ref count */
2729 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2730 skiplist_free(vpn_node_sl
);
2735 * Works for both VPN and ENCAP routes; timer_service_func is different
2739 rfapiBiStartWithdrawTimer(struct rfapi_import_table
*import_table
,
2740 struct agg_node
*rn
, struct bgp_path_info
*bpi
,
2741 afi_t afi
, safi_t safi
,
2742 void (*timer_service_func
)(struct thread
*))
2745 struct rfapi_withdraw
*wcb
;
2747 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
2749 * Already on the path to being withdrawn,
2750 * should already have a timer set up to
2753 vnc_zlog_debug_verbose(
2754 "%s: already being withdrawn, do nothing", __func__
);
2758 rfapiGetVncLifetime(bpi
->attr
, &lifetime
);
2759 vnc_zlog_debug_verbose("%s: VNC lifetime is %u", __func__
, lifetime
);
2762 * withdrawn routes get to hang around for a while
2764 SET_FLAG(bpi
->flags
, BGP_PATH_REMOVED
);
2766 /* set timer to remove the route later */
2767 lifetime
= rfapiGetHolddownFromLifetime(lifetime
);
2768 vnc_zlog_debug_verbose("%s: using timeout %u", __func__
, lifetime
);
2771 * Stash import_table, node, and info for use by timer
2772 * service routine, which is supposed to free the wcb.
2774 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2777 wcb
->import_table
= import_table
;
2778 bgp_attr_intern(bpi
->attr
);
2780 if (VNC_DEBUG(VERBOSE
)) {
2781 vnc_zlog_debug_verbose(
2782 "%s: wcb values: node=%p, info=%p, import_table=%p (bpi follows)",
2783 __func__
, wcb
->node
, wcb
->info
, wcb
->import_table
);
2784 rfapiPrintBi(NULL
, bpi
);
2789 if (lifetime
> UINT32_MAX
/ 1001) {
2790 /* sub-optimal case, but will probably never happen */
2791 bpi
->extra
->vnc
.import
.timer
= NULL
;
2792 thread_add_timer(bm
->master
, timer_service_func
, wcb
, lifetime
,
2793 &bpi
->extra
->vnc
.import
.timer
);
2795 static uint32_t jitter
;
2796 uint32_t lifetime_msec
;
2799 * the goal here is to spread out the timers so they are
2800 * sortable in the skip list
2802 if (++jitter
>= 1000)
2805 lifetime_msec
= (lifetime
* 1000) + jitter
;
2807 bpi
->extra
->vnc
.import
.timer
= NULL
;
2808 thread_add_timer_msec(bm
->master
, timer_service_func
, wcb
,
2810 &bpi
->extra
->vnc
.import
.timer
);
2813 /* re-sort route list (BGP_PATH_REMOVED routes are last) */
2814 if (((struct bgp_path_info
*)rn
->info
)->next
) {
2815 rfapiBgpInfoDetach(rn
, bpi
);
2816 rfapiBgpInfoAttachSorted(rn
, bpi
, afi
, safi
);
2821 typedef void(rfapi_bi_filtered_import_f
)(struct rfapi_import_table
*table
,
2822 int action
, struct peer
*peer
,
2823 void *rfd
, const struct prefix
*prefix
,
2824 const struct prefix
*aux_prefix
,
2825 afi_t afi
, struct prefix_rd
*prd
,
2826 struct attr
*attr
, uint8_t type
,
2827 uint8_t sub_type
, uint32_t *label
);
2830 static void rfapiExpireEncapNow(struct rfapi_import_table
*it
,
2831 struct agg_node
*rn
, struct bgp_path_info
*bpi
)
2833 struct rfapi_withdraw
*wcb
;
2837 * pretend we're an expiring timer
2839 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2842 wcb
->import_table
= it
;
2843 memset(&t
, 0, sizeof(t
));
2845 rfapiWithdrawTimerEncap(&t
); /* frees wcb */
2848 static int rfapiGetNexthop(struct attr
*attr
, struct prefix
*prefix
)
2850 switch (BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2852 prefix
->family
= AF_INET
;
2853 prefix
->prefixlen
= IPV4_MAX_BITLEN
;
2854 prefix
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2857 prefix
->family
= AF_INET6
;
2858 prefix
->prefixlen
= IPV6_MAX_BITLEN
;
2859 prefix
->u
.prefix6
= attr
->mp_nexthop_global
;
2862 vnc_zlog_debug_verbose("%s: unknown attr->mp_nexthop_len %d",
2863 __func__
, attr
->mp_nexthop_len
);
2870 * import a bgp_path_info if its route target list intersects with the
2871 * import table's route target list
2873 static void rfapiBgpInfoFilteredImportEncap(
2874 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
2875 void *rfd
, /* set for looped back routes */
2876 const struct prefix
*p
,
2877 const struct prefix
*aux_prefix
, /* Unused for encap routes */
2878 afi_t afi
, struct prefix_rd
*prd
,
2879 struct attr
*attr
, /* part of bgp_path_info */
2880 uint8_t type
, /* part of bgp_path_info */
2881 uint8_t sub_type
, /* part of bgp_path_info */
2882 uint32_t *label
) /* part of bgp_path_info */
2884 struct agg_table
*rt
= NULL
;
2885 struct agg_node
*rn
;
2886 struct bgp_path_info
*info_new
;
2887 struct bgp_path_info
*bpi
;
2888 struct bgp_path_info
*next
;
2891 struct prefix p_firstbpi_old
;
2892 struct prefix p_firstbpi_new
;
2894 const char *action_str
= NULL
;
2895 struct prefix un_prefix
;
2898 bgp
= bgp_get_default(); /* assume 1 instance for now */
2901 case FIF_ACTION_UPDATE
:
2902 action_str
= "update";
2904 case FIF_ACTION_WITHDRAW
:
2905 action_str
= "withdraw";
2907 case FIF_ACTION_KILL
:
2908 action_str
= "kill";
2915 vnc_zlog_debug_verbose(
2916 "%s: entry: %s: prefix %s/%d", __func__
, action_str
,
2917 inet_ntop(p
->family
, &p
->u
.prefix
, buf
, sizeof(buf
)),
2920 memset(&p_firstbpi_old
, 0, sizeof(p_firstbpi_old
));
2921 memset(&p_firstbpi_new
, 0, sizeof(p_firstbpi_new
));
2923 if (action
== FIF_ACTION_UPDATE
) {
2925 * Compare rt lists. If no intersection, don't import this route
2926 * On a withdraw, peer and RD are sufficient to determine if
2929 if (!attr
|| !bgp_attr_get_ecommunity(attr
)) {
2931 vnc_zlog_debug_verbose(
2932 "%s: attr, extra, or ecommunity missing, not importing",
2936 #ifdef RFAPI_REQUIRE_ENCAP_BEEC
2937 if (!rfapiEcommunitiesMatchBeec(
2938 bgp_attr_get_ecommunity(attr
))) {
2939 vnc_zlog_debug_verbose(
2940 "%s: it=%p: no match for BGP Encapsulation ecommunity",
2941 __func__
, import_table
);
2945 if (!rfapiEcommunitiesIntersect(
2946 import_table
->rt_import_list
,
2947 bgp_attr_get_ecommunity(attr
))) {
2949 vnc_zlog_debug_verbose(
2950 "%s: it=%p: no ecommunity intersection",
2951 __func__
, import_table
);
2956 * Updates must also have a nexthop address
2958 memset(&un_prefix
, 0,
2959 sizeof(un_prefix
)); /* keep valgrind happy */
2960 if (rfapiGetNexthop(attr
, &un_prefix
)) {
2961 vnc_zlog_debug_verbose("%s: missing nexthop address",
2968 * Figure out which radix tree the route would go into
2973 rt
= import_table
->imported_encap
[afi
];
2979 flog_err(EC_LIB_DEVELOPMENT
, "%s: bad afi %d", __func__
, afi
);
2984 * agg_node_lookup returns a node only if there is at least
2985 * one route attached.
2987 rn
= agg_node_lookup(rt
, p
);
2989 #ifdef DEBUG_ENCAP_MONITOR
2990 vnc_zlog_debug_verbose("%s: initial encap lookup(it=%p) rn=%p",
2991 __func__
, import_table
, rn
);
2996 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 1);
2997 agg_unlock_node(rn
); /* undo lock in agg_node_lookup */
3001 * capture nexthop of first bpi
3004 rfapiNexthop2Prefix(
3005 ((struct bgp_path_info
*)(rn
->info
))->attr
,
3009 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
3012 * Does this bgp_path_info refer to the same route
3013 * as we are trying to add?
3015 vnc_zlog_debug_verbose("%s: comparing BPI %p", __func__
,
3022 * RD of import table bpi is in
3023 * bpi->extra->vnc.import.rd RD of info_orig is in prd
3026 vnc_zlog_debug_verbose("%s: no bpi->extra",
3031 (struct prefix
*)&bpi
->extra
->vnc
.import
.rd
,
3032 (struct prefix
*)prd
)) {
3034 vnc_zlog_debug_verbose("%s: prd does not match",
3042 if (bpi
->peer
!= peer
) {
3043 vnc_zlog_debug_verbose(
3044 "%s: peer does not match", __func__
);
3048 vnc_zlog_debug_verbose("%s: found matching bpi",
3051 /* Same route. Delete this bpi, replace with new one */
3053 if (action
== FIF_ACTION_WITHDRAW
) {
3055 vnc_zlog_debug_verbose(
3056 "%s: withdrawing at prefix %pRN",
3059 rfapiBiStartWithdrawTimer(
3060 import_table
, rn
, bpi
, afi
, SAFI_ENCAP
,
3061 rfapiWithdrawTimerEncap
);
3064 vnc_zlog_debug_verbose(
3065 "%s: %s at prefix %pRN", __func__
,
3066 ((action
== FIF_ACTION_KILL
)
3072 * If this route is waiting to be deleted
3074 * a previous withdraw, we must cancel its
3077 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
3078 && bpi
->extra
->vnc
.import
.timer
) {
3079 struct rfapi_withdraw
*wcb
= THREAD_ARG(
3080 bpi
->extra
->vnc
.import
.timer
);
3082 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3084 bpi
->extra
->vnc
.import
.timer
);
3087 if (action
== FIF_ACTION_UPDATE
) {
3088 rfapiBgpInfoDetach(rn
, bpi
);
3089 rfapiBgpInfoFree(bpi
);
3093 * Kill: do export stuff when removing
3096 struct rfapi_withdraw
*wcb
;
3100 * pretend we're an expiring timer
3103 MTYPE_RFAPI_WITHDRAW
,
3104 sizeof(struct rfapi_withdraw
));
3107 wcb
->import_table
= import_table
;
3108 memset(&t
, 0, sizeof(t
));
3110 rfapiWithdrawTimerEncap(
3111 &t
); /* frees wcb */
3120 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, replacing
? 1 : 0);
3122 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
)
3126 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, NULL
);
3130 agg_lock_node(rn
); /* incr ref count for new BPI */
3132 rn
= agg_node_get(rt
, p
);
3135 vnc_zlog_debug_verbose("%s: (afi=%d, rn=%p) inserting at prefix %pRN",
3136 __func__
, afi
, rn
, rn
);
3138 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_ENCAP
);
3141 * Delete holddown routes from same NVE. See details in
3142 * rfapiBgpInfoFilteredImportVPN()
3144 for (bpi
= info_new
->next
; bpi
; bpi
= next
) {
3146 struct prefix pfx_un
;
3150 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
))
3154 * We already match the VN address (it is the prefix
3155 * of the route node)
3158 if (!rfapiGetNexthop(bpi
->attr
, &pfx_un
)
3159 && prefix_same(&pfx_un
, &un_prefix
)) {
3167 vnc_zlog_debug_verbose(
3168 "%s: removing holddown bpi matching NVE of new route",
3170 if (bpi
->extra
->vnc
.import
.timer
) {
3171 struct rfapi_withdraw
*wcb
=
3172 THREAD_ARG(bpi
->extra
->vnc
.import
.timer
);
3174 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3175 THREAD_OFF(bpi
->extra
->vnc
.import
.timer
);
3177 rfapiExpireEncapNow(import_table
, rn
, bpi
);
3180 rfapiNexthop2Prefix(((struct bgp_path_info
*)(rn
->info
))->attr
,
3184 * If the nexthop address of the selected Encap route (i.e.,
3185 * the UN address) has changed, then we must update the VPN
3186 * routes that refer to this Encap route and possibly force
3189 if (rfapiAttrNexthopAddrDifferent(&p_firstbpi_old
, &p_firstbpi_new
)) {
3191 struct rfapi_monitor_encap
*m
;
3192 struct rfapi_monitor_encap
*mnext
;
3194 struct agg_node
*referenced_vpn_prefix
;
3197 * Optimized approach: build radix tree on the fly to
3198 * hold list of VPN nodes referenced by the ENCAP monitors
3200 * The nodes in this table correspond to prefixes of VPN routes.
3201 * The "info" pointer of the node points to a chain of
3202 * struct rfapi_monitor_encap, each of which refers to a
3203 * specific VPN node.
3205 struct agg_table
*referenced_vpn_table
;
3207 referenced_vpn_table
= agg_table_init();
3210 * iterate over the set of monitors at this ENCAP node.
3212 #ifdef DEBUG_ENCAP_MONITOR
3213 vnc_zlog_debug_verbose("%s: examining monitors at rn=%p",
3216 for (m
= RFAPI_MONITOR_ENCAP(rn
); m
; m
= m
->next
) {
3217 const struct prefix
*p
;
3220 * For each referenced bpi/route, copy the ENCAP route's
3221 * nexthop to the VPN route's cached UN address field
3223 * the address family of the cached UN address field.
3225 rfapiCopyUnEncap2VPN(info_new
, m
->bpi
);
3226 if (!CHECK_FLAG(m
->bpi
->flags
, BGP_PATH_VALID
)) {
3227 SET_FLAG(m
->bpi
->flags
, BGP_PATH_VALID
);
3228 if (VALID_INTERIOR_TYPE(m
->bpi
->type
))
3229 RFAPI_MONITOR_EXTERIOR(m
->node
)
3230 ->valid_interior_count
++;
3231 vnc_import_bgp_exterior_add_route_interior(
3232 bgp
, import_table
, m
->node
, m
->bpi
);
3236 * Build a list of unique VPN nodes referenced by these
3239 * There could be more than one VPN node here with a
3241 * prefix. Those are currently in an unsorted linear
3245 p
= agg_node_get_prefix(m
->node
);
3246 referenced_vpn_prefix
=
3247 agg_node_get(referenced_vpn_table
, p
);
3248 assert(referenced_vpn_prefix
);
3249 for (mnext
= referenced_vpn_prefix
->info
; mnext
;
3250 mnext
= mnext
->next
) {
3252 if (mnext
->node
== m
->node
)
3258 * already have an entry for this VPN node
3260 agg_unlock_node(referenced_vpn_prefix
);
3263 MTYPE_RFAPI_MONITOR_ENCAP
,
3264 sizeof(struct rfapi_monitor_encap
));
3265 mnext
->node
= m
->node
;
3266 mnext
->next
= referenced_vpn_prefix
->info
;
3267 referenced_vpn_prefix
->info
= mnext
;
3272 * for each VPN node referenced in the ENCAP monitors:
3274 for (referenced_vpn_prefix
=
3275 agg_route_top(referenced_vpn_table
);
3276 referenced_vpn_prefix
;
3277 referenced_vpn_prefix
=
3278 agg_route_next(referenced_vpn_prefix
)) {
3280 while ((m
= referenced_vpn_prefix
->info
)) {
3284 rfapiMonitorMoveLonger(m
->node
);
3285 for (n
= m
->node
; n
; n
= agg_node_parent(n
)) {
3286 // rfapiDoRouteCallback(import_table, n,
3289 rfapiMonitorItNodeChanged(import_table
, m
->node
,
3292 referenced_vpn_prefix
->info
= m
->next
;
3293 agg_unlock_node(referenced_vpn_prefix
);
3294 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, m
);
3297 agg_table_finish(referenced_vpn_table
);
3300 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
3303 static void rfapiExpireVpnNow(struct rfapi_import_table
*it
,
3304 struct agg_node
*rn
, struct bgp_path_info
*bpi
,
3307 struct rfapi_withdraw
*wcb
;
3311 * pretend we're an expiring timer
3313 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
3316 wcb
->import_table
= it
;
3317 wcb
->lockoffset
= lockoffset
;
3318 memset(&t
, 0, sizeof(t
));
3320 rfapiWithdrawTimerVPN(&t
); /* frees wcb */
3325 * import a bgp_path_info if its route target list intersects with the
3326 * import table's route target list
3328 void rfapiBgpInfoFilteredImportVPN(
3329 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3330 void *rfd
, /* set for looped back routes */
3331 const struct prefix
*p
,
3332 const struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3333 afi_t afi
, struct prefix_rd
*prd
,
3334 struct attr
*attr
, /* part of bgp_path_info */
3335 uint8_t type
, /* part of bgp_path_info */
3336 uint8_t sub_type
, /* part of bgp_path_info */
3337 uint32_t *label
) /* part of bgp_path_info */
3339 struct agg_table
*rt
= NULL
;
3340 struct agg_node
*rn
;
3342 struct bgp_path_info
*info_new
;
3343 struct bgp_path_info
*bpi
;
3344 struct bgp_path_info
*next
;
3346 struct prefix vn_prefix
;
3347 struct prefix un_prefix
;
3348 int un_prefix_valid
= 0;
3349 struct agg_node
*ern
;
3351 int original_had_routes
= 0;
3352 struct prefix original_nexthop
;
3353 const char *action_str
= NULL
;
3357 bgp
= bgp_get_default(); /* assume 1 instance for now */
3360 case FIF_ACTION_UPDATE
:
3361 action_str
= "update";
3363 case FIF_ACTION_WITHDRAW
:
3364 action_str
= "withdraw";
3366 case FIF_ACTION_KILL
:
3367 action_str
= "kill";
3374 if (import_table
== bgp
->rfapi
->it_ce
)
3377 vnc_zlog_debug_verbose("%s: entry: %s%s: prefix %s/%d: it %p, afi %s",
3378 __func__
, (is_it_ce
? "CE-IT " : ""), action_str
,
3379 rfapi_ntop(p
->family
, &p
->u
.prefix
, buf
, BUFSIZ
),
3380 p
->prefixlen
, import_table
, afi2str(afi
));
3385 * Compare rt lists. If no intersection, don't import this route
3386 * On a withdraw, peer and RD are sufficient to determine if
3389 if (action
== FIF_ACTION_UPDATE
) {
3390 if (!attr
|| !bgp_attr_get_ecommunity(attr
)) {
3392 vnc_zlog_debug_verbose(
3393 "%s: attr, extra, or ecommunity missing, not importing",
3397 if ((import_table
!= bgp
->rfapi
->it_ce
) &&
3398 !rfapiEcommunitiesIntersect(
3399 import_table
->rt_import_list
,
3400 bgp_attr_get_ecommunity(attr
))) {
3402 vnc_zlog_debug_verbose(
3403 "%s: it=%p: no ecommunity intersection",
3404 __func__
, import_table
);
3408 memset(&vn_prefix
, 0,
3409 sizeof(vn_prefix
)); /* keep valgrind happy */
3410 if (rfapiGetNexthop(attr
, &vn_prefix
)) {
3411 /* missing nexthop address would be a bad, bad thing */
3412 vnc_zlog_debug_verbose("%s: missing nexthop", __func__
);
3418 * Figure out which radix tree the route would go into
3424 rt
= import_table
->imported_vpn
[afi
];
3429 flog_err(EC_LIB_DEVELOPMENT
, "%s: bad afi %d", __func__
, afi
);
3434 memset(&original_nexthop
, 0, sizeof(original_nexthop
));
3437 * agg_node_lookup returns a node only if there is at least
3438 * one route attached.
3440 rn
= agg_node_lookup(rt
, p
);
3442 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
3446 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
3447 agg_unlock_node(rn
); /* undo lock in agg_node_lookup */
3450 original_had_routes
= 1;
3452 if (VNC_DEBUG(VERBOSE
)) {
3453 vnc_zlog_debug_verbose("%s: showing IT node on entry",
3455 rfapiShowItNode(NULL
, rn
); /* debug */
3459 * Look for same route (will have same RD and peer)
3461 bpi
= rfapiItBiIndexSearch(rn
, prd
, peer
, aux_prefix
);
3466 * This was an old test when we iterated over the
3467 * BPIs linearly. Since we're now looking up with
3468 * RD and peer, comparing types should not be
3469 * needed. Changed to assertion.
3471 * Compare types. Doing so prevents a RFP-originated
3472 * route from matching an imported route, for example.
3474 if (VNC_DEBUG(VERBOSE
) && bpi
->type
!= type
)
3475 /* should be handled by RDs, but warn for now */
3476 zlog_warn("%s: type mismatch! (bpi=%d, arg=%d)",
3477 __func__
, bpi
->type
, type
);
3479 vnc_zlog_debug_verbose("%s: found matching bpi",
3483 * In the special CE table, withdrawals occur without
3486 if (import_table
== bgp
->rfapi
->it_ce
) {
3487 vnc_direct_bgp_del_route_ce(bgp
, rn
, bpi
);
3488 if (action
== FIF_ACTION_WITHDRAW
)
3489 action
= FIF_ACTION_KILL
;
3492 if (action
== FIF_ACTION_WITHDRAW
) {
3494 int washolddown
= CHECK_FLAG(bpi
->flags
,
3497 vnc_zlog_debug_verbose(
3498 "%s: withdrawing at prefix %pRN%s",
3501 ? " (already being withdrawn)"
3506 rfapiBiStartWithdrawTimer(
3507 import_table
, rn
, bpi
, afi
,
3509 rfapiWithdrawTimerVPN
);
3511 RFAPI_UPDATE_ITABLE_COUNT(
3512 bpi
, import_table
, afi
, -1);
3513 import_table
->holddown_count
[afi
] += 1;
3517 vnc_zlog_debug_verbose(
3518 "%s: %s at prefix %pRN", __func__
,
3519 ((action
== FIF_ACTION_KILL
)
3525 * If this route is waiting to be deleted
3527 * a previous withdraw, we must cancel its
3530 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
3531 && bpi
->extra
->vnc
.import
.timer
) {
3532 struct rfapi_withdraw
*wcb
= THREAD_ARG(
3533 bpi
->extra
->vnc
.import
.timer
);
3535 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3537 bpi
->extra
->vnc
.import
.timer
);
3539 import_table
->holddown_count
[afi
] -= 1;
3540 RFAPI_UPDATE_ITABLE_COUNT(
3541 bpi
, import_table
, afi
, 1);
3544 * decrement remote count (if route is remote)
3546 * we are going to remove it below
3548 RFAPI_UPDATE_ITABLE_COUNT(bpi
, import_table
,
3550 if (action
== FIF_ACTION_UPDATE
) {
3554 * make copy of original nexthop so we
3555 * can see if it changed
3557 rfapiGetNexthop(bpi
->attr
,
3561 * remove bpi without doing any export
3564 if (CHECK_FLAG(bpi
->flags
,
3566 && VALID_INTERIOR_TYPE(bpi
->type
))
3567 RFAPI_MONITOR_EXTERIOR(rn
)
3568 ->valid_interior_count
--;
3569 rfapiItBiIndexDel(rn
, bpi
);
3570 rfapiBgpInfoDetach(rn
, bpi
);
3571 rfapiMonitorEncapDelete(bpi
);
3572 vnc_import_bgp_exterior_del_route_interior(
3573 bgp
, import_table
, rn
, bpi
);
3574 rfapiBgpInfoFree(bpi
);
3578 * remove bpi and do export processing
3580 import_table
->holddown_count
[afi
] += 1;
3581 rfapiExpireVpnNow(import_table
, rn
, bpi
,
3589 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, replacing
? 1 : 0);
3591 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
) {
3597 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, label
);
3600 * lookup un address in encap table
3602 ern
= agg_node_match(import_table
->imported_encap
[afi
], &vn_prefix
);
3604 rfapiCopyUnEncap2VPN(ern
->info
, info_new
);
3605 agg_unlock_node(ern
); /* undo lock in route_note_match */
3607 /* Not a big deal, just means VPN route got here first */
3608 vnc_zlog_debug_verbose("%s: no encap route for vn addr %pFX",
3609 __func__
, &vn_prefix
);
3610 info_new
->extra
->vnc
.import
.un_family
= AF_UNSPEC
;
3618 * No need to increment reference count, so only "get"
3619 * if the node is not there already
3621 rn
= agg_node_get(rt
, p
);
3625 * For ethernet routes, if there is an accompanying IP address,
3626 * save it in the bpi
3628 if ((AFI_L2VPN
== afi
) && aux_prefix
) {
3630 vnc_zlog_debug_verbose("%s: setting BPI's aux_prefix",
3632 info_new
->extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
3635 vnc_zlog_debug_verbose("%s: inserting bpi %p at prefix %pRN #%d",
3636 __func__
, info_new
, rn
,
3637 agg_node_get_lock_count(rn
));
3639 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_MPLS_VPN
);
3640 rfapiItBiIndexAdd(rn
, info_new
);
3641 if (!rfapiGetUnAddrOfVpnBi(info_new
, NULL
)) {
3642 if (VALID_INTERIOR_TYPE(info_new
->type
))
3643 RFAPI_MONITOR_EXTERIOR(rn
)->valid_interior_count
++;
3644 SET_FLAG(info_new
->flags
, BGP_PATH_VALID
);
3646 RFAPI_UPDATE_ITABLE_COUNT(info_new
, import_table
, afi
, 1);
3647 vnc_import_bgp_exterior_add_route_interior(bgp
, import_table
, rn
,
3650 if (import_table
== bgp
->rfapi
->it_ce
)
3651 vnc_direct_bgp_add_route_ce(bgp
, rn
, info_new
);
3653 if (VNC_DEBUG(VERBOSE
)) {
3654 vnc_zlog_debug_verbose("%s: showing IT node", __func__
);
3655 rfapiShowItNode(NULL
, rn
); /* debug */
3658 rfapiMonitorEncapAdd(import_table
, &vn_prefix
, rn
, info_new
);
3660 if (!rfapiGetUnAddrOfVpnBi(info_new
, &un_prefix
)) {
3663 * if we have a valid UN address (either via Encap route
3664 * or via tunnel attribute), then we should attempt
3665 * to move any monitors at less-specific nodes to this node
3667 rfapiMonitorMoveLonger(rn
);
3669 un_prefix_valid
= 1;
3673 * 101129 Enhancement: if we add a route (implication: it is not
3674 * in holddown), delete all other routes from this nve at this
3675 * node that are in holddown, regardless of peer.
3677 * Reasons it's OK to do that:
3679 * - if the holddown route being deleted originally came from BGP VPN,
3680 * it is already gone from BGP (implication of holddown), so there
3681 * won't be any added inconsistency with the BGP RIB.
3683 * - once a fresh route is added at a prefix, any routes in holddown
3684 * at that prefix will not show up in RFP responses, so deleting
3685 * the holddown routes won't affect the contents of responses.
3687 * - lifetimes are supposed to be consistent, so there should not
3688 * be a case where the fresh route has a shorter lifetime than
3689 * the holddown route, so we don't expect the fresh route to
3690 * disappear and complete its holddown time before the existing
3691 * holddown routes time out. Therefore, we won't have a situation
3692 * where we expect the existing holddown routes to be hidden and
3693 * then to reappear sometime later (as holddown routes) in a
3696 * Among other things, this would enable us to skirt the problem
3697 * of local holddown routes that refer to NVE descriptors that
3698 * have already been closed (if the same NVE triggers a subsequent
3699 * rfapi_open(), the new peer is different and doesn't match the
3700 * peer of the holddown route, so the stale holddown route still
3701 * hangs around until it times out instead of just being replaced
3702 * by the fresh route).
3705 * We know that the new bpi will have been inserted before any routes
3706 * in holddown, so we can skip any that came before it
3708 for (bpi
= info_new
->next
; bpi
; bpi
= next
) {
3710 struct prefix pfx_vn
;
3711 struct prefix pfx_un
;
3713 int remote_peer_match
= 0;
3720 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
))
3724 * Must match VN address (nexthop of VPN route)
3726 if (rfapiGetNexthop(bpi
->attr
, &pfx_vn
))
3728 if (!prefix_same(&pfx_vn
, &vn_prefix
))
3731 if (un_prefix_valid
&& /* new route UN addr */
3732 !rfapiGetUnAddrOfVpnBi(bpi
, &pfx_un
)
3733 && /* old route UN addr */
3734 prefix_same(&pfx_un
, &un_prefix
)) { /* compare */
3737 if (!RFAPI_LOCAL_BI(bpi
) && !RFAPI_LOCAL_BI(info_new
)
3738 && sockunion_same(&bpi
->peer
->su
, &info_new
->peer
->su
)) {
3739 /* old & new are both remote, same peer */
3740 remote_peer_match
= 1;
3743 if (!un_match
&& !remote_peer_match
)
3746 vnc_zlog_debug_verbose(
3747 "%s: removing holddown bpi matching NVE of new route",
3749 if (bpi
->extra
->vnc
.import
.timer
) {
3750 struct rfapi_withdraw
*wcb
=
3751 THREAD_ARG(bpi
->extra
->vnc
.import
.timer
);
3753 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3754 THREAD_OFF(bpi
->extra
->vnc
.import
.timer
);
3756 rfapiExpireVpnNow(import_table
, rn
, bpi
, 0);
3759 if (!original_had_routes
) {
3761 * We went from 0 usable routes to 1 usable route. Perform the
3762 * "Adding a Route" export process.
3764 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3765 vnc_zebra_add_prefix(bgp
, import_table
, rn
);
3768 * Check for nexthop change event
3769 * Note: the prefix_same() test below detects two situations:
3770 * 1. route is replaced, new route has different nexthop
3771 * 2. new route is added (original_nexthop is 0)
3773 struct prefix new_nexthop
;
3775 rfapiGetNexthop(attr
, &new_nexthop
);
3776 if (!prefix_same(&original_nexthop
, &new_nexthop
)) {
3778 * nexthop change event
3779 * vnc_direct_bgp_add_prefix() will recompute VN addr
3782 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3786 if (!(bgp
->rfapi_cfg
->flags
& BGP_VNC_CONFIG_CALLBACK_DISABLE
)) {
3787 for (n
= rn
; n
; n
= agg_node_parent(n
)) {
3788 // rfapiDoRouteCallback(import_table, n, NULL);
3790 rfapiMonitorItNodeChanged(import_table
, rn
, NULL
);
3792 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 0);
3796 static void rfapiBgpInfoFilteredImportBadSafi(
3797 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3798 void *rfd
, /* set for looped back routes */
3799 const struct prefix
*p
,
3800 const struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3801 afi_t afi
, struct prefix_rd
*prd
,
3802 struct attr
*attr
, /* part of bgp_path_info */
3803 uint8_t type
, /* part of bgp_path_info */
3804 uint8_t sub_type
, /* part of bgp_path_info */
3805 uint32_t *label
) /* part of bgp_path_info */
3807 vnc_zlog_debug_verbose("%s: Error, bad safi", __func__
);
3810 static rfapi_bi_filtered_import_f
*
3811 rfapiBgpInfoFilteredImportFunction(safi_t safi
)
3815 return rfapiBgpInfoFilteredImportVPN
;
3818 return rfapiBgpInfoFilteredImportEncap
;
3822 case SAFI_MULTICAST
:
3824 case SAFI_LABELED_UNICAST
:
3828 flog_err(EC_LIB_DEVELOPMENT
, "%s: bad safi %d", __func__
, safi
);
3829 return rfapiBgpInfoFilteredImportBadSafi
;
3832 assert(!"Reached end of function when we were not expecting to");
3835 void rfapiProcessUpdate(struct peer
*peer
,
3836 void *rfd
, /* set when looped from RFP/RFAPI */
3837 const struct prefix
*p
, struct prefix_rd
*prd
,
3838 struct attr
*attr
, afi_t afi
, safi_t safi
, uint8_t type
,
3839 uint8_t sub_type
, uint32_t *label
)
3843 struct rfapi_import_table
*it
;
3844 int has_ip_route
= 1;
3847 bgp
= bgp_get_default(); /* assume 1 instance for now */
3854 * look at high-order byte of RD. FF means MAC
3855 * address is present (VNC L2VPN)
3857 if ((safi
== SAFI_MPLS_VPN
)
3858 && (decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
)) {
3859 struct prefix pfx_mac_buf
;
3860 struct prefix pfx_nexthop_buf
;
3864 * Set flag if prefix and nexthop are the same - don't
3865 * add the route to normal IP-based import tables
3867 if (!rfapiGetNexthop(attr
, &pfx_nexthop_buf
)) {
3868 if (!prefix_cmp(&pfx_nexthop_buf
, p
)) {
3873 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
3874 pfx_mac_buf
.family
= AF_ETHERNET
;
3875 pfx_mac_buf
.prefixlen
= 48;
3876 memcpy(&pfx_mac_buf
.u
.prefix_eth
.octet
, prd
->val
+ 2, 6);
3879 * Find rt containing LNI (Logical Network ID), which
3880 * _should_ always be present when mac address is present
3882 rc
= rfapiEcommunityGetLNI(bgp_attr_get_ecommunity(attr
), &lni
);
3884 vnc_zlog_debug_verbose(
3885 "%s: rfapiEcommunityGetLNI returned %d, lni=%d, attr=%p",
3886 __func__
, rc
, lni
, attr
);
3888 it
= rfapiMacImportTableGet(bgp
, lni
);
3890 rfapiBgpInfoFilteredImportVPN(
3891 it
, FIF_ACTION_UPDATE
, peer
, rfd
,
3892 &pfx_mac_buf
, /* prefix */
3893 p
, /* aux prefix: IP addr */
3894 AFI_L2VPN
, prd
, attr
, type
, sub_type
, label
);
3902 * Iterate over all import tables; do a filtered import
3903 * for the afi/safi combination
3905 for (it
= h
->imports
; it
; it
= it
->next
) {
3906 (*rfapiBgpInfoFilteredImportFunction(safi
))(
3907 it
, FIF_ACTION_UPDATE
, peer
, rfd
, p
, /* prefix */
3908 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3911 if (safi
== SAFI_MPLS_VPN
) {
3912 vnc_direct_bgp_rh_add_route(bgp
, afi
, p
, peer
, attr
);
3913 rfapiBgpInfoFilteredImportVPN(
3914 bgp
->rfapi
->it_ce
, FIF_ACTION_UPDATE
, peer
, rfd
,
3916 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3921 void rfapiProcessWithdraw(struct peer
*peer
, void *rfd
, const struct prefix
*p
,
3922 struct prefix_rd
*prd
, struct attr
*attr
, afi_t afi
,
3923 safi_t safi
, uint8_t type
, int kill
)
3927 struct rfapi_import_table
*it
;
3929 bgp
= bgp_get_default(); /* assume 1 instance for now */
3936 * look at high-order byte of RD. FF means MAC
3937 * address is present (VNC L2VPN)
3939 if (h
->import_mac
!= NULL
&& safi
== SAFI_MPLS_VPN
3940 && decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
) {
3941 struct prefix pfx_mac_buf
;
3942 void *cursor
= NULL
;
3945 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
3946 pfx_mac_buf
.family
= AF_ETHERNET
;
3947 pfx_mac_buf
.prefixlen
= 48;
3948 memcpy(&pfx_mac_buf
.u
.prefix_eth
, prd
->val
+ 2, 6);
3951 * withdraw does not contain attrs, so we don't have
3952 * access to the route's LNI, which would ordinarily
3953 * select the specific mac-based import table. Instead,
3954 * we must iterate over all mac-based tables and rely
3955 * on the RD to match.
3957 * If this approach is too slow, add an index where
3958 * key is {RD, peer} and value is the import table
3960 for (rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
3962 rc
== 0; rc
= skiplist_next(h
->import_mac
, NULL
,
3963 (void **)&it
, &cursor
)) {
3965 #ifdef DEBUG_L2_EXTRA
3966 vnc_zlog_debug_verbose(
3967 "%s: calling rfapiBgpInfoFilteredImportVPN(it=%p, afi=AFI_L2VPN)",
3971 rfapiBgpInfoFilteredImportVPN(
3973 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
3974 peer
, rfd
, &pfx_mac_buf
, /* prefix */
3975 p
, /* aux_prefix: IP */
3976 AFI_L2VPN
, prd
, attr
, type
, 0,
3977 NULL
); /* sub_type & label unused for withdraw
3983 * XXX For the case where the withdraw involves an L2
3984 * route with no IP information, we rely on the lack
3985 * of RT-list intersection to filter out the withdraw
3986 * from the IP-based import tables below
3990 * Iterate over all import tables; do a filtered import
3991 * for the afi/safi combination
3994 for (it
= h
->imports
; it
; it
= it
->next
) {
3995 (*rfapiBgpInfoFilteredImportFunction(safi
))(
3996 it
, (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
3997 peer
, rfd
, p
, /* prefix */
3998 NULL
, afi
, prd
, attr
, type
, 0,
3999 NULL
); /* sub_type & label unused for withdraw */
4002 /* TBD the deletion should happen after the lifetime expires */
4003 if (safi
== SAFI_MPLS_VPN
)
4004 vnc_direct_bgp_rh_del_route(bgp
, afi
, p
, peer
);
4006 if (safi
== SAFI_MPLS_VPN
) {
4007 rfapiBgpInfoFilteredImportVPN(
4009 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
), peer
,
4010 rfd
, p
, /* prefix */
4011 NULL
, afi
, prd
, attr
, type
, 0,
4012 NULL
); /* sub_type & label unused for withdraw */
4017 * TBD optimized withdraw timer algorithm for case of many
4018 * routes expiring at the same time due to peer drop.
4021 * 1. Visit all BPIs in all ENCAP import tables.
4023 * a. If a bpi's peer is the failed peer, remove the bpi.
4024 * b. If the removed ENCAP bpi was first in the list of
4025 * BPIs at this ENCAP node, loop over all monitors
4028 * (1) for each ENCAP monitor, loop over all its
4029 * VPN node monitors and set their RFAPI_MON_FLAG_NEEDCALLBACK
4032 * 2. Visit all BPIs in all VPN import tables.
4033 * a. If a bpi's peer is the failed peer, remove the bpi.
4034 * b. loop over all the VPN node monitors and set their
4035 * RFAPI_MON_FLAG_NEEDCALLBACK flags
4036 * c. If there are no BPIs left at this VPN node,
4041 /* surprise, this gets called from peer_delete(), from rfapi_close() */
4042 static void rfapiProcessPeerDownRt(struct peer
*peer
,
4043 struct rfapi_import_table
*import_table
,
4044 afi_t afi
, safi_t safi
)
4046 struct agg_node
*rn
;
4047 struct bgp_path_info
*bpi
;
4048 struct agg_table
*rt
= NULL
;
4049 void (*timer_service_func
)(struct thread
*) = NULL
;
4051 assert(afi
== AFI_IP
|| afi
== AFI_IP6
);
4057 rt
= import_table
->imported_vpn
[afi
];
4058 timer_service_func
= rfapiWithdrawTimerVPN
;
4061 rt
= import_table
->imported_encap
[afi
];
4062 timer_service_func
= rfapiWithdrawTimerEncap
;
4066 case SAFI_MULTICAST
:
4068 case SAFI_LABELED_UNICAST
:
4071 /* Suppress uninitialized variable warning */
4073 timer_service_func
= NULL
;
4077 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
4078 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
4079 if (bpi
->peer
== peer
) {
4081 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
4082 /* already in holddown, skip */
4086 if (safi
== SAFI_MPLS_VPN
) {
4087 RFAPI_UPDATE_ITABLE_COUNT(
4088 bpi
, import_table
, afi
, -1);
4089 import_table
->holddown_count
[afi
] += 1;
4091 rfapiBiStartWithdrawTimer(import_table
, rn
, bpi
,
4093 timer_service_func
);
4101 * This gets called when a peer connection drops. We have to remove
4102 * all the routes from this peer.
4104 * Current approach is crude. TBD Optimize by setting fewer timers and
4105 * grouping withdrawn routes so we can generate callbacks more
4108 void rfapiProcessPeerDown(struct peer
*peer
)
4112 struct rfapi_import_table
*it
;
4115 * If this peer is a "dummy" peer structure atached to a RFAPI
4116 * nve_descriptor, we don't need to walk the import tables
4117 * because the routes are already withdrawn by rfapi_close()
4119 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_IS_RFAPI_HD
))
4123 * 1. Visit all BPIs in all ENCAP import tables.
4124 * Start withdraw timer on the BPIs that match peer.
4126 * 2. Visit All BPIs in all VPN import tables.
4127 * Start withdraw timer on the BPIs that match peer.
4130 bgp
= bgp_get_default(); /* assume 1 instance for now */
4137 for (it
= h
->imports
; it
; it
= it
->next
) {
4138 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_ENCAP
);
4139 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_ENCAP
);
4140 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_MPLS_VPN
);
4141 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_MPLS_VPN
);
4145 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4146 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4151 * Import an entire RIB (for an afi/safi) to an import table RIB,
4152 * filtered according to the import table's RT list
4154 * TBD: does this function need additions to match rfapiProcessUpdate()
4155 * for, e.g., L2 handling?
4157 static void rfapiBgpTableFilteredImport(struct bgp
*bgp
,
4158 struct rfapi_import_table
*it
,
4159 afi_t afi
, safi_t safi
)
4161 struct bgp_dest
*dest1
;
4162 struct bgp_dest
*dest2
;
4164 /* Only these SAFIs have 2-level RIBS */
4165 assert(safi
== SAFI_MPLS_VPN
|| safi
== SAFI_ENCAP
);
4168 * Now visit all the rd nodes and the nodes of all the
4169 * route tables attached to them, and import the routes
4170 * if they have matching route targets
4172 for (dest1
= bgp_table_top(bgp
->rib
[afi
][safi
]); dest1
;
4173 dest1
= bgp_route_next(dest1
)) {
4175 if (bgp_dest_has_bgp_path_info_data(dest1
)) {
4177 for (dest2
= bgp_table_top(
4178 bgp_dest_get_bgp_table_info(dest1
));
4179 dest2
; dest2
= bgp_route_next(dest2
)) {
4181 struct bgp_path_info
*bpi
;
4183 for (bpi
= bgp_dest_get_bgp_path_info(dest2
);
4184 bpi
; bpi
= bpi
->next
) {
4187 if (CHECK_FLAG(bpi
->flags
,
4192 label
= decode_label(
4193 &bpi
->extra
->label
[0]);
4194 (*rfapiBgpInfoFilteredImportFunction(
4196 it
, /* which import table */
4197 FIF_ACTION_UPDATE
, bpi
->peer
,
4199 bgp_dest_get_prefix(dest2
),
4201 (struct prefix_rd
*)
4202 bgp_dest_get_prefix(
4204 bpi
->attr
, bpi
->type
,
4205 bpi
->sub_type
, &label
);
4213 /* per-bgp-instance rfapi data */
4214 struct rfapi
*bgp_rfapi_new(struct bgp
*bgp
)
4218 struct rfapi_rfp_cfg
*cfg
= NULL
;
4219 struct rfapi_rfp_cb_methods
*cbm
= NULL
;
4221 assert(bgp
->rfapi_cfg
== NULL
);
4223 h
= XCALLOC(MTYPE_RFAPI
, sizeof(struct rfapi
));
4225 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4226 h
->un
[afi
] = agg_table_init();
4230 * initialize the ce import table
4232 h
->it_ce
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4233 sizeof(struct rfapi_import_table
));
4234 h
->it_ce
->imported_vpn
[AFI_IP
] = agg_table_init();
4235 h
->it_ce
->imported_vpn
[AFI_IP6
] = agg_table_init();
4236 h
->it_ce
->imported_encap
[AFI_IP
] = agg_table_init();
4237 h
->it_ce
->imported_encap
[AFI_IP6
] = agg_table_init();
4238 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4239 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4242 * Set up work queue for deferred rfapi_close operations
4244 h
->deferred_close_q
=
4245 work_queue_new(bm
->master
, "rfapi deferred close");
4246 h
->deferred_close_q
->spec
.workfunc
= rfapi_deferred_close_workfunc
;
4247 h
->deferred_close_q
->spec
.data
= h
;
4249 h
->rfp
= rfp_start(bm
->master
, &cfg
, &cbm
);
4250 bgp
->rfapi_cfg
= bgp_rfapi_cfg_new(cfg
);
4252 h
->rfp_methods
= *cbm
;
4257 void bgp_rfapi_destroy(struct bgp
*bgp
, struct rfapi
*h
)
4261 if (bgp
== NULL
|| h
== NULL
)
4264 if (h
->resolve_nve_nexthop
) {
4265 skiplist_free(h
->resolve_nve_nexthop
);
4266 h
->resolve_nve_nexthop
= NULL
;
4269 agg_table_finish(h
->it_ce
->imported_vpn
[AFI_IP
]);
4270 agg_table_finish(h
->it_ce
->imported_vpn
[AFI_IP6
]);
4271 agg_table_finish(h
->it_ce
->imported_encap
[AFI_IP
]);
4272 agg_table_finish(h
->it_ce
->imported_encap
[AFI_IP6
]);
4274 if (h
->import_mac
) {
4275 struct rfapi_import_table
*it
;
4280 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4282 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4285 rfapiImportTableFlush(it
);
4286 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
4288 skiplist_free(h
->import_mac
);
4289 h
->import_mac
= NULL
;
4292 work_queue_free_and_null(&h
->deferred_close_q
);
4297 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4298 agg_table_finish(h
->un
[afi
]);
4301 XFREE(MTYPE_RFAPI_IMPORTTABLE
, h
->it_ce
);
4302 XFREE(MTYPE_RFAPI
, h
);
4305 struct rfapi_import_table
*
4306 rfapiImportTableRefAdd(struct bgp
*bgp
, struct ecommunity
*rt_import_list
,
4307 struct rfapi_nve_group_cfg
*rfg
)
4310 struct rfapi_import_table
*it
;
4316 for (it
= h
->imports
; it
; it
= it
->next
) {
4317 if (ecommunity_cmp(it
->rt_import_list
, rt_import_list
))
4321 vnc_zlog_debug_verbose("%s: matched it=%p", __func__
, it
);
4324 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4325 sizeof(struct rfapi_import_table
));
4326 it
->next
= h
->imports
;
4329 it
->rt_import_list
= ecommunity_dup(rt_import_list
);
4331 it
->monitor_exterior_orphans
=
4332 skiplist_new(0, NULL
, prefix_free_lists
);
4335 * fill import route tables from RIBs
4337 * Potential area for optimization. If this occurs when
4338 * tables are large (e.g., the operator adds a nve group
4339 * with a new RT list to a running system), it could take
4343 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4345 it
->imported_vpn
[afi
] = agg_table_init();
4346 it
->imported_encap
[afi
] = agg_table_init();
4348 rfapiBgpTableFilteredImport(bgp
, it
, afi
,
4350 rfapiBgpTableFilteredImport(bgp
, it
, afi
, SAFI_ENCAP
);
4352 vnc_import_bgp_exterior_redist_enable_it(bgp
, afi
, it
);
4362 * skiplist element free function
4364 static void delete_rem_pfx_na_free(void *na
)
4366 uint32_t *pCounter
= ((struct rfapi_nve_addr
*)na
)->info
;
4369 XFREE(MTYPE_RFAPI_NVE_ADDR
, na
);
4373 * Common deleter for IP and MAC import tables
4375 static void rfapiDeleteRemotePrefixesIt(
4376 struct bgp
*bgp
, struct rfapi_import_table
*it
, struct prefix
*un
,
4377 struct prefix
*vn
, struct prefix
*p
, int delete_active
,
4378 int delete_holddown
, uint32_t *pARcount
, uint32_t *pAHcount
,
4379 uint32_t *pHRcount
, uint32_t *pHHcount
,
4380 struct skiplist
*uniq_active_nves
, struct skiplist
*uniq_holddown_nves
)
4384 #ifdef DEBUG_L2_EXTRA
4386 char buf_pfx
[PREFIX_STRLEN
];
4389 prefix2str(p
, buf_pfx
, sizeof(buf_pfx
));
4395 vnc_zlog_debug_verbose(
4396 "%s: entry, p=%s, delete_active=%d, delete_holddown=%d",
4397 __func__
, buf_pfx
, delete_active
, delete_holddown
);
4401 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4403 struct agg_table
*rt
;
4404 struct agg_node
*rn
;
4406 if (p
&& (family2afi(p
->family
) != afi
)) {
4410 rt
= it
->imported_vpn
[afi
];
4414 vnc_zlog_debug_verbose("%s: scanning rt for afi=%d", __func__
,
4417 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
4418 struct bgp_path_info
*bpi
;
4419 struct bgp_path_info
*next
;
4420 const struct prefix
*rn_p
= agg_node_get_prefix(rn
);
4422 if (p
&& VNC_DEBUG(IMPORT_DEL_REMOTE
))
4423 vnc_zlog_debug_any("%s: want %pFX, have %pRN",
4426 if (p
&& prefix_cmp(p
, rn_p
))
4429 vnc_zlog_debug_verbose("%s: rn pfx=%pRN", __func__
, rn
);
4431 /* TBD is this valid for afi == AFI_L2VPN? */
4432 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
4434 for (bpi
= rn
->info
; bpi
; bpi
= next
) {
4443 vnc_zlog_debug_verbose("%s: examining bpi %p",
4446 if (!rfapiGetNexthop(bpi
->attr
, &qpt
))
4451 || !prefix_match(vn
, &qpt
)) {
4452 #ifdef DEBUG_L2_EXTRA
4453 vnc_zlog_debug_verbose(
4454 "%s: continue at vn && !qpt_valid || !prefix_match(vn, &qpt)",
4461 if (!rfapiGetUnAddrOfVpnBi(bpi
, &qct
))
4466 || !prefix_match(un
, &qct
)) {
4467 #ifdef DEBUG_L2_EXTRA
4468 vnc_zlog_debug_verbose(
4469 "%s: continue at un && !qct_valid || !prefix_match(un, &qct)",
4481 * If this route is waiting to be deleted
4483 * a previous withdraw, we must cancel its
4486 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
4487 if (!delete_holddown
)
4489 if (bpi
->extra
->vnc
.import
.timer
) {
4490 struct rfapi_withdraw
*wcb
=
4497 ->holddown_count
[afi
] -=
4499 RFAPI_UPDATE_ITABLE_COUNT(
4500 bpi
, wcb
->import_table
,
4502 XFREE(MTYPE_RFAPI_WITHDRAW
,
4505 bpi
->extra
->vnc
.import
4514 vnc_zlog_debug_verbose(
4515 "%s: deleting bpi %p (qct_valid=%d, qpt_valid=%d, delete_holddown=%d, delete_active=%d)",
4516 __func__
, bpi
, qct_valid
, qpt_valid
,
4517 delete_holddown
, delete_active
);
4523 if (qct_valid
&& qpt_valid
) {
4525 struct rfapi_nve_addr na
;
4526 struct rfapi_nve_addr
*nap
;
4528 memset(&na
, 0, sizeof(na
));
4529 assert(!rfapiQprefix2Raddr(&qct
,
4531 assert(!rfapiQprefix2Raddr(&qpt
,
4534 if (skiplist_search(
4537 : uniq_holddown_nves
),
4538 &na
, (void **)&nap
)) {
4542 MTYPE_RFAPI_NVE_ADDR
,
4546 nap
->info
= is_active
4552 : uniq_holddown_nves
),
4555 rfapiNveAddr2Str(nap
, line
,
4560 vnc_direct_bgp_rh_del_route(bgp
, afi
, rn_p
,
4563 RFAPI_UPDATE_ITABLE_COUNT(bpi
, it
, afi
, -1);
4564 it
->holddown_count
[afi
] += 1;
4565 rfapiExpireVpnNow(it
, rn
, bpi
, 1);
4567 vnc_zlog_debug_verbose(
4568 "%s: incrementing count (is_active=%d)",
4569 __func__
, is_active
);
4582 * For use by the "clear vnc prefixes" command
4584 /*------------------------------------------
4585 * rfapiDeleteRemotePrefixes
4587 * UI helper: For use by the "clear vnc prefixes" command
4590 * un if set, tunnel must match this prefix
4591 * vn if set, nexthop prefix must match this prefix
4592 * p if set, prefix must match this prefix
4593 * it if set, only look in this import table
4596 * pARcount number of active routes deleted
4597 * pAHcount number of active nves deleted
4598 * pHRcount number of holddown routes deleted
4599 * pHHcount number of holddown nves deleted
4603 --------------------------------------------*/
4604 void rfapiDeleteRemotePrefixes(struct prefix
*un
, struct prefix
*vn
,
4606 struct rfapi_import_table
*arg_it
,
4607 int delete_active
, int delete_holddown
,
4608 uint32_t *pARcount
, uint32_t *pAHcount
,
4609 uint32_t *pHRcount
, uint32_t *pHHcount
)
4613 struct rfapi_import_table
*it
;
4614 uint32_t deleted_holddown_route_count
= 0;
4615 uint32_t deleted_active_route_count
= 0;
4616 uint32_t deleted_holddown_nve_count
= 0;
4617 uint32_t deleted_active_nve_count
= 0;
4618 struct skiplist
*uniq_holddown_nves
;
4619 struct skiplist
*uniq_active_nves
;
4623 bgp
= bgp_get_default(); /* assume 1 instance for now */
4624 /* If no bgp instantiated yet, no vnc prefixes exist */
4631 uniq_holddown_nves
=
4632 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4634 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4637 * Iterate over all import tables; do a filtered import
4638 * for the afi/safi combination
4647 vnc_zlog_debug_verbose(
4648 "%s: calling rfapiDeleteRemotePrefixesIt() on (IP) import %p",
4651 rfapiDeleteRemotePrefixesIt(
4652 bgp
, it
, un
, vn
, p
, delete_active
, delete_holddown
,
4653 &deleted_active_route_count
, &deleted_active_nve_count
,
4654 &deleted_holddown_route_count
,
4655 &deleted_holddown_nve_count
, uniq_active_nves
,
4656 uniq_holddown_nves
);
4665 * Now iterate over L2 import tables
4667 if (h
->import_mac
&& !(p
&& (p
->family
!= AF_ETHERNET
))) {
4669 void *cursor
= NULL
;
4673 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4675 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4678 vnc_zlog_debug_verbose(
4679 "%s: calling rfapiDeleteRemotePrefixesIt() on import_mac %p",
4682 rfapiDeleteRemotePrefixesIt(
4683 bgp
, it
, un
, vn
, p
, delete_active
,
4684 delete_holddown
, &deleted_active_route_count
,
4685 &deleted_active_nve_count
,
4686 &deleted_holddown_route_count
,
4687 &deleted_holddown_nve_count
, uniq_active_nves
,
4688 uniq_holddown_nves
);
4693 * our custom element freeing function above counts as it deletes
4695 skiplist_free(uniq_holddown_nves
);
4696 skiplist_free(uniq_active_nves
);
4699 *pARcount
= deleted_active_route_count
;
4701 *pAHcount
= deleted_active_nve_count
;
4703 *pHRcount
= deleted_holddown_route_count
;
4705 *pHHcount
= deleted_holddown_nve_count
;
4710 /*------------------------------------------
4711 * rfapiCountRemoteRoutes
4713 * UI helper: count VRF routes from BGP side
4718 * pALRcount count of active local routes
4719 * pARRcount count of active remote routes
4720 * pHRcount count of holddown routes
4721 * pIRcount count of direct imported routes
4725 --------------------------------------------*/
4726 void rfapiCountAllItRoutes(int *pALRcount
, /* active local routes */
4727 int *pARRcount
, /* active remote routes */
4728 int *pHRcount
, /* holddown routes */
4729 int *pIRcount
) /* imported routes */
4733 struct rfapi_import_table
*it
;
4736 int total_active_local
= 0;
4737 int total_active_remote
= 0;
4738 int total_holddown
= 0;
4739 int total_imported
= 0;
4741 bgp
= bgp_get_default(); /* assume 1 instance for now */
4748 * Iterate over all import tables; do a filtered import
4749 * for the afi/safi combination
4752 for (it
= h
->imports
; it
; it
= it
->next
) {
4754 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4756 total_active_local
+= it
->local_count
[afi
];
4757 total_active_remote
+= it
->remote_count
[afi
];
4758 total_holddown
+= it
->holddown_count
[afi
];
4759 total_imported
+= it
->imported_count
[afi
];
4766 if (h
->import_mac
) {
4768 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4770 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4773 total_active_local
+= it
->local_count
[AFI_L2VPN
];
4774 total_active_remote
+= it
->remote_count
[AFI_L2VPN
];
4775 total_holddown
+= it
->holddown_count
[AFI_L2VPN
];
4776 total_imported
+= it
->imported_count
[AFI_L2VPN
];
4782 *pALRcount
= total_active_local
;
4785 *pARRcount
= total_active_remote
;
4788 *pHRcount
= total_holddown
;
4791 *pIRcount
= total_imported
;
4795 /*------------------------------------------
4796 * rfapiGetHolddownFromLifetime
4798 * calculate holddown value based on lifetime
4804 * Holddown value based on lifetime, holddown_factor,
4805 * and RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
4807 --------------------------------------------*/
4808 /* hold down time maxes out at RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY */
4809 uint32_t rfapiGetHolddownFromLifetime(uint32_t lifetime
)
4814 bgp
= bgp_get_default();
4815 if (bgp
&& bgp
->rfapi_cfg
)
4816 factor
= bgp
->rfapi_cfg
->rfp_cfg
.holddown_factor
;
4818 factor
= RFAPI_RFP_CFG_DEFAULT_HOLDDOWN_FACTOR
;
4820 if (factor
< 100 || lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4821 lifetime
= lifetime
* factor
/ 100;
4822 if (lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4825 return RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
;