3 * Copyright 2009-2016, LabN Consulting, L.L.C.
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version 2
9 * of the License, or (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License along
17 * with this program; see the file COPYING; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
22 * File: rfapi_import.c
23 * Purpose: Handle import of routes from BGP to RFAPI
28 #include "lib/zebra.h"
29 #include "lib/prefix.h"
30 #include "lib/table.h"
32 #include "lib/memory.h"
34 #include "lib/skiplist.h"
35 #include "lib/thread.h"
36 #include "lib/stream.h"
38 #include "bgpd/bgpd.h"
39 #include "bgpd/bgp_ecommunity.h"
40 #include "bgpd/bgp_attr.h"
41 #include "bgpd/bgp_route.h"
42 #include "bgpd/bgp_mplsvpn.h" /* prefix_rd2str() */
43 #include "bgpd/bgp_vnc_types.h"
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 route_node
*node
;
82 struct bgp_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 * route_top() or route_next() in a loop, so we need to pass this
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()
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 route_table
*rt
;
142 struct route_node
*rn
;
144 int holddown_count
= 0;
146 int imported_count
= 0;
147 int remote_count
= 0;
149 rt
= it
->imported_vpn
[afi
];
151 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
153 struct bgp_info
*next
;
155 for (bi
= rn
->info
; bi
; bi
= next
) {
158 if (CHECK_FLAG(bi
->flags
,
163 if (RFAPI_LOCAL_BI(bi
)) {
166 if (RFAPI_DIRECT_IMPORT_BI(
177 if (it
->holddown_count
[afi
] != holddown_count
) {
178 vnc_zlog_debug_verbose(
179 "%s: it->holddown_count %d != holddown_count %d",
180 __func__
, it
->holddown_count
[afi
],
184 if (it
->remote_count
[afi
] != remote_count
) {
185 vnc_zlog_debug_verbose(
186 "%s: it->remote_count %d != remote_count %d",
187 __func__
, it
->remote_count
[afi
],
191 if (it
->imported_count
[afi
] != imported_count
) {
192 vnc_zlog_debug_verbose(
193 "%s: it->imported_count %d != imported_count %d",
194 __func__
, it
->imported_count
[afi
],
202 #if DEBUG_ROUTE_COUNTERS
203 #define VNC_ITRCCK do {rfapiCheckRouteCount();} while (0)
209 * Validate reference count for a node in an import table
211 * Normally lockoffset is 0 for nodes in quiescent state. However,
212 * route_unlock_node will delete the node if it is called when
213 * node->lock == 1, and we have to validate the refcount before
214 * the node is deleted. In this case, we specify lockoffset 1.
216 void rfapiCheckRefcount(struct route_node
*rn
, safi_t safi
, int lockoffset
)
218 unsigned int count_bi
= 0;
219 unsigned int count_monitor
= 0;
221 struct rfapi_monitor_encap
*hme
;
222 struct rfapi_monitor_vpn
*hmv
;
224 for (bi
= rn
->info
; bi
; bi
= bi
->next
)
229 ++count_monitor
; /* rfapi_it_extra */
236 for (hme
= RFAPI_MONITOR_ENCAP(rn
); hme
;
243 for (hmv
= RFAPI_MONITOR_VPN(rn
); hmv
; hmv
= hmv
->next
)
246 if (RFAPI_MONITOR_EXTERIOR(rn
)->source
) {
247 ++count_monitor
; /* sl */
249 for (rc
= skiplist_next(
250 RFAPI_MONITOR_EXTERIOR(rn
)->source
,
251 NULL
, NULL
, &cursor
);
254 RFAPI_MONITOR_EXTERIOR(rn
)->source
,
255 NULL
, NULL
, &cursor
)) {
257 ++count_monitor
; /* sl entry */
267 if (count_bi
+ count_monitor
+ lockoffset
!= rn
->lock
) {
268 vnc_zlog_debug_verbose(
269 "%s: count_bi=%d, count_monitor=%d, lockoffset=%d, rn->lock=%d",
270 __func__
, count_bi
, count_monitor
, lockoffset
,
277 * Perform deferred rfapi_close operations that were queued
280 static wq_item_status
rfapi_deferred_close_workfunc(struct work_queue
*q
,
283 struct rfapi_descriptor
*rfd
= data
;
284 struct rfapi
*h
= q
->spec
.data
;
286 assert(!(h
->flags
& RFAPI_INCALLBACK
));
288 vnc_zlog_debug_verbose("%s: completed deferred close on handle %p",
294 * Extract layer 2 option from Encap TLVS in BGP attrs
296 int rfapiGetL2o(struct attr
*attr
, struct rfapi_l2address_option
*l2o
)
300 struct bgp_attr_encap_subtlv
*pEncap
;
302 for (pEncap
= attr
->vnc_subtlvs
; pEncap
;
303 pEncap
= pEncap
->next
) {
305 if (pEncap
->type
== BGP_VNC_SUBTLV_TYPE_RFPOPTION
) {
307 == RFAPI_VN_OPTION_TYPE_L2ADDR
) {
309 if (pEncap
->value
[1] == 14) {
310 memcpy(l2o
->macaddr
.octet
,
327 l2o
->logical_net_id
=
330 + ((pEncap
->value
[14]
333 + ((pEncap
->value
[13]
348 * Extract the lifetime from the Tunnel Encap attribute of a route in
351 int rfapiGetVncLifetime(struct attr
*attr
, uint32_t *lifetime
)
353 struct bgp_attr_encap_subtlv
*pEncap
;
355 *lifetime
= RFAPI_INFINITE_LIFETIME
; /* default to infinite */
359 for (pEncap
= attr
->vnc_subtlvs
; pEncap
;
360 pEncap
= pEncap
->next
) {
363 == BGP_VNC_SUBTLV_TYPE_LIFETIME
) { /* lifetime */
364 if (pEncap
->length
== 4) {
365 memcpy(lifetime
, pEncap
->value
, 4);
366 *lifetime
= ntohl(*lifetime
);
377 * Extract the tunnel type from the extended community
379 int rfapiGetTunnelType(struct attr
*attr
, bgp_encap_types
*type
)
381 *type
= BGP_ENCAP_TYPE_MPLS
; /* default to MPLS */
382 if (attr
&& attr
->ecommunity
) {
383 struct ecommunity
*ecom
= attr
->ecommunity
;
386 for (i
= 0; i
< (ecom
->size
* ECOMMUNITY_SIZE
);
387 i
+= ECOMMUNITY_SIZE
) {
391 if (ep
[0] == ECOMMUNITY_ENCODE_OPAQUE
392 && ep
[1] == ECOMMUNITY_OPAQUE_SUBTYPE_ENCAP
) {
393 *type
= (ep
[6] << 8) + ep
[7];
404 * Look for UN address in Encap attribute
406 int rfapiGetVncTunnelUnAddr(struct attr
*attr
, struct prefix
*p
)
408 struct bgp_attr_encap_subtlv
*pEncap
;
409 bgp_encap_types tun_type
;
411 rfapiGetTunnelType(attr
, &tun_type
);
412 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
415 /* MPLS carries UN address in next hop */
416 rfapiNexthop2Prefix(attr
, p
);
423 for (pEncap
= attr
->encap_subtlvs
; pEncap
;
424 pEncap
= pEncap
->next
) {
427 == BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
) { /* un
430 switch (pEncap
->length
) {
435 memcpy(p
->u
.val
, pEncap
->value
,
442 p
->family
= AF_INET6
;
444 memcpy(p
->u
.val
, pEncap
->value
,
457 * Get UN address wherever it might be
459 int rfapiGetUnAddrOfVpnBi(struct bgp_info
*bi
, struct prefix
*p
)
461 /* If it's in this route's VNC attribute, we're done */
462 if (!rfapiGetVncTunnelUnAddr(bi
->attr
, p
))
465 * Otherwise, see if it's cached from a corresponding ENCAP SAFI
469 switch (bi
->extra
->vnc
.import
.un_family
) {
472 p
->family
= bi
->extra
->vnc
.import
.un_family
;
473 p
->u
.prefix4
= bi
->extra
->vnc
.import
.un
.addr4
;
479 p
->family
= bi
->extra
->vnc
.import
.un_family
;
480 p
->u
.prefix6
= bi
->extra
->vnc
.import
.un
.addr6
;
487 #if DEBUG_ENCAP_MONITOR
488 vnc_zlog_debug_verbose(
489 "%s: bi->extra->vnc.import.un_family is 0, no UN addr",
501 * Make a new bgp_info from gathered parameters
503 static struct bgp_info
*rfapiBgpInfoCreate(struct attr
*attr
, struct peer
*peer
,
504 void *rfd
, struct prefix_rd
*prd
,
505 u_char type
, u_char sub_type
,
508 struct bgp_info
*new;
510 new = bgp_info_new();
515 new->attr
= bgp_attr_intern(attr
);
517 bgp_info_extra_get(new);
519 new->extra
->vnc
.import
.rd
= *prd
;
520 rfapi_time(&new->extra
->vnc
.import
.create_time
);
523 encode_label(*label
, &new->extra
->label
);
525 new->sub_type
= sub_type
;
533 * Frees bgp_info as used in import tables (parts are not
534 * allocated exactly the way they are in the main RIBs)
536 static void rfapiBgpInfoFree(struct bgp_info
*goner
)
542 vnc_zlog_debug_verbose("%s: calling peer_unlock(%p), #%d",
543 __func__
, goner
->peer
,
545 peer_unlock(goner
->peer
);
549 bgp_attr_unintern(&goner
->attr
);
552 assert(!goner
->extra
->damp_info
); /* Not used in import tbls */
553 XFREE(MTYPE_BGP_ROUTE_EXTRA
, goner
->extra
);
556 XFREE(MTYPE_BGP_ROUTE
, goner
);
559 struct rfapi_import_table
*rfapiMacImportTableGetNoAlloc(struct bgp
*bgp
,
563 struct rfapi_import_table
*it
= NULL
;
564 uintptr_t lni_as_ptr
= lni
;
573 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
))
579 struct rfapi_import_table
*rfapiMacImportTableGet(struct bgp
*bgp
, uint32_t lni
)
582 struct rfapi_import_table
*it
= NULL
;
583 uintptr_t lni_as_ptr
= lni
;
588 if (!h
->import_mac
) {
589 /* default cmp is good enough for LNI */
590 h
->import_mac
= skiplist_new(0, NULL
, NULL
);
593 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
)) {
595 struct ecommunity
*enew
;
596 struct ecommunity_val eval
;
599 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
600 sizeof(struct rfapi_import_table
));
601 /* set RT list of new import table based on LNI */
602 memset((char *)&eval
, 0, sizeof(eval
));
603 eval
.val
[0] = 0; /* VNC L2VPN */
604 eval
.val
[1] = 2; /* VNC L2VPN */
605 eval
.val
[5] = (lni
>> 16) & 0xff;
606 eval
.val
[6] = (lni
>> 8) & 0xff;
607 eval
.val
[7] = (lni
>> 0) & 0xff;
609 enew
= ecommunity_new();
610 ecommunity_add_val(enew
, &eval
);
611 it
->rt_import_list
= enew
;
613 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
614 it
->imported_vpn
[afi
] = route_table_init();
615 it
->imported_encap
[afi
] = route_table_init();
618 it
->l2_logical_net_id
= lni
;
620 skiplist_insert(h
->import_mac
, (void *)lni_as_ptr
, it
);
628 * Implement MONITOR_MOVE_SHORTER(original_node) from
629 * RFAPI-Import-Event-Handling.txt
631 * Returns pointer to the list of moved monitors
633 static struct rfapi_monitor_vpn
*
634 rfapiMonitorMoveShorter(struct route_node
*original_vpn_node
, int lockoffset
)
637 struct route_node
*par
;
638 struct rfapi_monitor_vpn
*m
;
639 struct rfapi_monitor_vpn
*mlast
;
640 struct rfapi_monitor_vpn
*moved
;
642 int parent_already_refcounted
= 0;
644 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
, lockoffset
);
646 #if DEBUG_MONITOR_MOVE_SHORTER
650 prefix2str(&original_vpn_node
->p
, buf
, BUFSIZ
);
652 vnc_zlog_debug_verbose("%s: called with node pfx=%s", __func__
,
658 * 1. If there is at least one bi (either regular route or
659 * route marked as withdrawn, with a pending timer) at
660 * original_node with a valid UN address, we're done. Return.
662 for (bi
= original_vpn_node
->info
; bi
; bi
= bi
->next
) {
665 if (!rfapiGetUnAddrOfVpnBi(bi
, &pfx
)) {
666 #if DEBUG_MONITOR_MOVE_SHORTER
667 vnc_zlog_debug_verbose(
668 "%s: have valid UN at original node, no change",
676 * 2. Travel up the tree (toward less-specific prefixes) from
677 * original_node to find the first node that has at least
678 * one route (even if it is only a withdrawn route) with a
679 * valid UN address. Call this node "Node P."
681 for (par
= original_vpn_node
->parent
; par
; par
= par
->parent
) {
682 for (bi
= par
->info
; bi
; bi
= bi
->next
) {
684 if (!rfapiGetUnAddrOfVpnBi(bi
, &pfx
)) {
693 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 0);
697 * If no less-specific routes, try to use the 0/0 node
700 /* this isn't necessarily 0/0 */
701 par
= route_top(original_vpn_node
->table
);
704 * If we got the top node but it wasn't 0/0,
707 if (par
&& par
->p
.prefixlen
) {
708 route_unlock_node(par
); /* maybe free */
713 ++parent_already_refcounted
;
718 * Create 0/0 node if it isn't there
721 struct prefix pfx_default
;
723 memset(&pfx_default
, 0, sizeof(pfx_default
));
724 pfx_default
.family
= original_vpn_node
->p
.family
;
726 /* creates default node if none exists */
727 par
= route_node_get(original_vpn_node
->table
, &pfx_default
);
728 ++parent_already_refcounted
;
732 * 3. Move each of the monitors found at original_node to Node P.
733 * These are "Moved Monitors."
738 * Attach at end so that the list pointer we return points
739 * only to the moved routes
741 for (m
= RFAPI_MONITOR_VPN(par
), mlast
= NULL
; m
;
742 mlast
= m
, m
= m
->next
)
746 moved
= mlast
->next
= RFAPI_MONITOR_VPN(original_vpn_node
);
748 moved
= RFAPI_MONITOR_VPN_W_ALLOC(par
) =
749 RFAPI_MONITOR_VPN(original_vpn_node
);
751 if (RFAPI_MONITOR_VPN(
752 original_vpn_node
)) /* check agg, so not allocated */
753 RFAPI_MONITOR_VPN_W_ALLOC(original_vpn_node
) = NULL
;
756 * update the node pointers on the monitors
758 for (m
= moved
; m
; m
= m
->next
) {
763 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
,
764 parent_already_refcounted
- movecount
);
765 while (movecount
> parent_already_refcounted
) {
766 route_lock_node(par
);
767 ++parent_already_refcounted
;
769 while (movecount
< parent_already_refcounted
) {
770 /* unlikely, but code defensively */
771 route_unlock_node(par
);
772 --parent_already_refcounted
;
774 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
,
775 movecount
+ lockoffset
);
776 while (movecount
--) {
777 route_unlock_node(original_vpn_node
);
780 #if DEBUG_MONITOR_MOVE_SHORTER
784 prefix2str(&par
->p
, buf
, BUFSIZ
);
786 vnc_zlog_debug_verbose("%s: moved to node pfx=%s", __func__
,
796 * Implement MONITOR_MOVE_LONGER(new_node) from
797 * RFAPI-Import-Event-Handling.txt
799 static void rfapiMonitorMoveLonger(struct route_node
*new_vpn_node
)
801 struct rfapi_monitor_vpn
*monitor
;
802 struct rfapi_monitor_vpn
*mlast
;
804 struct route_node
*par
;
806 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
809 * Make sure we have at least one valid route at the new node
811 for (bi
= new_vpn_node
->info
; bi
; bi
= bi
->next
) {
813 if (!rfapiGetUnAddrOfVpnBi(bi
, &pfx
))
818 vnc_zlog_debug_verbose(
819 "%s: no valid routes at node %p, so not attempting moves",
820 __func__
, new_vpn_node
);
825 * Find first parent node that has monitors
827 for (par
= new_vpn_node
->parent
; par
; par
= par
->parent
) {
828 if (RFAPI_MONITOR_VPN(par
))
833 vnc_zlog_debug_verbose(
834 "%s: no parent nodes with monitors, done", __func__
);
839 * Check each of these monitors to see of their longest-match
840 * is now the updated node. Move any such monitors to the more-
841 * specific updated node
843 for (mlast
= NULL
, monitor
= RFAPI_MONITOR_VPN(par
); monitor
;) {
846 * If new longest match for monitor prefix is the new
847 * route's prefix, move monitor to new route's prefix
849 if (prefix_match(&new_vpn_node
->p
, &monitor
->p
)) {
852 mlast
->next
= monitor
->next
;
854 RFAPI_MONITOR_VPN_W_ALLOC(par
) = monitor
->next
;
859 monitor
->next
= RFAPI_MONITOR_VPN(new_vpn_node
);
860 RFAPI_MONITOR_VPN_W_ALLOC(new_vpn_node
) = monitor
;
861 monitor
->node
= new_vpn_node
;
863 route_lock_node(new_vpn_node
); /* incr refcount */
865 monitor
= mlast
? mlast
->next
: RFAPI_MONITOR_VPN(par
);
867 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 1);
868 /* decr refcount after we're done with par as this might
870 route_unlock_node(par
);
875 monitor
= monitor
->next
;
878 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
882 static void rfapiBgpInfoChainFree(struct bgp_info
*bi
)
884 struct bgp_info
*next
;
889 * If there is a timer waiting to delete this bi, cancel
890 * the timer and delete immediately
892 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
893 && bi
->extra
->vnc
.import
.timer
) {
896 (struct thread
*)bi
->extra
->vnc
.import
.timer
;
897 struct rfapi_withdraw
*wcb
= t
->arg
;
899 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
905 rfapiBgpInfoFree(bi
);
910 static void rfapiImportTableFlush(struct rfapi_import_table
*it
)
917 ecommunity_free(&it
->rt_import_list
);
918 it
->rt_import_list
= NULL
;
920 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
922 struct route_node
*rn
;
924 for (rn
= route_top(it
->imported_vpn
[afi
]); rn
;
925 rn
= route_next(rn
)) {
927 * Each route_node has:
928 * aggregate: points to rfapi_it_extra with monitor
930 * info: points to chain of bgp_info
932 /* free bgp_info and its children */
933 rfapiBgpInfoChainFree(rn
->info
);
936 rfapiMonitorExtraFlush(SAFI_MPLS_VPN
, rn
);
939 for (rn
= route_top(it
->imported_encap
[afi
]); rn
;
940 rn
= route_next(rn
)) {
941 /* free bgp_info and its children */
942 rfapiBgpInfoChainFree(rn
->info
);
945 rfapiMonitorExtraFlush(SAFI_ENCAP
, rn
);
948 route_table_finish(it
->imported_vpn
[afi
]);
949 route_table_finish(it
->imported_encap
[afi
]);
951 if (it
->monitor_exterior_orphans
) {
952 skiplist_free(it
->monitor_exterior_orphans
);
956 void rfapiImportTableRefDelByIt(struct bgp
*bgp
,
957 struct rfapi_import_table
*it_target
)
960 struct rfapi_import_table
*it
;
961 struct rfapi_import_table
*prev
= NULL
;
968 for (it
= h
->imports
; it
; prev
= it
, it
= it
->next
) {
974 assert(it
->refcount
);
980 prev
->next
= it
->next
;
982 h
->imports
= it
->next
;
984 rfapiImportTableFlush(it
);
985 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
989 #if RFAPI_REQUIRE_ENCAP_BEEC
991 * Look for magic BGP Encapsulation Extended Community value
992 * Format in RFC 5512 Sect. 4.5
994 static int rfapiEcommunitiesMatchBeec(struct ecommunity
*ecom
,
995 bgp_encap_types type
)
1002 for (i
= 0; i
< (ecom
->size
* ECOMMUNITY_SIZE
); i
+= ECOMMUNITY_SIZE
) {
1008 if (ep
[0] == ECOMMUNITY_ENCODE_OPAQUE
1009 && ep
[1] == ECOMMUNITY_OPAQUE_SUBTYPE_ENCAP
1010 && ep
[6] == ((type
&& 0xff00) >> 8)
1011 && ep
[7] == (type
& 0xff)) {
1020 int rfapiEcommunitiesIntersect(struct ecommunity
*e1
, struct ecommunity
*e2
)
1029 s1
= ecommunity_ecom2str(e1
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
1030 s2
= ecommunity_ecom2str(e2
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
1031 vnc_zlog_debug_verbose("%s: e1[%s], e2[%s]", __func__
, s1
, s2
);
1032 XFREE(MTYPE_ECOMMUNITY_STR
, s1
);
1033 XFREE(MTYPE_ECOMMUNITY_STR
, s2
);
1036 for (i
= 0; i
< e1
->size
; ++i
) {
1037 for (j
= 0; j
< e2
->size
; ++j
) {
1038 if (!memcmp(e1
->val
+ (i
* ECOMMUNITY_SIZE
),
1039 e2
->val
+ (j
* ECOMMUNITY_SIZE
),
1049 int rfapiEcommunityGetLNI(struct ecommunity
*ecom
, uint32_t *lni
)
1053 for (i
= 0; i
< ecom
->size
; ++i
) {
1054 uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1056 if ((*(p
+ 0) == 0x00) && (*(p
+ 1) == 0x02)) {
1058 *lni
= (*(p
+ 5) << 16) | (*(p
+ 6) << 8)
1067 int rfapiEcommunityGetEthernetTag(struct ecommunity
*ecom
, uint16_t *tag_id
)
1069 struct bgp
*bgp
= bgp_get_default();
1070 *tag_id
= 0; /* default to untagged */
1073 for (i
= 0; i
< ecom
->size
; ++i
) {
1076 uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1078 /* High-order octet of type. */
1081 if (*p
++ == ECOMMUNITY_ROUTE_TARGET
) {
1082 if (encode
== ECOMMUNITY_ENCODE_AS4
) {
1083 p
= ptr_get_be32(p
, &as
);
1084 } else if (encode
== ECOMMUNITY_ENCODE_AS
) {
1088 2; /* skip next two, tag/vid
1089 always in lowest bytes */
1091 if (as
== bgp
->as
) {
1092 *tag_id
= *p
++ << 8;
1102 static int rfapiVpnBiNhEqualsPt(struct bgp_info
*bi
, struct rfapi_ip_addr
*hpt
)
1109 family
= BGP_MP_NEXTHOP_FAMILY(bi
->attr
->mp_nexthop_len
);
1111 if (hpt
->addr_family
!= family
)
1116 if (bi
->attr
->mp_nexthop_global_in
.s_addr
1117 != hpt
->addr
.v4
.s_addr
)
1122 if (IPV6_ADDR_CMP(&bi
->attr
->mp_nexthop_global
, &hpt
->addr
.v6
))
1136 * Compare 2 VPN BIs. Return true if they have the same VN and UN addresses
1138 static int rfapiVpnBiSamePtUn(struct bgp_info
*bi1
, struct bgp_info
*bi2
)
1140 struct prefix pfx_un1
;
1141 struct prefix pfx_un2
;
1146 if (!bi1
->attr
|| !bi2
->attr
)
1150 * VN address comparisons
1153 if (BGP_MP_NEXTHOP_FAMILY(bi1
->attr
->mp_nexthop_len
)
1154 != BGP_MP_NEXTHOP_FAMILY(bi2
->attr
->mp_nexthop_len
)) {
1158 switch (BGP_MP_NEXTHOP_FAMILY(bi1
->attr
->mp_nexthop_len
)) {
1160 if (bi1
->attr
->mp_nexthop_global_in
.s_addr
1161 != bi2
->attr
->mp_nexthop_global_in
.s_addr
)
1166 if (IPV6_ADDR_CMP(&bi1
->attr
->mp_nexthop_global
,
1167 &bi2
->attr
->mp_nexthop_global
))
1177 * UN address comparisons
1179 if (rfapiGetVncTunnelUnAddr(bi1
->attr
, &pfx_un1
)) {
1181 pfx_un1
.family
= bi1
->extra
->vnc
.import
.un_family
;
1182 switch (bi1
->extra
->vnc
.import
.un_family
) {
1185 bi1
->extra
->vnc
.import
.un
.addr4
;
1189 bi1
->extra
->vnc
.import
.un
.addr6
;
1198 if (rfapiGetVncTunnelUnAddr(bi2
->attr
, &pfx_un2
)) {
1200 pfx_un2
.family
= bi2
->extra
->vnc
.import
.un_family
;
1201 switch (bi2
->extra
->vnc
.import
.un_family
) {
1204 bi2
->extra
->vnc
.import
.un
.addr4
;
1208 bi2
->extra
->vnc
.import
.un
.addr6
;
1217 if (!pfx_un1
.family
|| !pfx_un2
.family
)
1220 if (pfx_un1
.family
!= pfx_un2
.family
)
1223 switch (pfx_un1
.family
) {
1225 if (!IPV4_ADDR_SAME(&pfx_un1
.u
.prefix4
,
1226 &pfx_un2
.u
.prefix4
))
1230 if (!IPV6_ADDR_SAME(&pfx_un1
.u
.prefix6
, &pfx_un2
.u
.prefix6
))
1239 uint8_t rfapiRfpCost(struct attr
*attr
)
1241 if (attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_LOCAL_PREF
)) {
1242 if (attr
->local_pref
> 255) {
1245 return 255 - attr
->local_pref
;
1251 /*------------------------------------------
1254 * Find Layer 2 options in an option chain
1260 * l2o layer 2 options extracted
1264 * 1 no options found
1266 --------------------------------------------*/
1267 int rfapi_extract_l2o(
1268 struct bgp_tea_options
*pHop
, /* chain of options */
1269 struct rfapi_l2address_option
*l2o
) /* return extracted value */
1271 struct bgp_tea_options
*p
;
1273 for (p
= pHop
; p
; p
= p
->next
) {
1274 if ((p
->type
== RFAPI_VN_OPTION_TYPE_L2ADDR
)
1275 && (p
->length
>= 8)) {
1279 memcpy(&l2o
->macaddr
, v
, 6);
1281 l2o
->label
= ((v
[6] << 12) & 0xff000)
1282 + ((v
[7] << 4) & 0xff0)
1283 + ((v
[8] >> 4) & 0xf);
1285 l2o
->local_nve_id
= (uint8_t)v
[10];
1287 l2o
->logical_net_id
=
1288 (v
[11] << 16) + (v
[12] << 8) + (v
[13] << 0);
1296 static struct rfapi_next_hop_entry
*
1297 rfapiRouteInfo2NextHopEntry(struct rfapi_ip_prefix
*rprefix
,
1298 struct bgp_info
*bi
, /* route to encode */
1299 uint32_t lifetime
, /* use this in nhe */
1300 struct route_node
*rn
) /* req for L2 eth addr */
1302 struct rfapi_next_hop_entry
*new;
1303 int have_vnc_tunnel_un
= 0;
1305 #if DEBUG_ENCAP_MONITOR
1306 vnc_zlog_debug_verbose("%s: entry, bi %p, rn %p", __func__
, bi
, rn
);
1309 new = XCALLOC(MTYPE_RFAPI_NEXTHOP
, sizeof(struct rfapi_next_hop_entry
));
1312 new->prefix
= *rprefix
;
1315 && decode_rd_type(bi
->extra
->vnc
.import
.rd
.val
)
1316 == RD_TYPE_VNC_ETH
) {
1319 struct rfapi_vn_option
*vo
;
1321 vo
= XCALLOC(MTYPE_RFAPI_VN_OPTION
,
1322 sizeof(struct rfapi_vn_option
));
1325 vo
->type
= RFAPI_VN_OPTION_TYPE_L2ADDR
;
1327 memcpy(&vo
->v
.l2addr
.macaddr
, &rn
->p
.u
.prefix_eth
.octet
,
1329 /* only low 3 bytes of this are significant */
1331 (void)rfapiEcommunityGetLNI(
1332 bi
->attr
->ecommunity
,
1333 &vo
->v
.l2addr
.logical_net_id
);
1334 (void)rfapiEcommunityGetEthernetTag(
1335 bi
->attr
->ecommunity
, &vo
->v
.l2addr
.tag_id
);
1338 /* local_nve_id comes from lower byte of RD type */
1339 vo
->v
.l2addr
.local_nve_id
= bi
->extra
->vnc
.import
.rd
.val
[1];
1341 /* label comes from MP_REACH_NLRI label */
1342 vo
->v
.l2addr
.label
= decode_label(&bi
->extra
->label
);
1344 new->vn_options
= vo
;
1347 * If there is an auxiliary prefix (i.e., host IP address),
1348 * use it as the nexthop prefix instead of the query prefix
1350 if (bi
->extra
->vnc
.import
.aux_prefix
.family
) {
1351 rfapiQprefix2Rprefix(&bi
->extra
->vnc
.import
.aux_prefix
,
1357 bgp_encap_types tun_type
;
1358 new->prefix
.cost
= rfapiRfpCost(bi
->attr
);
1360 struct bgp_attr_encap_subtlv
*pEncap
;
1362 switch (BGP_MP_NEXTHOP_FAMILY(bi
->attr
->mp_nexthop_len
)) {
1364 new->vn_address
.addr_family
= AF_INET
;
1365 new->vn_address
.addr
.v4
=
1366 bi
->attr
->mp_nexthop_global_in
;
1370 new->vn_address
.addr_family
= AF_INET6
;
1371 new->vn_address
.addr
.v6
= bi
->attr
->mp_nexthop_global
;
1375 zlog_warn("%s: invalid vpn nexthop length: %d",
1376 __func__
, bi
->attr
->mp_nexthop_len
);
1377 rfapi_free_next_hop_list(new);
1381 for (pEncap
= bi
->attr
->vnc_subtlvs
; pEncap
;
1382 pEncap
= pEncap
->next
) {
1383 switch (pEncap
->type
) {
1384 case BGP_VNC_SUBTLV_TYPE_LIFETIME
:
1385 /* use configured lifetime, not attr lifetime */
1389 zlog_warn("%s: unknown VNC option type %d",
1390 __func__
, pEncap
->type
);
1397 rfapiGetTunnelType(bi
->attr
, &tun_type
);
1398 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
1400 /* MPLS carries UN address in next hop */
1401 rfapiNexthop2Prefix(bi
->attr
, &p
);
1402 if (p
.family
!= 0) {
1403 rfapiQprefix2Raddr(&p
, &new->un_address
);
1404 have_vnc_tunnel_un
= 1;
1408 for (pEncap
= bi
->attr
->encap_subtlvs
; pEncap
;
1409 pEncap
= pEncap
->next
) {
1410 switch (pEncap
->type
) {
1411 case BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
:
1413 * Overrides ENCAP UN address, if any
1415 switch (pEncap
->length
) {
1418 new->un_address
.addr_family
= AF_INET
;
1419 memcpy(&new->un_address
.addr
.v4
,
1421 have_vnc_tunnel_un
= 1;
1425 new->un_address
.addr_family
= AF_INET6
;
1426 memcpy(&new->un_address
.addr
.v6
,
1428 have_vnc_tunnel_un
= 1;
1433 "%s: invalid tunnel subtlv UN addr length (%d) for bi %p",
1434 __func__
, pEncap
->length
, bi
);
1440 "%s: unknown Encap Attribute option type %d",
1441 __func__
, pEncap
->type
);
1448 new->un_options
= rfapi_encap_tlv_to_un_option(bi
->attr
);
1450 #if DEBUG_ENCAP_MONITOR
1451 vnc_zlog_debug_verbose("%s: line %d: have_vnc_tunnel_un=%d",
1452 __func__
, __LINE__
, have_vnc_tunnel_un
);
1455 if (!have_vnc_tunnel_un
&& bi
&& bi
->extra
) {
1457 * use cached UN address from ENCAP route
1459 new->un_address
.addr_family
=
1460 bi
->extra
->vnc
.import
.un_family
;
1461 switch (new->un_address
.addr_family
) {
1463 new->un_address
.addr
.v4
=
1464 bi
->extra
->vnc
.import
.un
.addr4
;
1467 new->un_address
.addr
.v6
=
1468 bi
->extra
->vnc
.import
.un
.addr6
;
1472 "%s: invalid UN addr family (%d) for bi %p",
1473 __func__
, new->un_address
.addr_family
,
1475 rfapi_free_next_hop_list(new);
1482 new->lifetime
= lifetime
;
1486 int rfapiHasNonRemovedRoutes(struct route_node
*rn
)
1488 struct bgp_info
*bi
;
1490 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
1493 if (!CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
1494 && (bi
->extra
&& !rfapiGetUnAddrOfVpnBi(bi
, &pfx
))) {
1506 void rfapiDumpNode(struct route_node
*rn
)
1508 struct bgp_info
*bi
;
1510 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
1511 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
1513 int ctrc
= rfapiGetUnAddrOfVpnBi(bi
, &pfx
);
1516 if (!CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
1517 && (bi
->extra
&& !ctrc
)) {
1524 vnc_zlog_debug_verbose(
1525 " bi=%p, nr=%d, flags=0x%x, extra=%p, ctrc=%d", bi
, nr
,
1526 bi
->flags
, bi
->extra
, ctrc
);
1531 static int rfapiNhlAddNodeRoutes(
1532 struct route_node
*rn
, /* in */
1533 struct rfapi_ip_prefix
*rprefix
, /* in */
1534 uint32_t lifetime
, /* in */
1535 int removed
, /* in */
1536 struct rfapi_next_hop_entry
**head
, /* in/out */
1537 struct rfapi_next_hop_entry
**tail
, /* in/out */
1538 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1539 struct route_node
*rfd_rib_node
, /* preload this NVE rib node */
1540 struct prefix
*pfx_target_original
) /* query target */
1542 struct bgp_info
*bi
;
1543 struct rfapi_next_hop_entry
*new;
1544 struct prefix pfx_un
;
1545 struct skiplist
*seen_nexthops
;
1547 int is_l2
= (rn
->p
.family
== AF_ETHERNET
);
1549 if (rfapiRibFTDFilterRecentPrefix(
1550 (struct rfapi_descriptor
*)(rfd_rib_node
->table
->info
), rn
,
1551 pfx_target_original
)) {
1556 skiplist_new(0, vnc_prefix_cmp
, (void (*)(void *))prefix_free
);
1558 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
1560 struct prefix pfx_vn
;
1561 struct prefix
*newpfx
;
1563 if (removed
&& !CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)) {
1564 #if DEBUG_RETURNED_NHL
1565 vnc_zlog_debug_verbose(
1566 "%s: want holddown, this route not holddown, skip",
1571 if (!removed
&& CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)) {
1580 * Check for excluded VN address
1582 if (rfapiVpnBiNhEqualsPt(bi
, exclude_vnaddr
))
1586 * Check for VN address (nexthop) copied already
1589 /* L2 routes: semantic nexthop in aux_prefix; VN addr
1591 pfx_vn
= bi
->extra
->vnc
.import
.aux_prefix
;
1593 rfapiNexthop2Prefix(bi
->attr
, &pfx_vn
);
1595 if (!skiplist_search(seen_nexthops
, &pfx_vn
, NULL
)) {
1596 #if DEBUG_RETURNED_NHL
1599 prefix2str(&pfx_vn
, buf
, BUFSIZ
);
1600 buf
[BUFSIZ
- 1] = 0; /* guarantee NUL-terminated */
1601 vnc_zlog_debug_verbose(
1602 "%s: already put VN/nexthop %s, skip", __func__
,
1608 if (rfapiGetUnAddrOfVpnBi(bi
, &pfx_un
)) {
1609 #if DEBUG_ENCAP_MONITOR
1610 vnc_zlog_debug_verbose(
1611 "%s: failed to get UN address of this VPN bi",
1617 newpfx
= prefix_new();
1619 skiplist_insert(seen_nexthops
, newpfx
, newpfx
);
1621 new = rfapiRouteInfo2NextHopEntry(rprefix
, bi
, lifetime
, rn
);
1623 if (rfapiRibPreloadBi(rfd_rib_node
, &pfx_vn
, &pfx_un
,
1625 /* duplicate filtered by RIB */
1626 rfapi_free_next_hop_list(new);
1633 (*tail
)->next
= new;
1642 skiplist_free(seen_nexthops
);
1651 * omit_node is meant for the situation where we are adding a subtree
1652 * of a parent of some original requested node. The response already
1653 * contains the original requested node, and we don't want to duplicate
1654 * its routes in the list, so we skip it if the right or left node
1655 * matches (of course, we still travel down its child subtrees).
1657 static int rfapiNhlAddSubtree(
1658 struct route_node
*rn
, /* in */
1659 uint32_t lifetime
, /* in */
1660 struct rfapi_next_hop_entry
**head
, /* in/out */
1661 struct rfapi_next_hop_entry
**tail
, /* in/out */
1662 struct route_node
*omit_node
, /* in */
1663 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1664 struct route_table
*rfd_rib_table
, /* preload here */
1665 struct prefix
*pfx_target_original
) /* query target */
1667 struct rfapi_ip_prefix rprefix
;
1670 /* FIXME: need to find a better way here to work without sticking our
1671 * hands in node->link */
1672 if (rn
->l_left
&& rn
->l_left
!= omit_node
) {
1673 if (rn
->l_left
->info
) {
1675 struct route_node
*rib_rn
= NULL
;
1677 rfapiQprefix2Rprefix(&rn
->l_left
->p
, &rprefix
);
1678 if (rfd_rib_table
) {
1679 rib_rn
= route_node_get(rfd_rib_table
,
1683 count
= rfapiNhlAddNodeRoutes(
1684 rn
->l_left
, &rprefix
, lifetime
, 0, head
, tail
,
1685 exclude_vnaddr
, rib_rn
, pfx_target_original
);
1687 count
= rfapiNhlAddNodeRoutes(
1688 rn
->l_left
, &rprefix
, lifetime
, 1, head
,
1689 tail
, exclude_vnaddr
, rib_rn
,
1690 pfx_target_original
);
1694 route_unlock_node(rib_rn
);
1698 if (rn
->l_right
&& rn
->l_right
!= omit_node
) {
1699 if (rn
->l_right
->info
) {
1701 struct route_node
*rib_rn
= NULL
;
1703 rfapiQprefix2Rprefix(&rn
->l_right
->p
, &rprefix
);
1704 if (rfd_rib_table
) {
1705 rib_rn
= route_node_get(rfd_rib_table
,
1708 count
= rfapiNhlAddNodeRoutes(
1709 rn
->l_right
, &rprefix
, lifetime
, 0, head
, tail
,
1710 exclude_vnaddr
, rib_rn
, pfx_target_original
);
1712 count
= rfapiNhlAddNodeRoutes(
1713 rn
->l_right
, &rprefix
, lifetime
, 1,
1714 head
, tail
, exclude_vnaddr
, rib_rn
,
1715 pfx_target_original
);
1719 route_unlock_node(rib_rn
);
1724 rcount
+= rfapiNhlAddSubtree(
1725 rn
->l_left
, lifetime
, head
, tail
, omit_node
,
1726 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1729 rcount
+= rfapiNhlAddSubtree(
1730 rn
->l_right
, lifetime
, head
, tail
, omit_node
,
1731 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1738 * Implementation of ROUTE_LIST(node) from RFAPI-Import-Event-Handling.txt
1740 * Construct an rfapi nexthop list based on the routes attached to
1741 * the specified node.
1743 * If there are any routes that do NOT have BGP_INFO_REMOVED set,
1744 * return those only. If there are ONLY routes with BGP_INFO_REMOVED,
1745 * then return those, and also include all the non-removed routes from the
1746 * next less-specific node (i.e., this node's parent) at the end.
1748 struct rfapi_next_hop_entry
*rfapiRouteNode2NextHopList(
1749 struct route_node
*rn
, uint32_t lifetime
, /* put into nexthop entries */
1750 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1751 struct route_table
*rfd_rib_table
, /* preload here */
1752 struct prefix
*pfx_target_original
) /* query target */
1754 struct rfapi_ip_prefix rprefix
;
1755 struct rfapi_next_hop_entry
*answer
= NULL
;
1756 struct rfapi_next_hop_entry
*last
= NULL
;
1757 struct route_node
*parent
;
1759 struct route_node
*rib_rn
;
1761 #if DEBUG_RETURNED_NHL
1765 prefix2str(&rn
->p
, buf
, BUFSIZ
);
1766 buf
[BUFSIZ
- 1] = 0;
1767 vnc_zlog_debug_verbose("%s: called with node pfx=%s", __func__
,
1770 rfapiDebugBacktrace();
1773 rfapiQprefix2Rprefix(&rn
->p
, &rprefix
);
1775 rib_rn
= rfd_rib_table
? route_node_get(rfd_rib_table
, &rn
->p
) : NULL
;
1778 * Add non-withdrawn routes at this node
1780 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 0, &answer
, &last
,
1781 exclude_vnaddr
, rib_rn
,
1782 pfx_target_original
);
1785 * If the list has at least one entry, it's finished
1788 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
, &last
, NULL
,
1789 exclude_vnaddr
, rfd_rib_table
,
1790 pfx_target_original
);
1791 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
,
1793 #if DEBUG_RETURNED_NHL
1794 rfapiPrintNhl(NULL
, answer
);
1797 route_unlock_node(rib_rn
);
1802 * Add withdrawn routes at this node
1804 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 1, &answer
, &last
,
1805 exclude_vnaddr
, rib_rn
,
1806 pfx_target_original
);
1808 route_unlock_node(rib_rn
);
1810 // rfapiPrintNhl(NULL, answer);
1813 * walk up the tree until we find a node with non-deleted
1814 * routes, then add them
1816 for (parent
= rn
->parent
; parent
; parent
= parent
->parent
) {
1817 if (rfapiHasNonRemovedRoutes(parent
)) {
1823 * Add non-withdrawn routes from less-specific prefix
1826 rib_rn
= rfd_rib_table
1827 ? route_node_get(rfd_rib_table
, &parent
->p
)
1829 rfapiQprefix2Rprefix(&parent
->p
, &rprefix
);
1830 count
+= rfapiNhlAddNodeRoutes(parent
, &rprefix
, lifetime
, 0,
1831 &answer
, &last
, exclude_vnaddr
,
1832 rib_rn
, pfx_target_original
);
1833 count
+= rfapiNhlAddSubtree(parent
, lifetime
, &answer
, &last
,
1834 rn
, exclude_vnaddr
, rfd_rib_table
,
1835 pfx_target_original
);
1837 route_unlock_node(rib_rn
);
1840 * There is no parent with non-removed routes. Still need to
1841 * add subtree of original node if it contributed routes to the
1845 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
,
1846 &last
, rn
, exclude_vnaddr
,
1848 pfx_target_original
);
1851 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
, count
,
1853 #if DEBUG_RETURNED_NHL
1854 rfapiPrintNhl(NULL
, answer
);
1860 * Construct nexthop list of all routes in table
1862 struct rfapi_next_hop_entry
*rfapiRouteTable2NextHopList(
1863 struct route_table
*rt
,
1864 uint32_t lifetime
, /* put into nexthop entries */
1865 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1866 struct route_table
*rfd_rib_table
, /* preload this NVE rib table */
1867 struct prefix
*pfx_target_original
) /* query target */
1869 struct route_node
*rn
;
1870 struct rfapi_next_hop_entry
*biglist
= NULL
;
1871 struct rfapi_next_hop_entry
*nhl
;
1872 struct rfapi_next_hop_entry
*tail
= NULL
;
1875 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
1877 nhl
= rfapiRouteNode2NextHopList(rn
, lifetime
, exclude_vnaddr
,
1879 pfx_target_original
);
1881 tail
= biglist
= nhl
;
1888 while (tail
->next
) {
1895 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1899 struct rfapi_next_hop_entry
*rfapiEthRouteNode2NextHopList(
1900 struct route_node
*rn
, struct rfapi_ip_prefix
*rprefix
,
1901 uint32_t lifetime
, /* put into nexthop entries */
1902 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1903 struct route_table
*rfd_rib_table
, /* preload NVE rib table */
1904 struct prefix
*pfx_target_original
) /* query target */
1907 struct rfapi_next_hop_entry
*answer
= NULL
;
1908 struct rfapi_next_hop_entry
*last
= NULL
;
1909 struct route_node
*rib_rn
;
1911 rib_rn
= rfd_rib_table
? route_node_get(rfd_rib_table
, &rn
->p
) : NULL
;
1913 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 0, &answer
, &last
,
1914 NULL
, rib_rn
, pfx_target_original
);
1916 #if DEBUG_ENCAP_MONITOR
1917 vnc_zlog_debug_verbose("%s: node %p: %d non-holddown routes", __func__
,
1922 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 1, &answer
,
1923 &last
, exclude_vnaddr
, rib_rn
,
1924 pfx_target_original
);
1925 vnc_zlog_debug_verbose("%s: node %p: %d holddown routes",
1926 __func__
, rn
, count
);
1930 route_unlock_node(rib_rn
);
1932 #if DEBUG_RETURNED_NHL
1933 rfapiPrintNhl(NULL
, answer
);
1941 * Construct nexthop list of all routes in table
1943 struct rfapi_next_hop_entry
*rfapiEthRouteTable2NextHopList(
1944 uint32_t logical_net_id
, struct rfapi_ip_prefix
*rprefix
,
1945 uint32_t lifetime
, /* put into nexthop entries */
1946 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1947 struct route_table
*rfd_rib_table
, /* preload NVE rib node */
1948 struct prefix
*pfx_target_original
) /* query target */
1950 struct rfapi_import_table
*it
;
1951 struct bgp
*bgp
= bgp_get_default();
1952 struct route_table
*rt
;
1953 struct route_node
*rn
;
1954 struct rfapi_next_hop_entry
*biglist
= NULL
;
1955 struct rfapi_next_hop_entry
*nhl
;
1956 struct rfapi_next_hop_entry
*tail
= NULL
;
1960 it
= rfapiMacImportTableGet(bgp
, logical_net_id
);
1961 rt
= it
->imported_vpn
[AFI_L2VPN
];
1963 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
1965 nhl
= rfapiEthRouteNode2NextHopList(
1966 rn
, rprefix
, lifetime
, exclude_vnaddr
, rfd_rib_table
,
1967 pfx_target_original
);
1969 tail
= biglist
= nhl
;
1976 while (tail
->next
) {
1983 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1988 * Insert a new bi to the imported route table node,
1989 * keeping the list of BIs sorted best route first
1991 static void rfapiBgpInfoAttachSorted(struct route_node
*rn
,
1992 struct bgp_info
*info_new
, afi_t afi
,
1996 struct bgp_info
*prev
;
1997 struct bgp_info
*next
;
1998 char pfx_buf
[PREFIX2STR_BUFFER
];
2001 bgp
= bgp_get_default(); /* assume 1 instance for now */
2003 if (VNC_DEBUG(IMPORT_BI_ATTACH
)) {
2004 vnc_zlog_debug_verbose("%s: info_new->peer=%p", __func__
,
2006 vnc_zlog_debug_verbose("%s: info_new->peer->su_remote=%p",
2007 __func__
, info_new
->peer
->su_remote
);
2010 for (prev
= NULL
, next
= rn
->info
; next
;
2011 prev
= next
, next
= next
->next
) {
2012 if (!bgp
|| (!CHECK_FLAG(info_new
->flags
, BGP_INFO_REMOVED
)
2013 && CHECK_FLAG(next
->flags
, BGP_INFO_REMOVED
))
2014 || bgp_info_cmp_compatible(bgp
, info_new
, next
, pfx_buf
,
2016 == -1) { /* -1 if 1st is better */
2020 vnc_zlog_debug_verbose("%s: prev=%p, next=%p", __func__
, prev
, next
);
2022 prev
->next
= info_new
;
2024 rn
->info
= info_new
;
2026 info_new
->prev
= prev
;
2027 info_new
->next
= next
;
2029 next
->prev
= info_new
;
2030 bgp_attr_intern(info_new
->attr
);
2033 static void rfapiBgpInfoDetach(struct route_node
*rn
, struct bgp_info
*bi
)
2036 * Remove the route (doubly-linked)
2038 // bgp_attr_unintern (&bi->attr);
2040 bi
->next
->prev
= bi
->prev
;
2042 bi
->prev
->next
= bi
->next
;
2044 rn
->info
= bi
->next
;
2048 * For L3-indexed import tables
2050 static int rfapi_bi_peer_rd_cmp(void *b1
, void *b2
)
2052 struct bgp_info
*bi1
= b1
;
2053 struct bgp_info
*bi2
= b2
;
2058 if (bi1
->peer
< bi2
->peer
)
2060 if (bi1
->peer
> bi2
->peer
)
2066 return vnc_prefix_cmp((struct prefix
*)&bi1
->extra
->vnc
.import
.rd
,
2067 (struct prefix
*)&bi2
->extra
->vnc
.import
.rd
);
2071 * For L2-indexed import tables
2072 * The BIs in these tables should ALWAYS have an aux_prefix set because
2073 * they arrive via IPv4 or IPv6 advertisements.
2075 static int rfapi_bi_peer_rd_aux_cmp(void *b1
, void *b2
)
2077 struct bgp_info
*bi1
= b1
;
2078 struct bgp_info
*bi2
= b2
;
2084 if (bi1
->peer
< bi2
->peer
)
2086 if (bi1
->peer
> bi2
->peer
)
2092 rc
= vnc_prefix_cmp((struct prefix
*)&bi1
->extra
->vnc
.import
.rd
,
2093 (struct prefix
*)&bi2
->extra
->vnc
.import
.rd
);
2099 * L2 import tables can have multiple entries with the
2100 * same MAC address, same RD, but different L3 addresses.
2102 * Use presence of aux_prefix with AF=ethernet and prefixlen=1
2103 * as magic value to signify explicit wildcarding of the aux_prefix.
2104 * This magic value will not appear in bona fide bi entries in
2105 * the import table, but is allowed in the "fake" bi used to
2106 * probe the table when searching. (We have to test both b1 and b2
2107 * because there is no guarantee of the order the test key and
2108 * the real key will be passed)
2110 if ((bi1
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2111 && (bi1
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))
2112 || (bi2
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2113 && (bi2
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))) {
2116 * wildcard aux address specified
2121 return vnc_prefix_cmp(&bi1
->extra
->vnc
.import
.aux_prefix
,
2122 &bi2
->extra
->vnc
.import
.aux_prefix
);
2127 * Index on RD and Peer
2129 static void rfapiItBiIndexAdd(struct route_node
*rn
, /* Import table VPN node */
2130 struct bgp_info
*bi
) /* new BI */
2132 struct skiplist
*sl
;
2140 prefix_rd2str(&bi
->extra
->vnc
.import
.rd
, buf
, BUFSIZ
);
2141 vnc_zlog_debug_verbose("%s: bi %p, peer %p, rd %s", __func__
,
2145 sl
= RFAPI_RDINDEX_W_ALLOC(rn
);
2147 if (AF_ETHERNET
== rn
->p
.family
) {
2148 sl
= skiplist_new(0, rfapi_bi_peer_rd_aux_cmp
, NULL
);
2150 sl
= skiplist_new(0, rfapi_bi_peer_rd_cmp
, NULL
);
2152 RFAPI_IT_EXTRA_GET(rn
)->u
.vpn
.idx_rd
= sl
;
2153 route_lock_node(rn
); /* for skiplist */
2155 assert(!skiplist_insert(sl
, (void *)bi
, (void *)bi
));
2156 route_lock_node(rn
); /* for skiplist entry */
2158 /* NB: BIs in import tables are not refcounted */
2161 static void rfapiItBiIndexDump(struct route_node
*rn
)
2163 struct skiplist
*sl
;
2164 void *cursor
= NULL
;
2169 sl
= RFAPI_RDINDEX(rn
);
2173 for (rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
); !rc
;
2174 rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
)) {
2177 char buf_aux_pfx
[BUFSIZ
];
2179 prefix_rd2str(&k
->extra
->vnc
.import
.rd
, buf
, BUFSIZ
);
2181 if (k
->extra
->vnc
.import
.aux_prefix
.family
) {
2182 prefix2str(&k
->extra
->vnc
.import
.aux_prefix
,
2183 buf_aux_pfx
, BUFSIZ
);
2185 strncpy(buf_aux_pfx
, "(none)", BUFSIZ
);
2186 buf_aux_pfx
[BUFSIZ
- 1] = 0;
2189 vnc_zlog_debug_verbose("bi %p, peer %p, rd %s, aux_prefix %s",
2190 k
, k
->peer
, buf
, buf_aux_pfx
);
2194 static struct bgp_info
*rfapiItBiIndexSearch(
2195 struct route_node
*rn
, /* Import table VPN node */
2196 struct prefix_rd
*prd
, struct peer
*peer
,
2197 struct prefix
*aux_prefix
) /* optional L3 addr for L2 ITs */
2199 struct skiplist
*sl
;
2201 struct bgp_info bi_fake
;
2202 struct bgp_info_extra bi_extra
;
2203 struct bgp_info
*bi_result
;
2205 sl
= RFAPI_RDINDEX(rn
);
2212 char buf_aux_pfx
[BUFSIZ
];
2214 prefix_rd2str(prd
, buf
, BUFSIZ
);
2216 prefix2str(aux_prefix
, buf_aux_pfx
, BUFSIZ
);
2218 strncpy(buf_aux_pfx
, "(nil)", BUFSIZ
- 1);
2219 buf_aux_pfx
[BUFSIZ
- 1] = 0;
2222 vnc_zlog_debug_verbose("%s want prd=%s, peer=%p, aux_prefix=%s",
2223 __func__
, buf
, peer
, buf_aux_pfx
);
2224 rfapiItBiIndexDump(rn
);
2228 /* threshold is a WAG */
2229 if (sl
->count
< 3) {
2231 vnc_zlog_debug_verbose("%s: short list algorithm", __func__
);
2233 /* if short list, linear search might be faster */
2234 for (bi_result
= rn
->info
; bi_result
;
2235 bi_result
= bi_result
->next
) {
2239 prefix_rd2str(&bi_result
->extra
->vnc
.import
.rd
,
2241 vnc_zlog_debug_verbose(
2242 "%s: bi has prd=%s, peer=%p", __func__
,
2243 buf
, bi_result
->peer
);
2246 if (peer
== bi_result
->peer
2247 && !prefix_cmp((struct prefix
*)&bi_result
->extra
2249 (struct prefix
*)prd
)) {
2252 vnc_zlog_debug_verbose(
2253 "%s: peer and RD same, doing aux_prefix check",
2257 || !prefix_cmp(aux_prefix
,
2258 &bi_result
->extra
->vnc
.import
2262 vnc_zlog_debug_verbose("%s: match",
2272 bi_fake
.peer
= peer
;
2273 bi_fake
.extra
= &bi_extra
;
2274 bi_fake
.extra
->vnc
.import
.rd
= *(struct prefix_rd
*)prd
;
2276 bi_fake
.extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
2279 bi_fake
.extra
->vnc
.import
.aux_prefix
.family
= AF_ETHERNET
;
2280 bi_fake
.extra
->vnc
.import
.aux_prefix
.prefixlen
= 1;
2283 rc
= skiplist_search(sl
, (void *)&bi_fake
, (void *)&bi_result
);
2287 vnc_zlog_debug_verbose("%s: no match", __func__
);
2293 vnc_zlog_debug_verbose("%s: matched bi=%p", __func__
, bi_result
);
2299 static void rfapiItBiIndexDel(struct route_node
*rn
, /* Import table VPN node */
2300 struct bgp_info
*bi
) /* old BI */
2302 struct skiplist
*sl
;
2307 prefix_rd2str(&bi
->extra
->vnc
.import
.rd
, buf
, BUFSIZ
);
2308 vnc_zlog_debug_verbose("%s: bi %p, peer %p, rd %s", __func__
,
2312 sl
= RFAPI_RDINDEX(rn
);
2315 rc
= skiplist_delete(sl
, (void *)(bi
), (void *)bi
);
2317 rfapiItBiIndexDump(rn
);
2321 route_unlock_node(rn
); /* for skiplist entry */
2323 /* NB: BIs in import tables are not refcounted */
2327 * Add a backreference at the ENCAP node to the VPN route that
2330 static void rfapiMonitorEncapAdd(struct rfapi_import_table
*import_table
,
2331 struct prefix
*p
, /* VN address */
2332 struct route_node
*vpn_rn
, /* VPN node */
2333 struct bgp_info
*vpn_bi
) /* VPN bi/route */
2335 afi_t afi
= family2afi(p
->family
);
2336 struct route_node
*rn
;
2337 struct rfapi_monitor_encap
*m
;
2340 rn
= route_node_get(import_table
->imported_encap
[afi
],
2344 m
= XCALLOC(MTYPE_RFAPI_MONITOR_ENCAP
,
2345 sizeof(struct rfapi_monitor_encap
));
2352 /* insert to encap node's list */
2353 m
->next
= RFAPI_MONITOR_ENCAP(rn
);
2356 RFAPI_MONITOR_ENCAP_W_ALLOC(rn
) = m
;
2358 /* for easy lookup when deleting vpn route */
2359 vpn_bi
->extra
->vnc
.import
.hme
= m
;
2361 vnc_zlog_debug_verbose(
2362 "%s: it=%p, vpn_bi=%p, afi=%d, encap rn=%p, setting vpn_bi->extra->vnc.import.hme=%p",
2363 __func__
, import_table
, vpn_bi
, afi
, rn
, m
);
2365 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
2366 bgp_attr_intern(vpn_bi
->attr
);
2369 static void rfapiMonitorEncapDelete(struct bgp_info
*vpn_bi
)
2372 * Remove encap monitor
2374 vnc_zlog_debug_verbose("%s: vpn_bi=%p", __func__
, vpn_bi
);
2375 if (vpn_bi
->extra
) {
2376 struct rfapi_monitor_encap
*hme
= vpn_bi
->extra
->vnc
.import
.hme
;
2380 vnc_zlog_debug_verbose("%s: hme=%p", __func__
, hme
);
2382 /* Refcount checking takes too long here */
2383 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 0);
2385 hme
->next
->prev
= hme
->prev
;
2387 hme
->prev
->next
= hme
->next
;
2389 RFAPI_MONITOR_ENCAP_W_ALLOC(hme
->rn
) =
2391 /* Refcount checking takes too long here */
2392 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 1);
2394 /* see if the struct rfapi_it_extra is empty and can be
2396 rfapiMonitorExtraPrune(SAFI_ENCAP
, hme
->rn
);
2398 route_unlock_node(hme
->rn
); /* decr ref count */
2399 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, hme
);
2400 vpn_bi
->extra
->vnc
.import
.hme
= NULL
;
2406 * quagga lib/thread.h says this must return int even though
2407 * it doesn't do anything with the return value
2409 static int rfapiWithdrawTimerVPN(struct thread
*t
)
2411 struct rfapi_withdraw
*wcb
= t
->arg
;
2412 struct bgp_info
*bi
= wcb
->info
;
2413 struct bgp
*bgp
= bgp_get_default();
2415 struct rfapi_monitor_vpn
*moved
;
2420 assert(wcb
->import_table
);
2423 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, wcb
->lockoffset
);
2428 vnc_zlog_debug_verbose(
2429 "%s: removing bi %p at prefix %s/%d", __func__
, bi
,
2430 rfapi_ntop(wcb
->node
->p
.family
, &wcb
->node
->p
.u
.prefix
,
2432 wcb
->node
->p
.prefixlen
);
2436 * Remove the route (doubly-linked)
2438 if (CHECK_FLAG(bi
->flags
, BGP_INFO_VALID
)
2439 && VALID_INTERIOR_TYPE(bi
->type
))
2440 RFAPI_MONITOR_EXTERIOR(wcb
->node
)->valid_interior_count
--;
2442 afi
= family2afi(wcb
->node
->p
.family
);
2443 wcb
->import_table
->holddown_count
[afi
] -= 1; /* keep count consistent */
2444 rfapiItBiIndexDel(wcb
->node
, bi
);
2445 rfapiBgpInfoDetach(wcb
->node
, bi
); /* with removed bi */
2447 vnc_import_bgp_exterior_del_route_interior(bgp
, wcb
->import_table
,
2452 * If VNC is configured to send response remove messages, AND
2453 * if the removed route had a UN address, do response removal
2456 if (!(bgp
->rfapi_cfg
->flags
2457 & BGP_VNC_CONFIG_RESPONSE_REMOVAL_DISABLE
)) {
2459 int has_valid_duplicate
= 0;
2460 struct bgp_info
*bii
;
2463 * First check if there are any OTHER routes at this node
2464 * that have the same nexthop and a valid UN address. If
2465 * there are (e.g., from other peers), then the route isn't
2466 * really gone, so skip sending a response removal message.
2468 for (bii
= wcb
->node
->info
; bii
; bii
= bii
->next
) {
2469 if (rfapiVpnBiSamePtUn(bi
, bii
)) {
2470 has_valid_duplicate
= 1;
2475 vnc_zlog_debug_verbose("%s: has_valid_duplicate=%d", __func__
,
2476 has_valid_duplicate
);
2478 if (!has_valid_duplicate
) {
2479 rfapiRibPendingDeleteRoute(bgp
, wcb
->import_table
, afi
,
2484 rfapiMonitorEncapDelete(bi
);
2487 * If there are no VPN monitors at this VPN Node A,
2490 if (!RFAPI_MONITOR_VPN(wcb
->node
)) {
2491 vnc_zlog_debug_verbose("%s: no VPN monitors at this node",
2497 * rfapiMonitorMoveShorter only moves monitors if there are
2498 * no remaining valid routes at the current node
2500 moved
= rfapiMonitorMoveShorter(wcb
->node
, 1);
2503 rfapiMonitorMovedUp(wcb
->import_table
, wcb
->node
, moved
->node
,
2511 rfapiBgpInfoFree(bi
);
2515 * If route count at this node has gone to 0, withdraw exported prefix
2517 if (!wcb
->node
->info
) {
2518 /* see if the struct rfapi_it_extra is empty and can be freed */
2519 rfapiMonitorExtraPrune(SAFI_MPLS_VPN
, wcb
->node
);
2520 vnc_direct_bgp_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2521 vnc_zebra_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2524 * nexthop change event
2525 * vnc_direct_bgp_add_prefix() will recompute the VN addr
2528 vnc_direct_bgp_add_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2531 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, 1 + wcb
->lockoffset
);
2532 route_unlock_node(wcb
->node
); /* decr ref count */
2533 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2538 * This works for multiprotocol extension, but not for plain ol'
2539 * unicast IPv4 because that nexthop is stored in attr->nexthop
2541 void rfapiNexthop2Prefix(struct attr
*attr
, struct prefix
*p
)
2546 memset(p
, 0, sizeof(struct prefix
));
2548 switch (p
->family
= BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2550 p
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2555 p
->u
.prefix6
= attr
->mp_nexthop_global
;
2560 vnc_zlog_debug_verbose("%s: Family is unknown = %d", __func__
,
2565 void rfapiUnicastNexthop2Prefix(afi_t afi
, struct attr
*attr
, struct prefix
*p
)
2567 if (afi
== AFI_IP
) {
2568 p
->family
= AF_INET
;
2570 p
->u
.prefix4
= attr
->nexthop
;
2572 rfapiNexthop2Prefix(attr
, p
);
2576 static int rfapiAttrNexthopAddrDifferent(struct prefix
*p1
, struct prefix
*p2
)
2579 vnc_zlog_debug_verbose("%s: p1 or p2 is NULL", __func__
);
2584 * Are address families the same?
2586 if (p1
->family
!= p2
->family
) {
2590 switch (p1
->family
) {
2592 if (IPV4_ADDR_SAME(&p1
->u
.prefix4
, &p2
->u
.prefix4
))
2597 if (IPV6_ADDR_SAME(&p1
->u
.prefix6
, &p2
->u
.prefix6
))
2608 static void rfapiCopyUnEncap2VPN(struct bgp_info
*encap_bi
,
2609 struct bgp_info
*vpn_bi
)
2611 if (!encap_bi
->attr
) {
2612 zlog_warn("%s: no encap bi attr/extra, can't copy UN address",
2617 if (!vpn_bi
|| !vpn_bi
->extra
) {
2618 zlog_warn("%s: no vpn bi attr/extra, can't copy UN address",
2623 switch (BGP_MP_NEXTHOP_FAMILY(encap_bi
->attr
->mp_nexthop_len
)) {
2627 * instrumentation to debug segfault of 091127
2629 vnc_zlog_debug_verbose("%s: vpn_bi=%p", __func__
, vpn_bi
);
2631 vnc_zlog_debug_verbose("%s: vpn_bi->extra=%p", __func__
,
2635 vpn_bi
->extra
->vnc
.import
.un_family
= AF_INET
;
2636 vpn_bi
->extra
->vnc
.import
.un
.addr4
=
2637 encap_bi
->attr
->mp_nexthop_global_in
;
2641 vpn_bi
->extra
->vnc
.import
.un_family
= AF_INET6
;
2642 vpn_bi
->extra
->vnc
.import
.un
.addr6
=
2643 encap_bi
->attr
->mp_nexthop_global
;
2647 zlog_warn("%s: invalid encap nexthop length: %d", __func__
,
2648 encap_bi
->attr
->mp_nexthop_len
);
2649 vpn_bi
->extra
->vnc
.import
.un_family
= 0;
2655 * returns 0 on success, nonzero on error
2657 static int rfapiWithdrawEncapUpdateCachedUn(
2658 struct rfapi_import_table
*import_table
, struct bgp_info
*encap_bi
,
2659 struct route_node
*vpn_rn
, struct bgp_info
*vpn_bi
)
2664 * clear cached UN address
2666 if (!vpn_bi
|| !vpn_bi
->extra
) {
2668 "%s: missing VPN bi/extra, can't clear UN addr",
2672 vpn_bi
->extra
->vnc
.import
.un_family
= 0;
2673 memset(&vpn_bi
->extra
->vnc
.import
.un
, 0,
2674 sizeof(vpn_bi
->extra
->vnc
.import
.un
));
2675 if (CHECK_FLAG(vpn_bi
->flags
, BGP_INFO_VALID
)) {
2676 if (rfapiGetVncTunnelUnAddr(vpn_bi
->attr
, NULL
)) {
2677 UNSET_FLAG(vpn_bi
->flags
, BGP_INFO_VALID
);
2678 if (VALID_INTERIOR_TYPE(vpn_bi
->type
))
2679 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2680 ->valid_interior_count
--;
2681 /* signal interior route withdrawal to
2682 * import-exterior */
2683 vnc_import_bgp_exterior_del_route_interior(
2684 bgp_get_default(), import_table
, vpn_rn
,
2691 zlog_warn("%s: missing VPN bi, can't clear UN addr",
2695 rfapiCopyUnEncap2VPN(encap_bi
, vpn_bi
);
2696 if (!CHECK_FLAG(vpn_bi
->flags
, BGP_INFO_VALID
)) {
2697 SET_FLAG(vpn_bi
->flags
, BGP_INFO_VALID
);
2698 if (VALID_INTERIOR_TYPE(vpn_bi
->type
))
2699 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2700 ->valid_interior_count
++;
2701 /* signal interior route withdrawal to import-exterior
2703 vnc_import_bgp_exterior_add_route_interior(
2704 bgp_get_default(), import_table
, vpn_rn
,
2711 static int rfapiWithdrawTimerEncap(struct thread
*t
)
2713 struct rfapi_withdraw
*wcb
= t
->arg
;
2714 struct bgp_info
*bi
= wcb
->info
;
2715 int was_first_route
= 0;
2716 struct rfapi_monitor_encap
*em
;
2717 struct skiplist
*vpn_node_sl
= skiplist_new(0, NULL
, NULL
);
2721 assert(wcb
->import_table
);
2723 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 0);
2725 if (wcb
->node
->info
== bi
)
2726 was_first_route
= 1;
2729 * Remove the route/bi and free it
2731 rfapiBgpInfoDetach(wcb
->node
, bi
);
2732 rfapiBgpInfoFree(bi
);
2734 if (!was_first_route
)
2737 for (em
= RFAPI_MONITOR_ENCAP(wcb
->node
); em
; em
= em
->next
) {
2740 * Update monitoring VPN BIs with new encap info at the
2741 * head of the encap bi chain (which could be NULL after
2742 * removing the expiring bi above)
2744 if (rfapiWithdrawEncapUpdateCachedUn(wcb
->import_table
,
2745 wcb
->node
->info
, em
->node
,
2750 * Build a list of unique VPN nodes referenced by these
2752 * Use a skiplist for speed.
2754 skiplist_insert(vpn_node_sl
, em
->node
, em
->node
);
2759 * for each VPN node referenced in the ENCAP monitors:
2761 struct route_node
*rn
;
2762 while (!skiplist_first(vpn_node_sl
, (void **)&rn
, NULL
)) {
2763 if (!wcb
->node
->info
) {
2764 struct rfapi_monitor_vpn
*moved
;
2766 moved
= rfapiMonitorMoveShorter(rn
, 0);
2768 // rfapiDoRouteCallback(wcb->import_table,
2769 // moved->node, moved);
2770 rfapiMonitorMovedUp(wcb
->import_table
, rn
,
2771 moved
->node
, moved
);
2774 // rfapiDoRouteCallback(wcb->import_table, rn, NULL);
2775 rfapiMonitorItNodeChanged(wcb
->import_table
, rn
, NULL
);
2777 skiplist_delete_first(vpn_node_sl
);
2781 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 1);
2782 route_unlock_node(wcb
->node
); /* decr ref count */
2783 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2784 skiplist_free(vpn_node_sl
);
2790 * Works for both VPN and ENCAP routes; timer_service_func is different
2794 rfapiBiStartWithdrawTimer(struct rfapi_import_table
*import_table
,
2795 struct route_node
*rn
, struct bgp_info
*bi
, afi_t afi
,
2797 int (*timer_service_func
)(struct thread
*))
2800 struct rfapi_withdraw
*wcb
;
2803 CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
2806 * Already on the path to being withdrawn,
2807 * should already have a timer set up to
2810 vnc_zlog_debug_verbose(
2811 "%s: already being withdrawn, do nothing",
2816 rfapiGetVncLifetime(bi
->attr
, &lifetime
);
2817 vnc_zlog_debug_verbose("%s: VNC lifetime is %u", __func__
, lifetime
);
2820 * withdrawn routes get to hang around for a while
2822 SET_FLAG(bi
->flags
, BGP_INFO_REMOVED
);
2824 /* set timer to remove the route later */
2825 lifetime
= rfapiGetHolddownFromLifetime(lifetime
);
2826 vnc_zlog_debug_verbose("%s: using timeout %u", __func__
, lifetime
);
2829 * Stash import_table, node, and info for use by timer
2830 * service routine, which is supposed to free the wcb.
2832 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2836 wcb
->import_table
= import_table
;
2837 bgp_attr_intern(bi
->attr
);
2839 if (VNC_DEBUG(VERBOSE
)) {
2840 vnc_zlog_debug_verbose(
2841 "%s: wcb values: node=%p, info=%p, import_table=%p (bi follows)",
2842 __func__
, wcb
->node
, wcb
->info
, wcb
->import_table
);
2843 rfapiPrintBi(NULL
, bi
);
2848 if (lifetime
> UINT32_MAX
/ 1001) {
2849 /* sub-optimal case, but will probably never happen */
2850 bi
->extra
->vnc
.import
.timer
= NULL
;
2851 thread_add_timer(bm
->master
, timer_service_func
, wcb
, lifetime
,
2852 &bi
->extra
->vnc
.import
.timer
);
2854 static uint32_t jitter
;
2855 uint32_t lifetime_msec
;
2858 * the goal here is to spread out the timers so they are
2859 * sortable in the skip list
2861 if (++jitter
>= 1000)
2864 lifetime_msec
= (lifetime
* 1000) + jitter
;
2866 bi
->extra
->vnc
.import
.timer
= NULL
;
2867 thread_add_timer_msec(bm
->master
, timer_service_func
, wcb
,
2869 &bi
->extra
->vnc
.import
.timer
);
2872 /* re-sort route list (BGP_INFO_REMOVED routes are last) */
2873 if (((struct bgp_info
*)rn
->info
)->next
) {
2874 rfapiBgpInfoDetach(rn
, bi
);
2875 rfapiBgpInfoAttachSorted(rn
, bi
, afi
, safi
);
2880 typedef void(rfapi_bi_filtered_import_f
)(struct rfapi_import_table
*, int,
2881 struct peer
*, void *, struct prefix
*,
2882 struct prefix
*, afi_t
,
2883 struct prefix_rd
*, struct attr
*,
2884 u_char
, u_char
, uint32_t *);
2887 static void rfapiExpireEncapNow(struct rfapi_import_table
*it
,
2888 struct route_node
*rn
, struct bgp_info
*bi
)
2890 struct rfapi_withdraw
*wcb
;
2894 * pretend we're an expiring timer
2896 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2899 wcb
->import_table
= it
;
2900 memset(&t
, 0, sizeof(t
));
2902 rfapiWithdrawTimerEncap(&t
); /* frees wcb */
2905 static int rfapiGetNexthop(struct attr
*attr
, struct prefix
*prefix
)
2907 switch (BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2909 prefix
->family
= AF_INET
;
2910 prefix
->prefixlen
= 32;
2911 prefix
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2914 prefix
->family
= AF_INET6
;
2915 prefix
->prefixlen
= 128;
2916 prefix
->u
.prefix6
= attr
->mp_nexthop_global
;
2919 vnc_zlog_debug_verbose("%s: unknown attr->mp_nexthop_len %d",
2920 __func__
, attr
->mp_nexthop_len
);
2927 * import a bgp_info if its route target list intersects with the
2928 * import table's route target list
2930 static void rfapiBgpInfoFilteredImportEncap(
2931 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
2932 void *rfd
, /* set for looped back routes */
2934 struct prefix
*aux_prefix
, /* Unused for encap routes */
2935 afi_t afi
, struct prefix_rd
*prd
,
2936 struct attr
*attr
, /* part of bgp_info */
2937 u_char type
, /* part of bgp_info */
2938 u_char sub_type
, /* part of bgp_info */
2939 uint32_t *label
) /* part of bgp_info */
2941 struct route_table
*rt
= NULL
;
2942 struct route_node
*rn
;
2943 struct bgp_info
*info_new
;
2944 struct bgp_info
*bi
;
2945 struct bgp_info
*next
;
2948 struct prefix p_firstbi_old
;
2949 struct prefix p_firstbi_new
;
2951 const char *action_str
= NULL
;
2952 struct prefix un_prefix
;
2955 bgp
= bgp_get_default(); /* assume 1 instance for now */
2958 case FIF_ACTION_UPDATE
:
2959 action_str
= "update";
2961 case FIF_ACTION_WITHDRAW
:
2962 action_str
= "withdraw";
2964 case FIF_ACTION_KILL
:
2965 action_str
= "kill";
2972 vnc_zlog_debug_verbose(
2973 "%s: entry: %s: prefix %s/%d", __func__
, action_str
,
2974 inet_ntop(p
->family
, &p
->u
.prefix
, buf
, BUFSIZ
), p
->prefixlen
);
2976 memset(&p_firstbi_old
, 0, sizeof(p_firstbi_old
));
2977 memset(&p_firstbi_new
, 0, sizeof(p_firstbi_new
));
2979 if (action
== FIF_ACTION_UPDATE
) {
2981 * Compare rt lists. If no intersection, don't import this route
2982 * On a withdraw, peer and RD are sufficient to determine if
2985 if (!attr
|| !attr
->ecommunity
) {
2987 vnc_zlog_debug_verbose(
2988 "%s: attr, extra, or ecommunity missing, not importing",
2992 #if RFAPI_REQUIRE_ENCAP_BEEC
2993 if (!rfapiEcommunitiesMatchBeec(attr
->ecommunity
)) {
2994 vnc_zlog_debug_verbose(
2995 "%s: it=%p: no match for BGP Encapsulation ecommunity",
2996 __func__
, import_table
);
3000 if (!rfapiEcommunitiesIntersect(import_table
->rt_import_list
,
3001 attr
->ecommunity
)) {
3003 vnc_zlog_debug_verbose(
3004 "%s: it=%p: no ecommunity intersection",
3005 __func__
, import_table
);
3010 * Updates must also have a nexthop address
3012 memset(&un_prefix
, 0,
3013 sizeof(un_prefix
)); /* keep valgrind happy */
3014 if (rfapiGetNexthop(attr
, &un_prefix
)) {
3015 vnc_zlog_debug_verbose("%s: missing nexthop address",
3022 * Figure out which radix tree the route would go into
3027 rt
= import_table
->imported_encap
[afi
];
3031 zlog_err("%s: bad afi %d", __func__
, afi
);
3036 * route_node_lookup returns a node only if there is at least
3037 * one route attached.
3039 rn
= route_node_lookup(rt
, p
);
3041 #if DEBUG_ENCAP_MONITOR
3042 vnc_zlog_debug_verbose("%s: initial encap lookup(it=%p) rn=%p",
3043 __func__
, import_table
, rn
);
3048 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 1);
3049 route_unlock_node(rn
); /* undo lock in route_node_lookup */
3053 * capture nexthop of first bi
3056 rfapiNexthop2Prefix(
3057 ((struct bgp_info
*)(rn
->info
))->attr
,
3061 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
3064 * Does this bgp_info refer to the same route
3065 * as we are trying to add?
3067 vnc_zlog_debug_verbose("%s: comparing BI %p", __func__
,
3074 * RD of import table bi is in bi->extra->vnc.import.rd
3075 * RD of info_orig is in prd
3078 vnc_zlog_debug_verbose("%s: no bi->extra",
3083 (struct prefix
*)&bi
->extra
->vnc
.import
.rd
,
3084 (struct prefix
*)prd
)) {
3086 vnc_zlog_debug_verbose("%s: prd does not match",
3094 if (bi
->peer
!= peer
) {
3095 vnc_zlog_debug_verbose(
3096 "%s: peer does not match", __func__
);
3100 vnc_zlog_debug_verbose("%s: found matching bi",
3103 /* Same route. Delete this bi, replace with new one */
3105 if (action
== FIF_ACTION_WITHDRAW
) {
3107 vnc_zlog_debug_verbose(
3108 "%s: withdrawing at prefix %s/%d",
3110 inet_ntop(rn
->p
.family
, &rn
->p
.u
.prefix
,
3114 rfapiBiStartWithdrawTimer(
3115 import_table
, rn
, bi
, afi
, SAFI_ENCAP
,
3116 rfapiWithdrawTimerEncap
);
3119 vnc_zlog_debug_verbose(
3120 "%s: %s at prefix %s/%d", __func__
,
3121 ((action
== FIF_ACTION_KILL
)
3124 inet_ntop(rn
->p
.family
, &rn
->p
.u
.prefix
,
3129 * If this route is waiting to be deleted
3131 * a previous withdraw, we must cancel its
3134 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
3135 && bi
->extra
->vnc
.import
.timer
) {
3138 (struct thread
*)bi
->extra
->vnc
3140 struct rfapi_withdraw
*wcb
= t
->arg
;
3142 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3146 if (action
== FIF_ACTION_UPDATE
) {
3147 rfapiBgpInfoDetach(rn
, bi
);
3148 rfapiBgpInfoFree(bi
);
3152 * Kill: do export stuff when removing
3155 struct rfapi_withdraw
*wcb
;
3159 * pretend we're an expiring timer
3162 MTYPE_RFAPI_WITHDRAW
,
3163 sizeof(struct rfapi_withdraw
));
3166 wcb
->import_table
= import_table
;
3167 memset(&t
, 0, sizeof(t
));
3169 rfapiWithdrawTimerEncap(
3170 &t
); /* frees wcb */
3179 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, replacing
? 1 : 0);
3181 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
)
3185 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, NULL
);
3189 route_lock_node(rn
); /* incr ref count for new BI */
3191 rn
= route_node_get(rt
, p
);
3194 vnc_zlog_debug_verbose(
3195 "%s: (afi=%d, rn=%p) inserting at prefix %s/%d", __func__
, afi
,
3196 rn
, inet_ntop(rn
->p
.family
, &rn
->p
.u
.prefix
, buf
, BUFSIZ
),
3199 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_ENCAP
);
3202 * Delete holddown routes from same NVE. See details in
3203 * rfapiBgpInfoFilteredImportVPN()
3205 for (bi
= info_new
->next
; bi
; bi
= next
) {
3207 struct prefix pfx_un
;
3211 if (!CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
))
3215 * We already match the VN address (it is the prefix
3216 * of the route node)
3219 if (!rfapiGetNexthop(bi
->attr
, &pfx_un
)
3220 && prefix_same(&pfx_un
, &un_prefix
)) {
3228 vnc_zlog_debug_verbose(
3229 "%s: removing holddown bi matching NVE of new route",
3231 if (bi
->extra
->vnc
.import
.timer
) {
3233 (struct thread
*)bi
->extra
->vnc
.import
.timer
;
3234 struct rfapi_withdraw
*wcb
= t
->arg
;
3236 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3239 rfapiExpireEncapNow(import_table
, rn
, bi
);
3242 rfapiNexthop2Prefix(((struct bgp_info
*)(rn
->info
))->attr
,
3246 * If the nexthop address of the selected Encap route (i.e.,
3247 * the UN address) has changed, then we must update the VPN
3248 * routes that refer to this Encap route and possibly force
3251 if (rfapiAttrNexthopAddrDifferent(&p_firstbi_old
, &p_firstbi_new
)) {
3253 struct rfapi_monitor_encap
*m
;
3254 struct rfapi_monitor_encap
*mnext
;
3256 struct route_node
*referenced_vpn_prefix
;
3259 * Optimized approach: build radix tree on the fly to
3260 * hold list of VPN nodes referenced by the ENCAP monitors
3262 * The nodes in this table correspond to prefixes of VPN routes.
3263 * The "info" pointer of the node points to a chain of
3264 * struct rfapi_monitor_encap, each of which refers to a
3265 * specific VPN node.
3267 struct route_table
*referenced_vpn_table
;
3269 referenced_vpn_table
= route_table_init();
3270 assert(referenced_vpn_table
);
3273 * iterate over the set of monitors at this ENCAP node.
3275 #if DEBUG_ENCAP_MONITOR
3276 vnc_zlog_debug_verbose("%s: examining monitors at rn=%p",
3279 for (m
= RFAPI_MONITOR_ENCAP(rn
); m
; m
= m
->next
) {
3282 * For each referenced bi/route, copy the ENCAP route's
3283 * nexthop to the VPN route's cached UN address field
3285 * the address family of the cached UN address field.
3287 rfapiCopyUnEncap2VPN(info_new
, m
->bi
);
3288 if (!CHECK_FLAG(m
->bi
->flags
, BGP_INFO_VALID
)) {
3289 SET_FLAG(m
->bi
->flags
, BGP_INFO_VALID
);
3290 if (VALID_INTERIOR_TYPE(m
->bi
->type
))
3291 RFAPI_MONITOR_EXTERIOR(m
->node
)
3292 ->valid_interior_count
++;
3293 vnc_import_bgp_exterior_add_route_interior(
3294 bgp
, import_table
, m
->node
, m
->bi
);
3298 * Build a list of unique VPN nodes referenced by these
3301 * There could be more than one VPN node here with a
3303 * prefix. Those are currently in an unsorted linear
3308 referenced_vpn_prefix
= route_node_get(
3309 referenced_vpn_table
, &m
->node
->p
);
3310 assert(referenced_vpn_prefix
);
3311 for (mnext
= referenced_vpn_prefix
->info
; mnext
;
3312 mnext
= mnext
->next
) {
3314 if (mnext
->node
== m
->node
)
3320 * already have an entry for this VPN node
3322 route_unlock_node(referenced_vpn_prefix
);
3325 MTYPE_RFAPI_MONITOR_ENCAP
,
3326 sizeof(struct rfapi_monitor_encap
));
3328 mnext
->node
= m
->node
;
3329 mnext
->next
= referenced_vpn_prefix
->info
;
3330 referenced_vpn_prefix
->info
= mnext
;
3335 * for each VPN node referenced in the ENCAP monitors:
3337 for (referenced_vpn_prefix
= route_top(referenced_vpn_table
);
3338 referenced_vpn_prefix
; referenced_vpn_prefix
= route_next(
3339 referenced_vpn_prefix
)) {
3341 while ((m
= referenced_vpn_prefix
->info
)) {
3343 struct route_node
*n
;
3345 rfapiMonitorMoveLonger(m
->node
);
3346 for (n
= m
->node
; n
; n
= n
->parent
) {
3347 // rfapiDoRouteCallback(import_table, n,
3350 rfapiMonitorItNodeChanged(import_table
, m
->node
,
3353 referenced_vpn_prefix
->info
= m
->next
;
3354 route_unlock_node(referenced_vpn_prefix
);
3355 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, m
);
3358 route_table_finish(referenced_vpn_table
);
3361 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
3364 static void rfapiExpireVpnNow(struct rfapi_import_table
*it
,
3365 struct route_node
*rn
, struct bgp_info
*bi
,
3368 struct rfapi_withdraw
*wcb
;
3372 * pretend we're an expiring timer
3374 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
3377 wcb
->import_table
= it
;
3378 wcb
->lockoffset
= lockoffset
;
3379 memset(&t
, 0, sizeof(t
));
3381 rfapiWithdrawTimerVPN(&t
); /* frees wcb */
3386 * import a bgp_info if its route target list intersects with the
3387 * import table's route target list
3389 void rfapiBgpInfoFilteredImportVPN(
3390 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3391 void *rfd
, /* set for looped back routes */
3393 struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3394 afi_t afi
, struct prefix_rd
*prd
,
3395 struct attr
*attr
, /* part of bgp_info */
3396 u_char type
, /* part of bgp_info */
3397 u_char sub_type
, /* part of bgp_info */
3398 uint32_t *label
) /* part of bgp_info */
3400 struct route_table
*rt
= NULL
;
3401 struct route_node
*rn
;
3402 struct route_node
*n
;
3403 struct bgp_info
*info_new
;
3404 struct bgp_info
*bi
;
3405 struct bgp_info
*next
;
3407 struct prefix vn_prefix
;
3408 struct prefix un_prefix
;
3409 int un_prefix_valid
= 0;
3410 struct route_node
*ern
;
3412 int original_had_routes
= 0;
3413 struct prefix original_nexthop
;
3414 const char *action_str
= NULL
;
3418 bgp
= bgp_get_default(); /* assume 1 instance for now */
3421 case FIF_ACTION_UPDATE
:
3422 action_str
= "update";
3424 case FIF_ACTION_WITHDRAW
:
3425 action_str
= "withdraw";
3427 case FIF_ACTION_KILL
:
3428 action_str
= "kill";
3435 if (import_table
== bgp
->rfapi
->it_ce
)
3438 vnc_zlog_debug_verbose("%s: entry: %s%s: prefix %s/%d: it %p, afi %s",
3439 __func__
, (is_it_ce
? "CE-IT " : ""), action_str
,
3440 rfapi_ntop(p
->family
, &p
->u
.prefix
, buf
, BUFSIZ
),
3441 p
->prefixlen
, import_table
, afi2str(afi
));
3446 * Compare rt lists. If no intersection, don't import this route
3447 * On a withdraw, peer and RD are sufficient to determine if
3450 if (action
== FIF_ACTION_UPDATE
) {
3451 if (!attr
|| !attr
->ecommunity
) {
3453 vnc_zlog_debug_verbose(
3454 "%s: attr, extra, or ecommunity missing, not importing",
3458 if ((import_table
!= bgp
->rfapi
->it_ce
)
3459 && !rfapiEcommunitiesIntersect(import_table
->rt_import_list
,
3460 attr
->ecommunity
)) {
3462 vnc_zlog_debug_verbose(
3463 "%s: it=%p: no ecommunity intersection",
3464 __func__
, import_table
);
3468 memset(&vn_prefix
, 0,
3469 sizeof(vn_prefix
)); /* keep valgrind happy */
3470 if (rfapiGetNexthop(attr
, &vn_prefix
)) {
3471 /* missing nexthop address would be a bad, bad thing */
3472 vnc_zlog_debug_verbose("%s: missing nexthop", __func__
);
3478 * Figure out which radix tree the route would go into
3484 rt
= import_table
->imported_vpn
[afi
];
3488 zlog_err("%s: bad afi %d", __func__
, afi
);
3493 memset(&original_nexthop
, 0, sizeof(original_nexthop
));
3496 * route_node_lookup returns a node only if there is at least
3497 * one route attached.
3499 rn
= route_node_lookup(rt
, p
);
3501 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
3505 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
3506 route_unlock_node(rn
); /* undo lock in route_node_lookup */
3509 original_had_routes
= 1;
3511 if (VNC_DEBUG(VERBOSE
)) {
3512 vnc_zlog_debug_verbose("%s: showing IT node on entry",
3514 rfapiShowItNode(NULL
, rn
); /* debug */
3518 * Look for same route (will have same RD and peer)
3520 bi
= rfapiItBiIndexSearch(rn
, prd
, peer
, aux_prefix
);
3525 * This was an old test when we iterated over the
3526 * BIs linearly. Since we're now looking up with
3527 * RD and peer, comparing types should not be
3528 * needed. Changed to assertion.
3530 * Compare types. Doing so prevents a RFP-originated
3531 * route from matching an imported route, for example.
3533 if (VNC_DEBUG(VERBOSE
) && bi
->type
!= type
)
3534 /* should be handled by RDs, but warn for now */
3535 zlog_warn("%s: type mismatch! (bi=%d, arg=%d)",
3536 __func__
, bi
->type
, type
);
3538 vnc_zlog_debug_verbose("%s: found matching bi",
3542 * In the special CE table, withdrawals occur without
3545 if (import_table
== bgp
->rfapi
->it_ce
) {
3546 vnc_direct_bgp_del_route_ce(bgp
, rn
, bi
);
3547 if (action
== FIF_ACTION_WITHDRAW
)
3548 action
= FIF_ACTION_KILL
;
3551 if (action
== FIF_ACTION_WITHDRAW
) {
3554 CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
);
3556 vnc_zlog_debug_verbose(
3557 "%s: withdrawing at prefix %s/%d%s",
3558 __func__
, rfapi_ntop(rn
->p
.family
,
3563 ? " (already being withdrawn)"
3568 rfapiBiStartWithdrawTimer(
3569 import_table
, rn
, bi
, afi
,
3571 rfapiWithdrawTimerVPN
);
3573 RFAPI_UPDATE_ITABLE_COUNT(
3574 bi
, import_table
, afi
, -1);
3575 import_table
->holddown_count
[afi
] += 1;
3579 vnc_zlog_debug_verbose(
3580 "%s: %s at prefix %s/%d", __func__
,
3581 ((action
== FIF_ACTION_KILL
)
3584 rfapi_ntop(rn
->p
.family
,
3585 &rn
->p
.u
.prefix
, buf
,
3590 * If this route is waiting to be deleted
3592 * a previous withdraw, we must cancel its
3595 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)
3596 && bi
->extra
->vnc
.import
.timer
) {
3599 (struct thread
*)bi
->extra
->vnc
3601 struct rfapi_withdraw
*wcb
= t
->arg
;
3603 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3606 import_table
->holddown_count
[afi
] -= 1;
3607 RFAPI_UPDATE_ITABLE_COUNT(
3608 bi
, import_table
, afi
, 1);
3611 * decrement remote count (if route is remote)
3613 * we are going to remove it below
3615 RFAPI_UPDATE_ITABLE_COUNT(bi
, import_table
, afi
,
3617 if (action
== FIF_ACTION_UPDATE
) {
3621 * make copy of original nexthop so we
3622 * can see if it changed
3624 rfapiGetNexthop(bi
->attr
,
3628 * remove bi without doing any export
3631 if (CHECK_FLAG(bi
->flags
,
3633 && VALID_INTERIOR_TYPE(bi
->type
))
3634 RFAPI_MONITOR_EXTERIOR(rn
)
3635 ->valid_interior_count
--;
3636 rfapiItBiIndexDel(rn
, bi
);
3637 rfapiBgpInfoDetach(rn
, bi
);
3638 rfapiMonitorEncapDelete(bi
);
3639 vnc_import_bgp_exterior_del_route_interior(
3640 bgp
, import_table
, rn
, bi
);
3641 rfapiBgpInfoFree(bi
);
3645 * remove bi and do export processing
3647 import_table
->holddown_count
[afi
] += 1;
3648 rfapiExpireVpnNow(import_table
, rn
, bi
,
3656 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, replacing
? 1 : 0);
3658 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
) {
3664 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, label
);
3667 * lookup un address in encap table
3669 ern
= route_node_match(import_table
->imported_encap
[afi
], &vn_prefix
);
3671 rfapiCopyUnEncap2VPN(ern
->info
, info_new
);
3672 route_unlock_node(ern
); /* undo lock in route_note_match */
3675 prefix2str(&vn_prefix
, buf
, sizeof(buf
));
3676 buf
[BUFSIZ
- 1] = 0;
3677 /* Not a big deal, just means VPN route got here first */
3678 vnc_zlog_debug_verbose("%s: no encap route for vn addr %s",
3680 info_new
->extra
->vnc
.import
.un_family
= 0;
3685 route_lock_node(rn
);
3688 * No need to increment reference count, so only "get"
3689 * if the node is not there already
3691 rn
= route_node_get(rt
, p
);
3695 * For ethernet routes, if there is an accompanying IP address,
3698 if ((AFI_L2VPN
== afi
) && aux_prefix
) {
3700 vnc_zlog_debug_verbose("%s: setting BI's aux_prefix", __func__
);
3701 info_new
->extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
3704 vnc_zlog_debug_verbose(
3705 "%s: inserting bi %p at prefix %s/%d #%d", __func__
, info_new
,
3706 rfapi_ntop(rn
->p
.family
, &rn
->p
.u
.prefix
, buf
, BUFSIZ
),
3707 rn
->p
.prefixlen
, rn
->lock
);
3709 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_MPLS_VPN
);
3710 rfapiItBiIndexAdd(rn
, info_new
);
3711 if (!rfapiGetUnAddrOfVpnBi(info_new
, NULL
)) {
3712 if (VALID_INTERIOR_TYPE(info_new
->type
))
3713 RFAPI_MONITOR_EXTERIOR(rn
)->valid_interior_count
++;
3714 SET_FLAG(info_new
->flags
, BGP_INFO_VALID
);
3716 RFAPI_UPDATE_ITABLE_COUNT(info_new
, import_table
, afi
, 1);
3717 vnc_import_bgp_exterior_add_route_interior(bgp
, import_table
, rn
,
3720 if (import_table
== bgp
->rfapi
->it_ce
)
3721 vnc_direct_bgp_add_route_ce(bgp
, rn
, info_new
);
3723 if (VNC_DEBUG(VERBOSE
)) {
3724 vnc_zlog_debug_verbose("%s: showing IT node", __func__
);
3725 rfapiShowItNode(NULL
, rn
); /* debug */
3728 rfapiMonitorEncapAdd(import_table
, &vn_prefix
, rn
, info_new
);
3730 if (!rfapiGetUnAddrOfVpnBi(info_new
, &un_prefix
)) {
3733 * if we have a valid UN address (either via Encap route
3734 * or via tunnel attribute), then we should attempt
3735 * to move any monitors at less-specific nodes to this node
3737 rfapiMonitorMoveLonger(rn
);
3739 un_prefix_valid
= 1;
3743 * 101129 Enhancement: if we add a route (implication: it is not
3744 * in holddown), delete all other routes from this nve at this
3745 * node that are in holddown, regardless of peer.
3747 * Reasons it's OK to do that:
3749 * - if the holddown route being deleted originally came from BGP VPN,
3750 * it is already gone from BGP (implication of holddown), so there
3751 * won't be any added inconsistency with the BGP RIB.
3753 * - once a fresh route is added at a prefix, any routes in holddown
3754 * at that prefix will not show up in RFP responses, so deleting
3755 * the holddown routes won't affect the contents of responses.
3757 * - lifetimes are supposed to be consistent, so there should not
3758 * be a case where the fresh route has a shorter lifetime than
3759 * the holddown route, so we don't expect the fresh route to
3760 * disappear and complete its holddown time before the existing
3761 * holddown routes time out. Therefore, we won't have a situation
3762 * where we expect the existing holddown routes to be hidden and
3763 * then to reappear sometime later (as holddown routes) in a
3766 * Among other things, this would enable us to skirt the problem
3767 * of local holddown routes that refer to NVE descriptors that
3768 * have already been closed (if the same NVE triggers a subsequent
3769 * rfapi_open(), the new peer is different and doesn't match the
3770 * peer of the holddown route, so the stale holddown route still
3771 * hangs around until it times out instead of just being replaced
3772 * by the fresh route).
3775 * We know that the new bi will have been inserted before any routes
3776 * in holddown, so we can skip any that came before it
3778 for (bi
= info_new
->next
; bi
; bi
= next
) {
3780 struct prefix pfx_vn
;
3781 struct prefix pfx_un
;
3783 int remote_peer_match
= 0;
3790 if (!CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
))
3794 * Must match VN address (nexthop of VPN route)
3796 if (rfapiGetNexthop(bi
->attr
, &pfx_vn
))
3798 if (!prefix_same(&pfx_vn
, &vn_prefix
))
3801 if (un_prefix_valid
&& /* new route UN addr */
3802 !rfapiGetUnAddrOfVpnBi(bi
, &pfx_un
)
3803 && /* old route UN addr */
3804 prefix_same(&pfx_un
, &un_prefix
)) { /* compare */
3807 if (!RFAPI_LOCAL_BI(bi
) && !RFAPI_LOCAL_BI(info_new
)
3808 && sockunion_same(&bi
->peer
->su
, &info_new
->peer
->su
)) {
3809 /* old & new are both remote, same peer */
3810 remote_peer_match
= 1;
3813 if (!un_match
& !remote_peer_match
)
3816 vnc_zlog_debug_verbose(
3817 "%s: removing holddown bi matching NVE of new route",
3819 if (bi
->extra
->vnc
.import
.timer
) {
3821 (struct thread
*)bi
->extra
->vnc
.import
.timer
;
3822 struct rfapi_withdraw
*wcb
= t
->arg
;
3824 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3827 rfapiExpireVpnNow(import_table
, rn
, bi
, 0);
3830 if (!original_had_routes
) {
3832 * We went from 0 usable routes to 1 usable route. Perform the
3833 * "Adding a Route" export process.
3835 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3836 vnc_zebra_add_prefix(bgp
, import_table
, rn
);
3839 * Check for nexthop change event
3840 * Note: the prefix_same() test below detects two situations:
3841 * 1. route is replaced, new route has different nexthop
3842 * 2. new route is added (original_nexthop is 0)
3844 struct prefix new_nexthop
;
3846 rfapiGetNexthop(attr
, &new_nexthop
);
3847 if (!prefix_same(&original_nexthop
, &new_nexthop
)) {
3849 * nexthop change event
3850 * vnc_direct_bgp_add_prefix() will recompute VN addr
3853 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3857 if (!(bgp
->rfapi_cfg
->flags
& BGP_VNC_CONFIG_CALLBACK_DISABLE
)) {
3858 for (n
= rn
; n
; n
= n
->parent
) {
3859 // rfapiDoRouteCallback(import_table, n, NULL);
3861 rfapiMonitorItNodeChanged(import_table
, rn
, NULL
);
3863 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 0);
3867 static void rfapiBgpInfoFilteredImportBadSafi(
3868 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3869 void *rfd
, /* set for looped back routes */
3871 struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3872 afi_t afi
, struct prefix_rd
*prd
,
3873 struct attr
*attr
, /* part of bgp_info */
3874 u_char type
, /* part of bgp_info */
3875 u_char sub_type
, /* part of bgp_info */
3876 uint32_t *label
) /* part of bgp_info */
3878 vnc_zlog_debug_verbose("%s: Error, bad safi", __func__
);
3881 static rfapi_bi_filtered_import_f
*
3882 rfapiBgpInfoFilteredImportFunction(safi_t safi
)
3886 return rfapiBgpInfoFilteredImportVPN
;
3889 return rfapiBgpInfoFilteredImportEncap
;
3893 zlog_err("%s: bad safi %d", __func__
, safi
);
3894 return rfapiBgpInfoFilteredImportBadSafi
;
3898 void rfapiProcessUpdate(struct peer
*peer
,
3899 void *rfd
, /* set when looped from RFP/RFAPI */
3900 struct prefix
*p
, struct prefix_rd
*prd
,
3901 struct attr
*attr
, afi_t afi
, safi_t safi
, u_char type
,
3902 u_char sub_type
, uint32_t *label
)
3906 struct rfapi_import_table
*it
;
3907 int has_ip_route
= 1;
3910 bgp
= bgp_get_default(); /* assume 1 instance for now */
3917 * look at high-order byte of RD. FF means MAC
3918 * address is present (VNC L2VPN)
3920 if ((safi
== SAFI_MPLS_VPN
)
3921 && (decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
)) {
3922 struct prefix pfx_mac_buf
;
3923 struct prefix pfx_nexthop_buf
;
3927 * Set flag if prefix and nexthop are the same - don't
3928 * add the route to normal IP-based import tables
3930 if (!rfapiGetNexthop(attr
, &pfx_nexthop_buf
)) {
3931 if (!prefix_cmp(&pfx_nexthop_buf
, p
)) {
3936 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
3937 pfx_mac_buf
.family
= AF_ETHERNET
;
3938 pfx_mac_buf
.prefixlen
= 48;
3939 memcpy(&pfx_mac_buf
.u
.prefix_eth
.octet
, prd
->val
+ 2, 6);
3942 * Find rt containing LNI (Logical Network ID), which
3943 * _should_ always be present when mac address is present
3945 rc
= rfapiEcommunityGetLNI(attr
->ecommunity
, &lni
);
3947 vnc_zlog_debug_verbose(
3948 "%s: rfapiEcommunityGetLNI returned %d, lni=%d, attr=%p",
3949 __func__
, rc
, lni
, attr
);
3951 it
= rfapiMacImportTableGet(bgp
, lni
);
3953 rfapiBgpInfoFilteredImportVPN(
3954 it
, FIF_ACTION_UPDATE
, peer
, rfd
,
3955 &pfx_mac_buf
, /* prefix */
3956 p
, /* aux prefix: IP addr */
3957 AFI_L2VPN
, prd
, attr
, type
, sub_type
, label
);
3965 * Iterate over all import tables; do a filtered import
3966 * for the afi/safi combination
3968 for (it
= h
->imports
; it
; it
= it
->next
) {
3969 (*rfapiBgpInfoFilteredImportFunction(safi
))(
3970 it
, FIF_ACTION_UPDATE
, peer
, rfd
, p
, /* prefix */
3971 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3974 if (safi
== SAFI_MPLS_VPN
) {
3975 vnc_direct_bgp_rh_add_route(bgp
, afi
, p
, peer
, attr
);
3978 if (safi
== SAFI_MPLS_VPN
) {
3979 rfapiBgpInfoFilteredImportVPN(
3980 bgp
->rfapi
->it_ce
, FIF_ACTION_UPDATE
, peer
, rfd
,
3982 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3987 void rfapiProcessWithdraw(struct peer
*peer
, void *rfd
, struct prefix
*p
,
3988 struct prefix_rd
*prd
, struct attr
*attr
, afi_t afi
,
3989 safi_t safi
, u_char type
, int kill
)
3993 struct rfapi_import_table
*it
;
3995 bgp
= bgp_get_default(); /* assume 1 instance for now */
4002 * look at high-order byte of RD. FF means MAC
4003 * address is present (VNC L2VPN)
4005 if (h
->import_mac
!= NULL
&& safi
== SAFI_MPLS_VPN
4006 && decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
) {
4007 struct prefix pfx_mac_buf
;
4008 void *cursor
= NULL
;
4011 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
4012 pfx_mac_buf
.family
= AF_ETHERNET
;
4013 pfx_mac_buf
.prefixlen
= 48;
4014 memcpy(&pfx_mac_buf
.u
.prefix_eth
, prd
->val
+ 2, 6);
4017 * withdraw does not contain attrs, so we don't have
4018 * access to the route's LNI, which would ordinarily
4019 * select the specific mac-based import table. Instead,
4020 * we must iterate over all mac-based tables and rely
4021 * on the RD to match.
4023 * If this approach is too slow, add an index where
4024 * key is {RD, peer} and value is the import table
4026 for (rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4028 rc
== 0; rc
= skiplist_next(h
->import_mac
, NULL
,
4029 (void **)&it
, &cursor
)) {
4032 vnc_zlog_debug_verbose(
4033 "%s: calling rfapiBgpInfoFilteredImportVPN(it=%p, afi=AFI_L2VPN)",
4037 rfapiBgpInfoFilteredImportVPN(
4039 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
4040 peer
, rfd
, &pfx_mac_buf
, /* prefix */
4041 p
, /* aux_prefix: IP */
4042 AFI_L2VPN
, prd
, attr
, type
, 0,
4043 NULL
); /* sub_type & label unused for withdraw
4049 * XXX For the case where the withdraw involves an L2
4050 * route with no IP information, we rely on the lack
4051 * of RT-list intersection to filter out the withdraw
4052 * from the IP-based import tables below
4056 * Iterate over all import tables; do a filtered import
4057 * for the afi/safi combination
4060 for (it
= h
->imports
; it
; it
= it
->next
) {
4061 (*rfapiBgpInfoFilteredImportFunction(safi
))(
4062 it
, (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
4063 peer
, rfd
, p
, /* prefix */
4064 NULL
, afi
, prd
, attr
, type
, 0,
4065 NULL
); /* sub_type & label unused for withdraw */
4068 /* TBD the deletion should happen after the lifetime expires */
4069 if (safi
== SAFI_MPLS_VPN
)
4070 vnc_direct_bgp_rh_del_route(bgp
, afi
, p
, peer
);
4072 if (safi
== SAFI_MPLS_VPN
) {
4073 rfapiBgpInfoFilteredImportVPN(
4075 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
), peer
,
4076 rfd
, p
, /* prefix */
4077 NULL
, afi
, prd
, attr
, type
, 0,
4078 NULL
); /* sub_type & label unused for withdraw */
4083 * TBD optimized withdraw timer algorithm for case of many
4084 * routes expiring at the same time due to peer drop.
4087 * 1. Visit all BIs in all ENCAP import tables.
4089 * a. If a bi's peer is the failed peer, remove the bi.
4090 * b. If the removed ENCAP bi was first in the list of
4091 * BIs at this ENCAP node, loop over all monitors
4094 * (1) for each ENCAP monitor, loop over all its
4095 * VPN node monitors and set their RFAPI_MON_FLAG_NEEDCALLBACK
4098 * 2. Visit all BIs in all VPN import tables.
4099 * a. If a bi's peer is the failed peer, remove the bi.
4100 * b. loop over all the VPN node monitors and set their
4101 * RFAPI_MON_FLAG_NEEDCALLBACK flags
4102 * c. If there are no BIs left at this VPN node,
4107 /* surprise, this gets called from peer_delete(), from rfapi_close() */
4108 static void rfapiProcessPeerDownRt(struct peer
*peer
,
4109 struct rfapi_import_table
*import_table
,
4110 afi_t afi
, safi_t safi
)
4112 struct route_node
*rn
;
4113 struct bgp_info
*bi
;
4114 struct route_table
*rt
;
4115 int (*timer_service_func
)(struct thread
*);
4117 assert(afi
== AFI_IP
|| afi
== AFI_IP6
);
4123 rt
= import_table
->imported_vpn
[afi
];
4124 timer_service_func
= rfapiWithdrawTimerVPN
;
4127 rt
= import_table
->imported_encap
[afi
];
4128 timer_service_func
= rfapiWithdrawTimerEncap
;
4135 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
4136 for (bi
= rn
->info
; bi
; bi
= bi
->next
) {
4137 if (bi
->peer
== peer
) {
4139 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)) {
4140 /* already in holddown, skip */
4144 if (safi
== SAFI_MPLS_VPN
) {
4145 RFAPI_UPDATE_ITABLE_COUNT(
4146 bi
, import_table
, afi
, -1);
4147 import_table
->holddown_count
[afi
] += 1;
4149 rfapiBiStartWithdrawTimer(import_table
, rn
, bi
,
4151 timer_service_func
);
4159 * This gets called when a peer connection drops. We have to remove
4160 * all the routes from this peer.
4162 * Current approach is crude. TBD Optimize by setting fewer timers and
4163 * grouping withdrawn routes so we can generate callbacks more
4166 void rfapiProcessPeerDown(struct peer
*peer
)
4170 struct rfapi_import_table
*it
;
4173 * If this peer is a "dummy" peer structure atached to a RFAPI
4174 * nve_descriptor, we don't need to walk the import tables
4175 * because the routes are already withdrawn by rfapi_close()
4177 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_IS_RFAPI_HD
))
4181 * 1. Visit all BIs in all ENCAP import tables.
4182 * Start withdraw timer on the BIs that match peer.
4184 * 2. Visit All BIs in all VPN import tables.
4185 * Start withdraw timer on the BIs that match peer.
4188 bgp
= bgp_get_default(); /* assume 1 instance for now */
4195 for (it
= h
->imports
; it
; it
= it
->next
) {
4196 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_ENCAP
);
4197 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_ENCAP
);
4198 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_MPLS_VPN
);
4199 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_MPLS_VPN
);
4203 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4204 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4209 * Import an entire RIB (for an afi/safi) to an import table RIB,
4210 * filtered according to the import table's RT list
4212 * TBD: does this function need additions to match rfapiProcessUpdate()
4213 * for, e.g., L2 handling?
4215 static void rfapiBgpTableFilteredImport(struct bgp
*bgp
,
4216 struct rfapi_import_table
*it
,
4217 afi_t afi
, safi_t safi
)
4219 struct bgp_node
*rn1
;
4220 struct bgp_node
*rn2
;
4222 /* Only these SAFIs have 2-level RIBS */
4223 assert(safi
== SAFI_MPLS_VPN
|| safi
== SAFI_ENCAP
);
4226 * Now visit all the rd nodes and the nodes of all the
4227 * route tables attached to them, and import the routes
4228 * if they have matching route targets
4230 for (rn1
= bgp_table_top(bgp
->rib
[afi
][safi
]); rn1
;
4231 rn1
= bgp_route_next(rn1
)) {
4234 for (rn2
= bgp_table_top(rn1
->info
); rn2
;
4235 rn2
= bgp_route_next(rn2
)) {
4237 struct bgp_info
*bi
;
4239 for (bi
= rn2
->info
; bi
; bi
= bi
->next
) {
4240 u_int32_t label
= 0;
4242 if (CHECK_FLAG(bi
->flags
,
4247 label
= decode_label(
4249 (*rfapiBgpInfoFilteredImportFunction(
4251 it
, /* which import table */
4252 FIF_ACTION_UPDATE
, bi
->peer
,
4253 NULL
, &rn2
->p
, /* prefix */
4255 (struct prefix_rd
*)&rn1
->p
,
4257 bi
->sub_type
, &label
);
4265 /* per-bgp-instance rfapi data */
4266 struct rfapi
*bgp_rfapi_new(struct bgp
*bgp
)
4270 struct rfapi_rfp_cfg
*cfg
= NULL
;
4271 struct rfapi_rfp_cb_methods
*cbm
= NULL
;
4273 assert(bgp
->rfapi_cfg
== NULL
);
4275 h
= (struct rfapi
*)XCALLOC(MTYPE_RFAPI
, sizeof(struct rfapi
));
4277 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4278 h
->un
[afi
] = route_table_init();
4282 * initialize the ce import table
4284 h
->it_ce
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4285 sizeof(struct rfapi_import_table
));
4286 h
->it_ce
->imported_vpn
[AFI_IP
] = route_table_init();
4287 h
->it_ce
->imported_vpn
[AFI_IP6
] = route_table_init();
4288 h
->it_ce
->imported_encap
[AFI_IP
] = route_table_init();
4289 h
->it_ce
->imported_encap
[AFI_IP6
] = route_table_init();
4290 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4291 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4294 * Set up work queue for deferred rfapi_close operations
4296 h
->deferred_close_q
=
4297 work_queue_new(bm
->master
, "rfapi deferred close");
4298 h
->deferred_close_q
->spec
.workfunc
= rfapi_deferred_close_workfunc
;
4299 h
->deferred_close_q
->spec
.data
= h
;
4301 h
->rfp
= rfp_start(bm
->master
, &cfg
, &cbm
);
4302 bgp
->rfapi_cfg
= bgp_rfapi_cfg_new(cfg
);
4304 h
->rfp_methods
= *cbm
;
4309 void bgp_rfapi_destroy(struct bgp
*bgp
, struct rfapi
*h
)
4313 if (bgp
== NULL
|| h
== NULL
)
4316 if (h
->resolve_nve_nexthop
) {
4317 skiplist_free(h
->resolve_nve_nexthop
);
4318 h
->resolve_nve_nexthop
= NULL
;
4321 route_table_finish(h
->it_ce
->imported_vpn
[AFI_IP
]);
4322 route_table_finish(h
->it_ce
->imported_vpn
[AFI_IP6
]);
4323 route_table_finish(h
->it_ce
->imported_encap
[AFI_IP
]);
4324 route_table_finish(h
->it_ce
->imported_encap
[AFI_IP6
]);
4326 if (h
->import_mac
) {
4327 struct rfapi_import_table
*it
;
4332 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4334 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4337 rfapiImportTableFlush(it
);
4338 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
4340 skiplist_free(h
->import_mac
);
4341 h
->import_mac
= NULL
;
4344 work_queue_free(h
->deferred_close_q
);
4349 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4350 route_table_finish(h
->un
[afi
]);
4353 XFREE(MTYPE_RFAPI_IMPORTTABLE
, h
->it_ce
);
4354 XFREE(MTYPE_RFAPI
, h
);
4357 struct rfapi_import_table
*
4358 rfapiImportTableRefAdd(struct bgp
*bgp
, struct ecommunity
*rt_import_list
,
4359 struct rfapi_nve_group_cfg
*rfg
)
4362 struct rfapi_import_table
*it
;
4368 for (it
= h
->imports
; it
; it
= it
->next
) {
4369 if (ecommunity_cmp(it
->rt_import_list
, rt_import_list
))
4373 vnc_zlog_debug_verbose("%s: matched it=%p", __func__
, it
);
4376 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4377 sizeof(struct rfapi_import_table
));
4379 it
->next
= h
->imports
;
4382 it
->rt_import_list
= ecommunity_dup(rt_import_list
);
4384 it
->monitor_exterior_orphans
=
4385 skiplist_new(0, NULL
, (void (*)(void *))prefix_free
);
4388 * fill import route tables from RIBs
4390 * Potential area for optimization. If this occurs when
4391 * tables are large (e.g., the operator adds a nve group
4392 * with a new RT list to a running system), it could take
4396 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4398 it
->imported_vpn
[afi
] = route_table_init();
4399 it
->imported_encap
[afi
] = route_table_init();
4401 rfapiBgpTableFilteredImport(bgp
, it
, afi
,
4403 rfapiBgpTableFilteredImport(bgp
, it
, afi
, SAFI_ENCAP
);
4405 vnc_import_bgp_exterior_redist_enable_it(bgp
, afi
, it
);
4415 * skiplist element free function
4417 static void delete_rem_pfx_na_free(void *na
)
4419 uint32_t *pCounter
= ((struct rfapi_nve_addr
*)na
)->info
;
4422 XFREE(MTYPE_RFAPI_NVE_ADDR
, na
);
4426 * Common deleter for IP and MAC import tables
4428 static void rfapiDeleteRemotePrefixesIt(
4429 struct bgp
*bgp
, struct rfapi_import_table
*it
, struct prefix
*un
,
4430 struct prefix
*vn
, struct prefix
*p
, int delete_active
,
4431 int delete_holddown
, uint32_t *pARcount
, uint32_t *pAHcount
,
4432 uint32_t *pHRcount
, uint32_t *pHHcount
,
4433 struct skiplist
*uniq_active_nves
, struct skiplist
*uniq_holddown_nves
)
4439 char buf_pfx
[BUFSIZ
];
4442 prefix2str(p
, buf_pfx
, BUFSIZ
);
4448 vnc_zlog_debug_verbose(
4449 "%s: entry, p=%s, delete_active=%d, delete_holddown=%d",
4450 __func__
, buf_pfx
, delete_active
, delete_holddown
);
4454 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4456 struct route_table
*rt
;
4457 struct route_node
*rn
;
4459 if (p
&& (family2afi(p
->family
) != afi
)) {
4463 rt
= it
->imported_vpn
[afi
];
4467 vnc_zlog_debug_verbose("%s: scanning rt for afi=%d", __func__
,
4470 for (rn
= route_top(rt
); rn
; rn
= route_next(rn
)) {
4471 struct bgp_info
*bi
;
4472 struct bgp_info
*next
;
4474 if (VNC_DEBUG(IMPORT_DEL_REMOTE
)) {
4475 char p1line
[BUFSIZ
];
4476 char p2line
[BUFSIZ
];
4478 prefix2str(p
, p1line
, BUFSIZ
);
4479 prefix2str(&rn
->p
, p2line
, BUFSIZ
);
4480 vnc_zlog_debug_any("%s: want %s, have %s",
4481 __func__
, p1line
, p2line
);
4484 if (p
&& prefix_cmp(p
, &rn
->p
))
4488 char buf_pfx
[BUFSIZ
];
4489 prefix2str(&rn
->p
, buf_pfx
, BUFSIZ
);
4490 vnc_zlog_debug_verbose("%s: rn pfx=%s",
4494 /* TBD is this valid for afi == AFI_L2VPN? */
4495 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
4497 for (bi
= rn
->info
; bi
; bi
= next
) {
4506 vnc_zlog_debug_verbose("%s: examining bi %p",
4510 if (!rfapiGetNexthop(bi
->attr
, &qpt
))
4515 || !prefix_match(vn
, &qpt
)) {
4517 vnc_zlog_debug_verbose(
4518 "%s: continue at vn && !qpt_valid || !prefix_match(vn, &qpt)",
4525 if (!rfapiGetUnAddrOfVpnBi(bi
, &qct
))
4530 || !prefix_match(un
, &qct
)) {
4532 vnc_zlog_debug_verbose(
4533 "%s: continue at un && !qct_valid || !prefix_match(un, &qct)",
4545 * If this route is waiting to be deleted
4547 * a previous withdraw, we must cancel its
4550 if (CHECK_FLAG(bi
->flags
, BGP_INFO_REMOVED
)) {
4551 if (!delete_holddown
)
4553 if (bi
->extra
->vnc
.import
.timer
) {
4559 struct rfapi_withdraw
*wcb
=
4563 ->holddown_count
[afi
] -=
4565 RFAPI_UPDATE_ITABLE_COUNT(
4566 bi
, wcb
->import_table
,
4568 XFREE(MTYPE_RFAPI_WITHDRAW
,
4578 vnc_zlog_debug_verbose(
4579 "%s: deleting bi %p (qct_valid=%d, qpt_valid=%d, delete_holddown=%d, delete_active=%d)",
4580 __func__
, bi
, qct_valid
, qpt_valid
,
4581 delete_holddown
, delete_active
);
4587 if (qct_valid
&& qpt_valid
) {
4589 struct rfapi_nve_addr na
;
4590 struct rfapi_nve_addr
*nap
;
4592 memset(&na
, 0, sizeof(na
));
4593 assert(!rfapiQprefix2Raddr(&qct
,
4595 assert(!rfapiQprefix2Raddr(&qpt
,
4598 if (skiplist_search(
4601 : uniq_holddown_nves
),
4602 &na
, (void **)&nap
)) {
4606 MTYPE_RFAPI_NVE_ADDR
,
4611 nap
->info
= is_active
4617 : uniq_holddown_nves
),
4620 rfapiNveAddr2Str(nap
, line
,
4625 vnc_direct_bgp_rh_del_route(bgp
, afi
, &rn
->p
,
4628 RFAPI_UPDATE_ITABLE_COUNT(bi
, it
, afi
, -1);
4629 it
->holddown_count
[afi
] += 1;
4630 rfapiExpireVpnNow(it
, rn
, bi
, 1);
4632 vnc_zlog_debug_verbose(
4633 "%s: incrementing count (is_active=%d)",
4634 __func__
, is_active
);
4647 * For use by the "clear vnc prefixes" command
4649 /*------------------------------------------
4650 * rfapiDeleteRemotePrefixes
4652 * UI helper: For use by the "clear vnc prefixes" command
4655 * un if set, tunnel must match this prefix
4656 * vn if set, nexthop prefix must match this prefix
4657 * p if set, prefix must match this prefix
4658 * it if set, only look in this import table
4661 * pARcount number of active routes deleted
4662 * pAHcount number of active nves deleted
4663 * pHRcount number of holddown routes deleted
4664 * pHHcount number of holddown nves deleted
4668 --------------------------------------------*/
4669 void rfapiDeleteRemotePrefixes(struct prefix
*un
, struct prefix
*vn
,
4671 struct rfapi_import_table
*arg_it
,
4672 int delete_active
, int delete_holddown
,
4673 uint32_t *pARcount
, uint32_t *pAHcount
,
4674 uint32_t *pHRcount
, uint32_t *pHHcount
)
4678 struct rfapi_import_table
*it
;
4679 uint32_t deleted_holddown_route_count
= 0;
4680 uint32_t deleted_active_route_count
= 0;
4681 uint32_t deleted_holddown_nve_count
= 0;
4682 uint32_t deleted_active_nve_count
= 0;
4683 struct skiplist
*uniq_holddown_nves
;
4684 struct skiplist
*uniq_active_nves
;
4688 bgp
= bgp_get_default(); /* assume 1 instance for now */
4689 /* If no bgp instantiated yet, no vnc prefixes exist */
4696 uniq_holddown_nves
=
4697 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4699 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4702 * Iterate over all import tables; do a filtered import
4703 * for the afi/safi combination
4712 vnc_zlog_debug_verbose(
4713 "%s: calling rfapiDeleteRemotePrefixesIt() on (IP) import %p",
4716 rfapiDeleteRemotePrefixesIt(
4717 bgp
, it
, un
, vn
, p
, delete_active
, delete_holddown
,
4718 &deleted_active_route_count
, &deleted_active_nve_count
,
4719 &deleted_holddown_route_count
,
4720 &deleted_holddown_nve_count
, uniq_active_nves
,
4721 uniq_holddown_nves
);
4730 * Now iterate over L2 import tables
4732 if (h
->import_mac
&& !(p
&& (p
->family
!= AF_ETHERNET
))) {
4734 void *cursor
= NULL
;
4738 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4740 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4743 vnc_zlog_debug_verbose(
4744 "%s: calling rfapiDeleteRemotePrefixesIt() on import_mac %p",
4747 rfapiDeleteRemotePrefixesIt(
4748 bgp
, it
, un
, vn
, p
, delete_active
,
4749 delete_holddown
, &deleted_active_route_count
,
4750 &deleted_active_nve_count
,
4751 &deleted_holddown_route_count
,
4752 &deleted_holddown_nve_count
, uniq_active_nves
,
4753 uniq_holddown_nves
);
4758 * our custom element freeing function above counts as it deletes
4760 skiplist_free(uniq_holddown_nves
);
4761 skiplist_free(uniq_active_nves
);
4764 *pARcount
= deleted_active_route_count
;
4766 *pAHcount
= deleted_active_nve_count
;
4768 *pHRcount
= deleted_holddown_route_count
;
4770 *pHHcount
= deleted_holddown_nve_count
;
4775 /*------------------------------------------
4776 * rfapiCountRemoteRoutes
4778 * UI helper: count VRF routes from BGP side
4783 * pALRcount count of active local routes
4784 * pARRcount count of active remote routes
4785 * pHRcount count of holddown routes
4786 * pIRcount count of direct imported routes
4790 --------------------------------------------*/
4791 void rfapiCountAllItRoutes(int *pALRcount
, /* active local routes */
4792 int *pARRcount
, /* active remote routes */
4793 int *pHRcount
, /* holddown routes */
4794 int *pIRcount
) /* imported routes */
4798 struct rfapi_import_table
*it
;
4801 int total_active_local
= 0;
4802 int total_active_remote
= 0;
4803 int total_holddown
= 0;
4804 int total_imported
= 0;
4806 bgp
= bgp_get_default(); /* assume 1 instance for now */
4813 * Iterate over all import tables; do a filtered import
4814 * for the afi/safi combination
4817 for (it
= h
->imports
; it
; it
= it
->next
) {
4819 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4821 total_active_local
+= it
->local_count
[afi
];
4822 total_active_remote
+= it
->remote_count
[afi
];
4823 total_holddown
+= it
->holddown_count
[afi
];
4824 total_imported
+= it
->imported_count
[afi
];
4831 if (h
->import_mac
) {
4833 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4835 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4838 total_active_local
+= it
->local_count
[AFI_L2VPN
];
4839 total_active_remote
+= it
->remote_count
[AFI_L2VPN
];
4840 total_holddown
+= it
->holddown_count
[AFI_L2VPN
];
4841 total_imported
+= it
->imported_count
[AFI_L2VPN
];
4847 *pALRcount
= total_active_local
;
4850 *pARRcount
= total_active_remote
;
4853 *pHRcount
= total_holddown
;
4856 *pIRcount
= total_imported
;
4860 /*------------------------------------------
4861 * rfapiGetHolddownFromLifetime
4863 * calculate holddown value based on lifetime
4869 * Holddown value based on lifetime, holddown_factor,
4870 * and RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
4872 --------------------------------------------*/
4873 /* hold down time maxes out at RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY */
4874 uint32_t rfapiGetHolddownFromLifetime(uint32_t lifetime
)
4879 bgp
= bgp_get_default();
4880 if (bgp
&& bgp
->rfapi_cfg
)
4881 factor
= bgp
->rfapi_cfg
->rfp_cfg
.holddown_factor
;
4883 factor
= RFAPI_RFP_CFG_DEFAULT_HOLDDOWN_FACTOR
;
4885 if (factor
< 100 || lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4886 lifetime
= lifetime
* factor
/ 100;
4887 if (lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4890 return RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
;