1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Copyright 2009-2016, LabN Consulting, L.L.C.
8 * Purpose: Handle import of routes from BGP to RFAPI
11 #include "lib/zebra.h"
12 #include "lib/prefix.h"
13 #include "lib/agg_table.h"
15 #include "lib/memory.h"
17 #include "lib/skiplist.h"
18 #include "lib/thread.h"
19 #include "lib/stream.h"
20 #include "lib/lib_errors.h"
22 #include "bgpd/bgpd.h"
23 #include "bgpd/bgp_ecommunity.h"
24 #include "bgpd/bgp_attr.h"
25 #include "bgpd/bgp_route.h"
26 #include "bgpd/bgp_mplsvpn.h" /* prefix_rd2str() */
27 #include "bgpd/bgp_vnc_types.h"
28 #include "bgpd/bgp_rd.h"
30 #include "bgpd/rfapi/rfapi.h"
31 #include "bgpd/rfapi/bgp_rfapi_cfg.h"
32 #include "bgpd/rfapi/rfapi_backend.h"
33 #include "bgpd/rfapi/rfapi_import.h"
34 #include "bgpd/rfapi/rfapi_private.h"
35 #include "bgpd/rfapi/rfapi_monitor.h"
36 #include "bgpd/rfapi/rfapi_nve_addr.h"
37 #include "bgpd/rfapi/rfapi_vty.h"
38 #include "bgpd/rfapi/vnc_export_bgp.h"
39 #include "bgpd/rfapi/vnc_export_bgp_p.h"
40 #include "bgpd/rfapi/vnc_zebra.h"
41 #include "bgpd/rfapi/vnc_import_bgp.h"
42 #include "bgpd/rfapi/vnc_import_bgp_p.h"
43 #include "bgpd/rfapi/rfapi_rib.h"
44 #include "bgpd/rfapi/rfapi_encap_tlv.h"
45 #include "bgpd/rfapi/vnc_debug.h"
47 #ifdef HAVE_GLIBC_BACKTRACE
48 /* for backtrace and friends */
50 #endif /* HAVE_GLIBC_BACKTRACE */
52 #undef DEBUG_MONITOR_MOVE_SHORTER
53 #undef DEBUG_RETURNED_NHL
54 #undef DEBUG_ROUTE_COUNTERS
55 #undef DEBUG_ENCAP_MONITOR
58 #undef DEBUG_BI_SEARCH
61 * Allocated for each withdraw timer instance; freed when the timer
62 * expires or is canceled
64 struct rfapi_withdraw
{
65 struct rfapi_import_table
*import_table
;
66 struct agg_node
*node
;
67 struct bgp_path_info
*info
;
68 safi_t safi
; /* used only for bulk operations */
70 * For import table node reference count checking (i.e., debugging).
71 * Normally when a timer expires, lockoffset should be 0. However, if
72 * the timer expiration function is called directly (e.g.,
73 * rfapiExpireVpnNow), the node could be locked by a preceding
74 * agg_route_top() or agg_route_next() in a loop, so we need to pass
82 * It's evil and fiendish. It's compiler-dependent.
83 * ? Might need LDFLAGS -rdynamic to produce all function names
85 void rfapiDebugBacktrace(void)
87 #ifdef HAVE_GLIBC_BACKTRACE
88 #define RFAPI_DEBUG_BACKTRACE_NENTRIES 200
89 void *buf
[RFAPI_DEBUG_BACKTRACE_NENTRIES
];
94 size
= backtrace(buf
, RFAPI_DEBUG_BACKTRACE_NENTRIES
);
95 syms
= backtrace_symbols(buf
, size
);
97 for (i
= 0; i
< size
&& i
< RFAPI_DEBUG_BACKTRACE_NENTRIES
; ++i
) {
98 vnc_zlog_debug_verbose("backtrace[%2zu]: %s", i
, syms
[i
]);
108 * Count remote routes and compare with actively-maintained values.
109 * Abort if they disagree.
111 void rfapiCheckRouteCount(void)
113 struct bgp
*bgp
= bgp_get_default();
115 struct rfapi_import_table
*it
;
123 for (it
= h
->imports
; it
; it
= it
->next
) {
124 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
126 struct agg_table
*rt
;
129 int holddown_count
= 0;
131 int imported_count
= 0;
132 int remote_count
= 0;
134 rt
= it
->imported_vpn
[afi
];
136 for (rn
= agg_route_top(rt
); rn
;
137 rn
= agg_route_next(rn
)) {
138 struct bgp_path_info
*bpi
;
139 struct bgp_path_info
*next
;
141 for (bpi
= rn
->info
; bpi
; bpi
= next
) {
144 if (CHECK_FLAG(bpi
->flags
,
149 if (RFAPI_LOCAL_BI(bpi
)) {
152 if (RFAPI_DIRECT_IMPORT_BI(
163 if (it
->holddown_count
[afi
] != holddown_count
) {
164 vnc_zlog_debug_verbose(
165 "%s: it->holddown_count %d != holddown_count %d",
166 __func__
, it
->holddown_count
[afi
],
170 if (it
->remote_count
[afi
] != remote_count
) {
171 vnc_zlog_debug_verbose(
172 "%s: it->remote_count %d != remote_count %d",
173 __func__
, it
->remote_count
[afi
],
177 if (it
->imported_count
[afi
] != imported_count
) {
178 vnc_zlog_debug_verbose(
179 "%s: it->imported_count %d != imported_count %d",
180 __func__
, it
->imported_count
[afi
],
188 #ifdef DEBUG_ROUTE_COUNTERS
189 #define VNC_ITRCCK do {rfapiCheckRouteCount();} while (0)
195 * Validate reference count for a node in an import table
197 * Normally lockoffset is 0 for nodes in quiescent state. However,
198 * agg_unlock_node will delete the node if it is called when
199 * node->lock == 1, and we have to validate the refcount before
200 * the node is deleted. In this case, we specify lockoffset 1.
202 void rfapiCheckRefcount(struct agg_node
*rn
, safi_t safi
, int lockoffset
)
204 unsigned int count_bpi
= 0;
205 unsigned int count_monitor
= 0;
206 struct bgp_path_info
*bpi
;
207 struct rfapi_monitor_encap
*hme
;
208 struct rfapi_monitor_vpn
*hmv
;
210 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
)
215 ++count_monitor
; /* rfapi_it_extra */
222 for (hme
= RFAPI_MONITOR_ENCAP(rn
); hme
;
229 for (hmv
= RFAPI_MONITOR_VPN(rn
); hmv
; hmv
= hmv
->next
)
232 if (RFAPI_MONITOR_EXTERIOR(rn
)->source
) {
233 ++count_monitor
; /* sl */
235 for (rc
= skiplist_next(
236 RFAPI_MONITOR_EXTERIOR(rn
)->source
,
237 NULL
, NULL
, &cursor
);
240 RFAPI_MONITOR_EXTERIOR(rn
)->source
,
241 NULL
, NULL
, &cursor
)) {
243 ++count_monitor
; /* sl entry */
252 case SAFI_LABELED_UNICAST
:
255 assert(!"Passed in safi should be impossible");
259 if (count_bpi
+ count_monitor
+ lockoffset
260 != agg_node_get_lock_count(rn
)) {
261 vnc_zlog_debug_verbose(
262 "%s: count_bpi=%d, count_monitor=%d, lockoffset=%d, rn->lock=%d",
263 __func__
, count_bpi
, count_monitor
, lockoffset
,
264 agg_node_get_lock_count(rn
));
270 * Perform deferred rfapi_close operations that were queued
273 static wq_item_status
rfapi_deferred_close_workfunc(struct work_queue
*q
,
276 struct rfapi_descriptor
*rfd
= data
;
277 struct rfapi
*h
= q
->spec
.data
;
279 assert(!(h
->flags
& RFAPI_INCALLBACK
));
281 vnc_zlog_debug_verbose("%s: completed deferred close on handle %p",
287 * Extract layer 2 option from Encap TLVS in BGP attrs
289 int rfapiGetL2o(struct attr
*attr
, struct rfapi_l2address_option
*l2o
)
292 struct bgp_attr_encap_subtlv
*pEncap
;
294 for (pEncap
= bgp_attr_get_vnc_subtlvs(attr
); pEncap
;
295 pEncap
= pEncap
->next
) {
297 if (pEncap
->type
== BGP_VNC_SUBTLV_TYPE_RFPOPTION
) {
299 == RFAPI_VN_OPTION_TYPE_L2ADDR
) {
301 if (pEncap
->value
[1] == 14) {
302 memcpy(l2o
->macaddr
.octet
,
319 l2o
->logical_net_id
=
322 + ((pEncap
->value
[14]
325 + ((pEncap
->value
[13]
340 * Extract the lifetime from the Tunnel Encap attribute of a route in
343 int rfapiGetVncLifetime(struct attr
*attr
, uint32_t *lifetime
)
345 struct bgp_attr_encap_subtlv
*pEncap
;
347 *lifetime
= RFAPI_INFINITE_LIFETIME
; /* default to infinite */
351 for (pEncap
= bgp_attr_get_vnc_subtlvs(attr
); pEncap
;
352 pEncap
= pEncap
->next
) {
355 == BGP_VNC_SUBTLV_TYPE_LIFETIME
) { /* lifetime */
356 if (pEncap
->length
== 4) {
357 memcpy(lifetime
, pEncap
->value
, 4);
358 *lifetime
= ntohl(*lifetime
);
369 * Look for UN address in Encap attribute
371 int rfapiGetVncTunnelUnAddr(struct attr
*attr
, struct prefix
*p
)
373 struct bgp_attr_encap_subtlv
*pEncap
;
374 bgp_encap_types tun_type
= BGP_ENCAP_TYPE_MPLS
;/*Default tunnel type*/
376 bgp_attr_extcom_tunnel_type(attr
, &tun_type
);
377 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
380 /* MPLS carries UN address in next hop */
381 rfapiNexthop2Prefix(attr
, p
);
382 if (p
->family
!= AF_UNSPEC
)
388 for (pEncap
= attr
->encap_subtlvs
; pEncap
;
389 pEncap
= pEncap
->next
) {
392 == BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
) { /* un
395 switch (pEncap
->length
) {
399 p
->prefixlen
= IPV4_MAX_BITLEN
;
400 memcpy(p
->u
.val
, pEncap
->value
,
407 p
->family
= AF_INET6
;
408 p
->prefixlen
= IPV6_MAX_BITLEN
;
409 memcpy(p
->u
.val
, pEncap
->value
,
422 * Get UN address wherever it might be
424 int rfapiGetUnAddrOfVpnBi(struct bgp_path_info
*bpi
, struct prefix
*p
)
426 /* If it's in this route's VNC attribute, we're done */
427 if (!rfapiGetVncTunnelUnAddr(bpi
->attr
, p
))
430 * Otherwise, see if it's cached from a corresponding ENCAP SAFI
434 switch (bpi
->extra
->vnc
.import
.un_family
) {
437 p
->family
= bpi
->extra
->vnc
.import
.un_family
;
438 p
->u
.prefix4
= bpi
->extra
->vnc
.import
.un
.addr4
;
439 p
->prefixlen
= IPV4_MAX_BITLEN
;
444 p
->family
= bpi
->extra
->vnc
.import
.un_family
;
445 p
->u
.prefix6
= bpi
->extra
->vnc
.import
.un
.addr6
;
446 p
->prefixlen
= IPV6_MAX_BITLEN
;
451 p
->family
= AF_UNSPEC
;
452 #ifdef DEBUG_ENCAP_MONITOR
453 vnc_zlog_debug_verbose(
454 "%s: bpi->extra->vnc.import.un_family is 0, no UN addr",
466 * Make a new bgp_path_info from gathered parameters
468 static struct bgp_path_info
*rfapiBgpInfoCreate(struct attr
*attr
,
469 struct peer
*peer
, void *rfd
,
470 struct prefix_rd
*prd
,
471 uint8_t type
, uint8_t sub_type
,
474 struct bgp_path_info
*new;
476 new = info_make(type
, sub_type
, 0, peer
, attr
, NULL
);
478 new->attr
= bgp_attr_intern(attr
);
480 bgp_path_info_extra_get(new);
482 new->extra
->vnc
.import
.rd
= *prd
;
483 new->extra
->vnc
.import
.create_time
= monotime(NULL
);
486 encode_label(*label
, &new->extra
->label
[0]);
494 * Frees bgp_path_info as used in import tables (parts are not
495 * allocated exactly the way they are in the main RIBs)
497 static void rfapiBgpInfoFree(struct bgp_path_info
*goner
)
503 vnc_zlog_debug_verbose("%s: calling peer_unlock(%p), #%d",
504 __func__
, goner
->peer
,
506 peer_unlock(goner
->peer
);
509 bgp_attr_unintern(&goner
->attr
);
512 bgp_path_info_extra_free(&goner
->extra
);
513 XFREE(MTYPE_BGP_ROUTE
, goner
);
516 struct rfapi_import_table
*rfapiMacImportTableGetNoAlloc(struct bgp
*bgp
,
520 struct rfapi_import_table
*it
= NULL
;
521 uintptr_t lni_as_ptr
= lni
;
530 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
))
536 struct rfapi_import_table
*rfapiMacImportTableGet(struct bgp
*bgp
, uint32_t lni
)
539 struct rfapi_import_table
*it
= NULL
;
540 uintptr_t lni_as_ptr
= lni
;
545 if (!h
->import_mac
) {
546 /* default cmp is good enough for LNI */
547 h
->import_mac
= skiplist_new(0, NULL
, NULL
);
550 if (skiplist_search(h
->import_mac
, (void *)lni_as_ptr
, (void **)&it
)) {
552 struct ecommunity
*enew
;
553 struct ecommunity_val eval
;
556 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
557 sizeof(struct rfapi_import_table
));
558 /* set RT list of new import table based on LNI */
559 memset((char *)&eval
, 0, sizeof(eval
));
560 eval
.val
[0] = 0; /* VNC L2VPN */
561 eval
.val
[1] = 2; /* VNC L2VPN */
562 eval
.val
[5] = (lni
>> 16) & 0xff;
563 eval
.val
[6] = (lni
>> 8) & 0xff;
564 eval
.val
[7] = (lni
>> 0) & 0xff;
566 enew
= ecommunity_new();
567 ecommunity_add_val(enew
, &eval
, false, false);
568 it
->rt_import_list
= enew
;
570 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
571 it
->imported_vpn
[afi
] = agg_table_init();
572 it
->imported_encap
[afi
] = agg_table_init();
575 it
->l2_logical_net_id
= lni
;
577 skiplist_insert(h
->import_mac
, (void *)lni_as_ptr
, it
);
585 * Implement MONITOR_MOVE_SHORTER(original_node) from
586 * RFAPI-Import-Event-Handling.txt
588 * Returns pointer to the list of moved monitors
590 static struct rfapi_monitor_vpn
*
591 rfapiMonitorMoveShorter(struct agg_node
*original_vpn_node
, int lockoffset
)
593 struct bgp_path_info
*bpi
;
594 struct agg_node
*par
;
595 struct rfapi_monitor_vpn
*m
;
596 struct rfapi_monitor_vpn
*mlast
;
597 struct rfapi_monitor_vpn
*moved
;
599 int parent_already_refcounted
= 0;
601 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
, lockoffset
);
603 #ifdef DEBUG_MONITOR_MOVE_SHORTER
605 vnc_zlog_debug_verbose("%s: called with node pfx=%pFX",
606 __func__
, &original_vpn_node
->p
);
611 * 1. If there is at least one bpi (either regular route or
612 * route marked as withdrawn, with a pending timer) at
613 * original_node with a valid UN address, we're done. Return.
615 for (bpi
= original_vpn_node
->info
; bpi
; bpi
= bpi
->next
) {
618 if (!rfapiGetUnAddrOfVpnBi(bpi
, &pfx
)) {
619 #ifdef DEBUG_MONITOR_MOVE_SHORTER
620 vnc_zlog_debug_verbose(
621 "%s: have valid UN at original node, no change",
629 * 2. Travel up the tree (toward less-specific prefixes) from
630 * original_node to find the first node that has at least
631 * one route (even if it is only a withdrawn route) with a
632 * valid UN address. Call this node "Node P."
634 for (par
= agg_node_parent(original_vpn_node
); par
;
635 par
= agg_node_parent(par
)) {
636 for (bpi
= par
->info
; bpi
; bpi
= bpi
->next
) {
638 if (!rfapiGetUnAddrOfVpnBi(bpi
, &pfx
)) {
647 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 0);
651 * If no less-specific routes, try to use the 0/0 node
654 const struct prefix
*p
;
655 /* this isn't necessarily 0/0 */
656 par
= agg_route_table_top(original_vpn_node
);
659 p
= agg_node_get_prefix(par
);
661 * If we got the top node but it wasn't 0/0,
664 if (par
&& p
->prefixlen
) {
665 agg_unlock_node(par
); /* maybe free */
670 ++parent_already_refcounted
;
675 * Create 0/0 node if it isn't there
678 struct prefix pfx_default
;
679 const struct prefix
*p
= agg_node_get_prefix(original_vpn_node
);
681 memset(&pfx_default
, 0, sizeof(pfx_default
));
682 pfx_default
.family
= p
->family
;
684 /* creates default node if none exists */
685 par
= agg_node_get(agg_get_table(original_vpn_node
),
687 ++parent_already_refcounted
;
691 * 3. Move each of the monitors found at original_node to Node P.
692 * These are "Moved Monitors."
697 * Attach at end so that the list pointer we return points
698 * only to the moved routes
700 for (m
= RFAPI_MONITOR_VPN(par
), mlast
= NULL
; m
;
701 mlast
= m
, m
= m
->next
)
705 moved
= mlast
->next
= RFAPI_MONITOR_VPN(original_vpn_node
);
707 moved
= RFAPI_MONITOR_VPN_W_ALLOC(par
) =
708 RFAPI_MONITOR_VPN(original_vpn_node
);
710 if (RFAPI_MONITOR_VPN(
711 original_vpn_node
)) /* check agg, so not allocated */
712 RFAPI_MONITOR_VPN_W_ALLOC(original_vpn_node
) = NULL
;
715 * update the node pointers on the monitors
717 for (m
= moved
; m
; m
= m
->next
) {
722 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
,
723 parent_already_refcounted
- movecount
);
724 while (movecount
> parent_already_refcounted
) {
726 ++parent_already_refcounted
;
728 while (movecount
< parent_already_refcounted
) {
729 /* unlikely, but code defensively */
730 agg_unlock_node(par
);
731 --parent_already_refcounted
;
733 RFAPI_CHECK_REFCOUNT(original_vpn_node
, SAFI_MPLS_VPN
,
734 movecount
+ lockoffset
);
735 while (movecount
--) {
736 agg_unlock_node(original_vpn_node
);
739 #ifdef DEBUG_MONITOR_MOVE_SHORTER
741 vnc_zlog_debug_verbose("%s: moved to node pfx=%pFX", __func__
,
751 * Implement MONITOR_MOVE_LONGER(new_node) from
752 * RFAPI-Import-Event-Handling.txt
754 static void rfapiMonitorMoveLonger(struct agg_node
*new_vpn_node
)
756 struct rfapi_monitor_vpn
*monitor
;
757 struct rfapi_monitor_vpn
*mlast
;
758 struct bgp_path_info
*bpi
;
759 struct agg_node
*par
;
760 const struct prefix
*new_vpn_node_p
= agg_node_get_prefix(new_vpn_node
);
762 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
765 * Make sure we have at least one valid route at the new node
767 for (bpi
= new_vpn_node
->info
; bpi
; bpi
= bpi
->next
) {
769 if (!rfapiGetUnAddrOfVpnBi(bpi
, &pfx
))
774 vnc_zlog_debug_verbose(
775 "%s: no valid routes at node %p, so not attempting moves",
776 __func__
, new_vpn_node
);
781 * Find first parent node that has monitors
783 for (par
= agg_node_parent(new_vpn_node
); par
;
784 par
= agg_node_parent(par
)) {
785 if (RFAPI_MONITOR_VPN(par
))
790 vnc_zlog_debug_verbose(
791 "%s: no parent nodes with monitors, done", __func__
);
796 * Check each of these monitors to see of their longest-match
797 * is now the updated node. Move any such monitors to the more-
798 * specific updated node
800 for (mlast
= NULL
, monitor
= RFAPI_MONITOR_VPN(par
); monitor
;) {
802 * If new longest match for monitor prefix is the new
803 * route's prefix, move monitor to new route's prefix
805 if (prefix_match(new_vpn_node_p
, &monitor
->p
)) {
808 mlast
->next
= monitor
->next
;
810 RFAPI_MONITOR_VPN_W_ALLOC(par
) = monitor
->next
;
815 monitor
->next
= RFAPI_MONITOR_VPN(new_vpn_node
);
816 RFAPI_MONITOR_VPN_W_ALLOC(new_vpn_node
) = monitor
;
817 monitor
->node
= new_vpn_node
;
819 agg_lock_node(new_vpn_node
); /* incr refcount */
821 monitor
= mlast
? mlast
->next
: RFAPI_MONITOR_VPN(par
);
823 RFAPI_CHECK_REFCOUNT(par
, SAFI_MPLS_VPN
, 1);
824 /* decr refcount after we're done with par as this might
826 agg_unlock_node(par
);
831 monitor
= monitor
->next
;
834 RFAPI_CHECK_REFCOUNT(new_vpn_node
, SAFI_MPLS_VPN
, 0);
838 static void rfapiBgpInfoChainFree(struct bgp_path_info
*bpi
)
840 struct bgp_path_info
*next
;
845 * If there is a timer waiting to delete this bpi, cancel
846 * the timer and delete immediately
848 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
849 && bpi
->extra
->vnc
.import
.timer
) {
850 struct rfapi_withdraw
*wcb
=
851 THREAD_ARG(bpi
->extra
->vnc
.import
.timer
);
853 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
854 THREAD_OFF(bpi
->extra
->vnc
.import
.timer
);
859 rfapiBgpInfoFree(bpi
);
864 static void rfapiImportTableFlush(struct rfapi_import_table
*it
)
871 ecommunity_free(&it
->rt_import_list
);
872 it
->rt_import_list
= NULL
;
874 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
878 for (rn
= agg_route_top(it
->imported_vpn
[afi
]); rn
;
879 rn
= agg_route_next(rn
)) {
881 * Each route_node has:
882 * aggregate: points to rfapi_it_extra with monitor
884 * info: points to chain of bgp_path_info
886 /* free bgp_path_info and its children */
887 rfapiBgpInfoChainFree(rn
->info
);
890 rfapiMonitorExtraFlush(SAFI_MPLS_VPN
, rn
);
893 for (rn
= agg_route_top(it
->imported_encap
[afi
]); rn
;
894 rn
= agg_route_next(rn
)) {
895 /* free bgp_path_info and its children */
896 rfapiBgpInfoChainFree(rn
->info
);
899 rfapiMonitorExtraFlush(SAFI_ENCAP
, rn
);
902 agg_table_finish(it
->imported_vpn
[afi
]);
903 agg_table_finish(it
->imported_encap
[afi
]);
905 if (it
->monitor_exterior_orphans
) {
906 skiplist_free(it
->monitor_exterior_orphans
);
910 void rfapiImportTableRefDelByIt(struct bgp
*bgp
,
911 struct rfapi_import_table
*it_target
)
914 struct rfapi_import_table
*it
;
915 struct rfapi_import_table
*prev
= NULL
;
922 for (it
= h
->imports
; it
; prev
= it
, it
= it
->next
) {
928 assert(it
->refcount
);
934 prev
->next
= it
->next
;
936 h
->imports
= it
->next
;
938 rfapiImportTableFlush(it
);
939 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
943 #ifdef RFAPI_REQUIRE_ENCAP_BEEC
945 * Look for magic BGP Encapsulation Extended Community value
946 * Format in RFC 5512 Sect. 4.5
948 static int rfapiEcommunitiesMatchBeec(struct ecommunity
*ecom
,
949 bgp_encap_types type
)
956 for (i
= 0; i
< (ecom
->size
* ECOMMUNITY_SIZE
); i
+= ECOMMUNITY_SIZE
) {
962 if (ep
[0] == ECOMMUNITY_ENCODE_OPAQUE
963 && ep
[1] == ECOMMUNITY_OPAQUE_SUBTYPE_ENCAP
964 && ep
[6] == ((type
&& 0xff00) >> 8)
965 && ep
[7] == (type
& 0xff)) {
974 int rfapiEcommunitiesIntersect(struct ecommunity
*e1
, struct ecommunity
*e2
)
983 s1
= ecommunity_ecom2str(e1
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
984 s2
= ecommunity_ecom2str(e2
, ECOMMUNITY_FORMAT_DISPLAY
, 0);
985 vnc_zlog_debug_verbose("%s: e1[%s], e2[%s]", __func__
, s1
, s2
);
986 XFREE(MTYPE_ECOMMUNITY_STR
, s1
);
987 XFREE(MTYPE_ECOMMUNITY_STR
, s2
);
990 for (i
= 0; i
< e1
->size
; ++i
) {
991 for (j
= 0; j
< e2
->size
; ++j
) {
992 if (!memcmp(e1
->val
+ (i
* ECOMMUNITY_SIZE
),
993 e2
->val
+ (j
* ECOMMUNITY_SIZE
),
1003 int rfapiEcommunityGetLNI(struct ecommunity
*ecom
, uint32_t *lni
)
1008 for (i
= 0; i
< ecom
->size
; ++i
) {
1009 uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1011 if ((*(p
+ 0) == 0x00) && (*(p
+ 1) == 0x02)) {
1013 *lni
= (*(p
+ 5) << 16) | (*(p
+ 6) << 8)
1022 int rfapiEcommunityGetEthernetTag(struct ecommunity
*ecom
, uint16_t *tag_id
)
1024 struct bgp
*bgp
= bgp_get_default();
1025 *tag_id
= 0; /* default to untagged */
1029 for (i
= 0; i
< ecom
->size
; ++i
) {
1032 const uint8_t *p
= ecom
->val
+ (i
* ECOMMUNITY_SIZE
);
1034 /* High-order octet of type. */
1037 if (*p
++ == ECOMMUNITY_ROUTE_TARGET
) {
1038 if (encode
== ECOMMUNITY_ENCODE_AS4
) {
1039 p
= ptr_get_be32(p
, &as
);
1040 } else if (encode
== ECOMMUNITY_ENCODE_AS
) {
1043 p
+= 2; /* skip next two, tag/vid
1044 always in lowest bytes */
1046 if (as
== bgp
->as
) {
1047 *tag_id
= *p
++ << 8;
1057 static int rfapiVpnBiNhEqualsPt(struct bgp_path_info
*bpi
,
1058 struct rfapi_ip_addr
*hpt
)
1065 family
= BGP_MP_NEXTHOP_FAMILY(bpi
->attr
->mp_nexthop_len
);
1067 if (hpt
->addr_family
!= family
)
1072 if (bpi
->attr
->mp_nexthop_global_in
.s_addr
1073 != hpt
->addr
.v4
.s_addr
)
1078 if (IPV6_ADDR_CMP(&bpi
->attr
->mp_nexthop_global
, &hpt
->addr
.v6
))
1091 * Compare 2 VPN BIs. Return true if they have the same VN and UN addresses
1093 static int rfapiVpnBiSamePtUn(struct bgp_path_info
*bpi1
,
1094 struct bgp_path_info
*bpi2
)
1096 struct prefix pfx_un1
;
1097 struct prefix pfx_un2
;
1103 * VN address comparisons
1106 if (BGP_MP_NEXTHOP_FAMILY(bpi1
->attr
->mp_nexthop_len
)
1107 != BGP_MP_NEXTHOP_FAMILY(bpi2
->attr
->mp_nexthop_len
)) {
1111 switch (BGP_MP_NEXTHOP_FAMILY(bpi1
->attr
->mp_nexthop_len
)) {
1113 if (bpi1
->attr
->mp_nexthop_global_in
.s_addr
1114 != bpi2
->attr
->mp_nexthop_global_in
.s_addr
)
1119 if (IPV6_ADDR_CMP(&bpi1
->attr
->mp_nexthop_global
,
1120 &bpi2
->attr
->mp_nexthop_global
))
1128 memset(&pfx_un1
, 0, sizeof(pfx_un1
));
1129 memset(&pfx_un2
, 0, sizeof(pfx_un2
));
1132 * UN address comparisons
1134 if (rfapiGetVncTunnelUnAddr(bpi1
->attr
, &pfx_un1
)) {
1136 pfx_un1
.family
= bpi1
->extra
->vnc
.import
.un_family
;
1137 switch (bpi1
->extra
->vnc
.import
.un_family
) {
1140 bpi1
->extra
->vnc
.import
.un
.addr4
;
1144 bpi1
->extra
->vnc
.import
.un
.addr6
;
1147 pfx_un1
.family
= AF_UNSPEC
;
1153 if (rfapiGetVncTunnelUnAddr(bpi2
->attr
, &pfx_un2
)) {
1155 pfx_un2
.family
= bpi2
->extra
->vnc
.import
.un_family
;
1156 switch (bpi2
->extra
->vnc
.import
.un_family
) {
1159 bpi2
->extra
->vnc
.import
.un
.addr4
;
1163 bpi2
->extra
->vnc
.import
.un
.addr6
;
1166 pfx_un2
.family
= AF_UNSPEC
;
1172 if (pfx_un1
.family
== AF_UNSPEC
|| pfx_un2
.family
== AF_UNSPEC
)
1175 if (pfx_un1
.family
!= pfx_un2
.family
)
1178 switch (pfx_un1
.family
) {
1180 if (!IPV4_ADDR_SAME(&pfx_un1
.u
.prefix4
, &pfx_un2
.u
.prefix4
))
1184 if (!IPV6_ADDR_SAME(&pfx_un1
.u
.prefix6
, &pfx_un2
.u
.prefix6
))
1193 uint8_t rfapiRfpCost(struct attr
*attr
)
1195 if (attr
->flag
& ATTR_FLAG_BIT(BGP_ATTR_LOCAL_PREF
)) {
1196 if (attr
->local_pref
> 255) {
1199 return 255 - attr
->local_pref
;
1205 /*------------------------------------------
1208 * Find Layer 2 options in an option chain
1214 * l2o layer 2 options extracted
1218 * 1 no options found
1220 --------------------------------------------*/
1221 int rfapi_extract_l2o(
1222 struct bgp_tea_options
*pHop
, /* chain of options */
1223 struct rfapi_l2address_option
*l2o
) /* return extracted value */
1225 struct bgp_tea_options
*p
;
1227 for (p
= pHop
; p
; p
= p
->next
) {
1228 if ((p
->type
== RFAPI_VN_OPTION_TYPE_L2ADDR
)
1229 && (p
->length
>= 8)) {
1233 memcpy(&l2o
->macaddr
, v
, 6);
1235 l2o
->label
= ((v
[6] << 12) & 0xff000)
1236 + ((v
[7] << 4) & 0xff0)
1237 + ((v
[8] >> 4) & 0xf);
1239 l2o
->local_nve_id
= (uint8_t)v
[10];
1241 l2o
->logical_net_id
=
1242 (v
[11] << 16) + (v
[12] << 8) + (v
[13] << 0);
1250 static struct rfapi_next_hop_entry
*
1251 rfapiRouteInfo2NextHopEntry(struct rfapi_ip_prefix
*rprefix
,
1252 struct bgp_path_info
*bpi
, /* route to encode */
1253 uint32_t lifetime
, /* use this in nhe */
1254 struct agg_node
*rn
) /* req for L2 eth addr */
1256 struct rfapi_next_hop_entry
*new;
1257 int have_vnc_tunnel_un
= 0;
1258 const struct prefix
*p
= agg_node_get_prefix(rn
);
1260 #ifdef DEBUG_ENCAP_MONITOR
1261 vnc_zlog_debug_verbose("%s: entry, bpi %p, rn %p", __func__
, bpi
, rn
);
1264 new = XCALLOC(MTYPE_RFAPI_NEXTHOP
, sizeof(struct rfapi_next_hop_entry
));
1266 new->prefix
= *rprefix
;
1269 && decode_rd_type(bpi
->extra
->vnc
.import
.rd
.val
)
1270 == RD_TYPE_VNC_ETH
) {
1273 struct rfapi_vn_option
*vo
;
1275 vo
= XCALLOC(MTYPE_RFAPI_VN_OPTION
,
1276 sizeof(struct rfapi_vn_option
));
1278 vo
->type
= RFAPI_VN_OPTION_TYPE_L2ADDR
;
1280 memcpy(&vo
->v
.l2addr
.macaddr
, &p
->u
.prefix_eth
.octet
, ETH_ALEN
);
1281 /* only low 3 bytes of this are significant */
1282 (void)rfapiEcommunityGetLNI(bgp_attr_get_ecommunity(bpi
->attr
),
1283 &vo
->v
.l2addr
.logical_net_id
);
1284 (void)rfapiEcommunityGetEthernetTag(
1285 bgp_attr_get_ecommunity(bpi
->attr
),
1286 &vo
->v
.l2addr
.tag_id
);
1288 /* local_nve_id comes from lower byte of RD type */
1289 vo
->v
.l2addr
.local_nve_id
= bpi
->extra
->vnc
.import
.rd
.val
[1];
1291 /* label comes from MP_REACH_NLRI label */
1292 vo
->v
.l2addr
.label
= decode_label(&bpi
->extra
->label
[0]);
1294 new->vn_options
= vo
;
1297 * If there is an auxiliary prefix (i.e., host IP address),
1298 * use it as the nexthop prefix instead of the query prefix
1300 if (bpi
->extra
->vnc
.import
.aux_prefix
.family
) {
1301 rfapiQprefix2Rprefix(&bpi
->extra
->vnc
.import
.aux_prefix
,
1306 bgp_encap_types tun_type
= BGP_ENCAP_TYPE_MPLS
; /*Default*/
1307 new->prefix
.cost
= rfapiRfpCost(bpi
->attr
);
1309 struct bgp_attr_encap_subtlv
*pEncap
;
1311 switch (BGP_MP_NEXTHOP_FAMILY(bpi
->attr
->mp_nexthop_len
)) {
1313 new->vn_address
.addr_family
= AF_INET
;
1314 new->vn_address
.addr
.v4
= bpi
->attr
->mp_nexthop_global_in
;
1318 new->vn_address
.addr_family
= AF_INET6
;
1319 new->vn_address
.addr
.v6
= bpi
->attr
->mp_nexthop_global
;
1323 zlog_warn("%s: invalid vpn nexthop length: %d", __func__
,
1324 bpi
->attr
->mp_nexthop_len
);
1325 rfapi_free_next_hop_list(new);
1329 for (pEncap
= bgp_attr_get_vnc_subtlvs(bpi
->attr
); pEncap
;
1330 pEncap
= pEncap
->next
) {
1331 switch (pEncap
->type
) {
1332 case BGP_VNC_SUBTLV_TYPE_LIFETIME
:
1333 /* use configured lifetime, not attr lifetime */
1337 zlog_warn("%s: unknown VNC option type %d", __func__
,
1344 bgp_attr_extcom_tunnel_type(bpi
->attr
, &tun_type
);
1345 if (tun_type
== BGP_ENCAP_TYPE_MPLS
) {
1347 /* MPLS carries UN address in next hop */
1348 rfapiNexthop2Prefix(bpi
->attr
, &p
);
1349 if (p
.family
!= AF_UNSPEC
) {
1350 rfapiQprefix2Raddr(&p
, &new->un_address
);
1351 have_vnc_tunnel_un
= 1;
1355 for (pEncap
= bpi
->attr
->encap_subtlvs
; pEncap
; pEncap
= pEncap
->next
) {
1356 switch (pEncap
->type
) {
1357 case BGP_ENCAP_SUBTLV_TYPE_REMOTE_ENDPOINT
:
1359 * Overrides ENCAP UN address, if any
1361 switch (pEncap
->length
) {
1364 new->un_address
.addr_family
= AF_INET
;
1365 memcpy(&new->un_address
.addr
.v4
, pEncap
->value
,
1367 have_vnc_tunnel_un
= 1;
1371 new->un_address
.addr_family
= AF_INET6
;
1372 memcpy(&new->un_address
.addr
.v6
, pEncap
->value
,
1374 have_vnc_tunnel_un
= 1;
1379 "%s: invalid tunnel subtlv UN addr length (%d) for bpi %p",
1380 __func__
, pEncap
->length
, bpi
);
1385 zlog_warn("%s: unknown Encap Attribute option type %d",
1386 __func__
, pEncap
->type
);
1391 new->un_options
= rfapi_encap_tlv_to_un_option(bpi
->attr
);
1393 #ifdef DEBUG_ENCAP_MONITOR
1394 vnc_zlog_debug_verbose("%s: line %d: have_vnc_tunnel_un=%d", __func__
,
1395 __LINE__
, have_vnc_tunnel_un
);
1398 if (!have_vnc_tunnel_un
&& bpi
->extra
) {
1400 * use cached UN address from ENCAP route
1402 new->un_address
.addr_family
= bpi
->extra
->vnc
.import
.un_family
;
1403 switch (new->un_address
.addr_family
) {
1405 new->un_address
.addr
.v4
=
1406 bpi
->extra
->vnc
.import
.un
.addr4
;
1409 new->un_address
.addr
.v6
=
1410 bpi
->extra
->vnc
.import
.un
.addr6
;
1413 zlog_warn("%s: invalid UN addr family (%d) for bpi %p",
1414 __func__
, new->un_address
.addr_family
, bpi
);
1415 rfapi_free_next_hop_list(new);
1420 new->lifetime
= lifetime
;
1424 int rfapiHasNonRemovedRoutes(struct agg_node
*rn
)
1426 struct bgp_path_info
*bpi
;
1428 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
1431 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
1432 && (bpi
->extra
&& !rfapiGetUnAddrOfVpnBi(bpi
, &pfx
))) {
1440 #ifdef DEBUG_IT_NODES
1444 void rfapiDumpNode(struct agg_node
*rn
)
1446 struct bgp_path_info
*bpi
;
1448 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
1449 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
1451 int ctrc
= rfapiGetUnAddrOfVpnBi(bpi
, &pfx
);
1454 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
1455 && (bpi
->extra
&& !ctrc
)) {
1462 vnc_zlog_debug_verbose(
1463 " bpi=%p, nr=%d, flags=0x%x, extra=%p, ctrc=%d", bpi
,
1464 nr
, bpi
->flags
, bpi
->extra
, ctrc
);
1469 static int rfapiNhlAddNodeRoutes(
1470 struct agg_node
*rn
, /* in */
1471 struct rfapi_ip_prefix
*rprefix
, /* in */
1472 uint32_t lifetime
, /* in */
1473 int removed
, /* in */
1474 struct rfapi_next_hop_entry
**head
, /* in/out */
1475 struct rfapi_next_hop_entry
**tail
, /* in/out */
1476 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1477 struct agg_node
*rfd_rib_node
, /* preload this NVE rib node */
1478 struct prefix
*pfx_target_original
) /* query target */
1480 struct bgp_path_info
*bpi
;
1481 struct rfapi_next_hop_entry
*new;
1482 struct prefix pfx_un
;
1483 struct skiplist
*seen_nexthops
;
1485 const struct prefix
*p
= agg_node_get_prefix(rn
);
1486 int is_l2
= (p
->family
== AF_ETHERNET
);
1489 struct agg_table
*atable
= agg_get_table(rfd_rib_node
);
1490 struct rfapi_descriptor
*rfd
;
1493 rfd
= agg_get_table_info(atable
);
1495 if (rfapiRibFTDFilterRecentPrefix(rfd
, rn
,
1496 pfx_target_original
))
1502 skiplist_new(0, vnc_prefix_cmp
, prefix_free_lists
);
1504 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
1506 struct prefix pfx_vn
;
1507 struct prefix
*newpfx
;
1509 if (removed
&& !CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
1510 #ifdef DEBUG_RETURNED_NHL
1511 vnc_zlog_debug_verbose(
1512 "%s: want holddown, this route not holddown, skip",
1517 if (!removed
&& CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
1526 * Check for excluded VN address
1528 if (rfapiVpnBiNhEqualsPt(bpi
, exclude_vnaddr
))
1532 * Check for VN address (nexthop) copied already
1535 /* L2 routes: semantic nexthop in aux_prefix; VN addr
1537 pfx_vn
= bpi
->extra
->vnc
.import
.aux_prefix
;
1539 rfapiNexthop2Prefix(bpi
->attr
, &pfx_vn
);
1541 if (!skiplist_search(seen_nexthops
, &pfx_vn
, NULL
)) {
1542 #ifdef DEBUG_RETURNED_NHL
1543 vnc_zlog_debug_verbose(
1544 "%s: already put VN/nexthop %pFX, skip",
1550 if (rfapiGetUnAddrOfVpnBi(bpi
, &pfx_un
)) {
1551 #ifdef DEBUG_ENCAP_MONITOR
1552 vnc_zlog_debug_verbose(
1553 "%s: failed to get UN address of this VPN bpi",
1559 newpfx
= prefix_new();
1561 skiplist_insert(seen_nexthops
, newpfx
, newpfx
);
1563 new = rfapiRouteInfo2NextHopEntry(rprefix
, bpi
, lifetime
, rn
);
1565 if (rfapiRibPreloadBi(rfd_rib_node
, &pfx_vn
, &pfx_un
,
1567 /* duplicate filtered by RIB */
1568 rfapi_free_next_hop_list(new);
1575 (*tail
)->next
= new;
1584 skiplist_free(seen_nexthops
);
1593 * omit_node is meant for the situation where we are adding a subtree
1594 * of a parent of some original requested node. The response already
1595 * contains the original requested node, and we don't want to duplicate
1596 * its routes in the list, so we skip it if the right or left node
1597 * matches (of course, we still travel down its child subtrees).
1599 static int rfapiNhlAddSubtree(
1600 struct agg_node
*rn
, /* in */
1601 uint32_t lifetime
, /* in */
1602 struct rfapi_next_hop_entry
**head
, /* in/out */
1603 struct rfapi_next_hop_entry
**tail
, /* in/out */
1604 struct agg_node
*omit_node
, /* in */
1605 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1606 struct agg_table
*rfd_rib_table
, /* preload here */
1607 struct prefix
*pfx_target_original
) /* query target */
1609 struct rfapi_ip_prefix rprefix
;
1612 /* FIXME: need to find a better way here to work without sticking our
1613 * hands in node->link */
1614 if (agg_node_left(rn
) && agg_node_left(rn
) != omit_node
) {
1615 if (agg_node_left(rn
)->info
) {
1616 const struct prefix
*p
=
1617 agg_node_get_prefix(agg_node_left(rn
));
1619 struct agg_node
*rib_rn
= NULL
;
1621 rfapiQprefix2Rprefix(p
, &rprefix
);
1623 rib_rn
= agg_node_get(rfd_rib_table
, p
);
1625 count
= rfapiNhlAddNodeRoutes(
1626 agg_node_left(rn
), &rprefix
, lifetime
, 0, head
,
1627 tail
, exclude_vnaddr
, rib_rn
,
1628 pfx_target_original
);
1630 count
= rfapiNhlAddNodeRoutes(
1631 agg_node_left(rn
), &rprefix
, lifetime
,
1632 1, head
, tail
, exclude_vnaddr
, rib_rn
,
1633 pfx_target_original
);
1637 agg_unlock_node(rib_rn
);
1641 if (agg_node_right(rn
) && agg_node_right(rn
) != omit_node
) {
1642 if (agg_node_right(rn
)->info
) {
1643 const struct prefix
*p
=
1644 agg_node_get_prefix(agg_node_right(rn
));
1646 struct agg_node
*rib_rn
= NULL
;
1648 rfapiQprefix2Rprefix(p
, &rprefix
);
1650 rib_rn
= agg_node_get(rfd_rib_table
, p
);
1652 count
= rfapiNhlAddNodeRoutes(
1653 agg_node_right(rn
), &rprefix
, lifetime
, 0, head
,
1654 tail
, exclude_vnaddr
, rib_rn
,
1655 pfx_target_original
);
1657 count
= rfapiNhlAddNodeRoutes(
1658 agg_node_right(rn
), &rprefix
, lifetime
,
1659 1, head
, tail
, exclude_vnaddr
, rib_rn
,
1660 pfx_target_original
);
1664 agg_unlock_node(rib_rn
);
1668 if (agg_node_left(rn
)) {
1669 rcount
+= rfapiNhlAddSubtree(
1670 agg_node_left(rn
), lifetime
, head
, tail
, omit_node
,
1671 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1673 if (agg_node_right(rn
)) {
1674 rcount
+= rfapiNhlAddSubtree(
1675 agg_node_right(rn
), lifetime
, head
, tail
, omit_node
,
1676 exclude_vnaddr
, rfd_rib_table
, pfx_target_original
);
1683 * Implementation of ROUTE_LIST(node) from RFAPI-Import-Event-Handling.txt
1685 * Construct an rfapi nexthop list based on the routes attached to
1686 * the specified node.
1688 * If there are any routes that do NOT have BGP_PATH_REMOVED set,
1689 * return those only. If there are ONLY routes with BGP_PATH_REMOVED,
1690 * then return those, and also include all the non-removed routes from the
1691 * next less-specific node (i.e., this node's parent) at the end.
1693 struct rfapi_next_hop_entry
*rfapiRouteNode2NextHopList(
1694 struct agg_node
*rn
, uint32_t lifetime
, /* put into nexthop entries */
1695 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1696 struct agg_table
*rfd_rib_table
, /* preload here */
1697 struct prefix
*pfx_target_original
) /* query target */
1699 struct rfapi_ip_prefix rprefix
;
1700 struct rfapi_next_hop_entry
*answer
= NULL
;
1701 struct rfapi_next_hop_entry
*last
= NULL
;
1702 struct agg_node
*parent
;
1703 const struct prefix
*p
= agg_node_get_prefix(rn
);
1705 struct agg_node
*rib_rn
;
1707 #ifdef DEBUG_RETURNED_NHL
1708 vnc_zlog_debug_verbose("%s: called with node pfx=%rRN", __func__
, rn
);
1709 rfapiDebugBacktrace();
1712 rfapiQprefix2Rprefix(p
, &rprefix
);
1714 rib_rn
= rfd_rib_table
? agg_node_get(rfd_rib_table
, p
) : NULL
;
1717 * Add non-withdrawn routes at this node
1719 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 0, &answer
, &last
,
1720 exclude_vnaddr
, rib_rn
,
1721 pfx_target_original
);
1724 * If the list has at least one entry, it's finished
1727 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
, &last
, NULL
,
1728 exclude_vnaddr
, rfd_rib_table
,
1729 pfx_target_original
);
1730 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
,
1732 #ifdef DEBUG_RETURNED_NHL
1733 rfapiPrintNhl(NULL
, answer
);
1736 agg_unlock_node(rib_rn
);
1741 * Add withdrawn routes at this node
1743 count
= rfapiNhlAddNodeRoutes(rn
, &rprefix
, lifetime
, 1, &answer
, &last
,
1744 exclude_vnaddr
, rib_rn
,
1745 pfx_target_original
);
1747 agg_unlock_node(rib_rn
);
1749 // rfapiPrintNhl(NULL, answer);
1752 * walk up the tree until we find a node with non-deleted
1753 * routes, then add them
1755 for (parent
= agg_node_parent(rn
); parent
;
1756 parent
= agg_node_parent(parent
)) {
1757 if (rfapiHasNonRemovedRoutes(parent
)) {
1763 * Add non-withdrawn routes from less-specific prefix
1766 const struct prefix
*p
= agg_node_get_prefix(parent
);
1768 rib_rn
= rfd_rib_table
? agg_node_get(rfd_rib_table
, p
) : NULL
;
1769 rfapiQprefix2Rprefix(p
, &rprefix
);
1770 count
+= rfapiNhlAddNodeRoutes(parent
, &rprefix
, lifetime
, 0,
1771 &answer
, &last
, exclude_vnaddr
,
1772 rib_rn
, pfx_target_original
);
1773 count
+= rfapiNhlAddSubtree(parent
, lifetime
, &answer
, &last
,
1774 rn
, exclude_vnaddr
, rfd_rib_table
,
1775 pfx_target_original
);
1777 agg_unlock_node(rib_rn
);
1780 * There is no parent with non-removed routes. Still need to
1781 * add subtree of original node if it contributed routes to the
1785 count
+= rfapiNhlAddSubtree(rn
, lifetime
, &answer
,
1786 &last
, rn
, exclude_vnaddr
,
1788 pfx_target_original
);
1791 vnc_zlog_debug_verbose("%s: %d nexthops, answer=%p", __func__
, count
,
1793 #ifdef DEBUG_RETURNED_NHL
1794 rfapiPrintNhl(NULL
, answer
);
1800 * Construct nexthop list of all routes in table
1802 struct rfapi_next_hop_entry
*rfapiRouteTable2NextHopList(
1803 struct agg_table
*rt
, uint32_t lifetime
, /* put into nexthop entries */
1804 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1805 struct agg_table
*rfd_rib_table
, /* preload this NVE rib table */
1806 struct prefix
*pfx_target_original
) /* query target */
1808 struct agg_node
*rn
;
1809 struct rfapi_next_hop_entry
*biglist
= NULL
;
1810 struct rfapi_next_hop_entry
*nhl
;
1811 struct rfapi_next_hop_entry
*tail
= NULL
;
1814 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
1816 nhl
= rfapiRouteNode2NextHopList(rn
, lifetime
, exclude_vnaddr
,
1818 pfx_target_original
);
1820 tail
= biglist
= nhl
;
1827 while (tail
->next
) {
1834 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1838 struct rfapi_next_hop_entry
*rfapiEthRouteNode2NextHopList(
1839 struct agg_node
*rn
, struct rfapi_ip_prefix
*rprefix
,
1840 uint32_t lifetime
, /* put into nexthop entries */
1841 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1842 struct agg_table
*rfd_rib_table
, /* preload NVE rib table */
1843 struct prefix
*pfx_target_original
) /* query target */
1846 struct rfapi_next_hop_entry
*answer
= NULL
;
1847 struct rfapi_next_hop_entry
*last
= NULL
;
1848 struct agg_node
*rib_rn
;
1850 rib_rn
= rfd_rib_table
1851 ? agg_node_get(rfd_rib_table
, agg_node_get_prefix(rn
))
1854 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 0, &answer
, &last
,
1855 NULL
, rib_rn
, pfx_target_original
);
1857 #ifdef DEBUG_ENCAP_MONITOR
1858 vnc_zlog_debug_verbose("%s: node %p: %d non-holddown routes", __func__
,
1863 count
= rfapiNhlAddNodeRoutes(rn
, rprefix
, lifetime
, 1, &answer
,
1864 &last
, exclude_vnaddr
, rib_rn
,
1865 pfx_target_original
);
1866 vnc_zlog_debug_verbose("%s: node %p: %d holddown routes",
1867 __func__
, rn
, count
);
1871 agg_unlock_node(rib_rn
);
1873 #ifdef DEBUG_RETURNED_NHL
1874 rfapiPrintNhl(NULL
, answer
);
1882 * Construct nexthop list of all routes in table
1884 struct rfapi_next_hop_entry
*rfapiEthRouteTable2NextHopList(
1885 uint32_t logical_net_id
, struct rfapi_ip_prefix
*rprefix
,
1886 uint32_t lifetime
, /* put into nexthop entries */
1887 struct rfapi_ip_addr
*exclude_vnaddr
, /* omit routes to same NVE */
1888 struct agg_table
*rfd_rib_table
, /* preload NVE rib node */
1889 struct prefix
*pfx_target_original
) /* query target */
1891 struct rfapi_import_table
*it
;
1892 struct bgp
*bgp
= bgp_get_default();
1893 struct agg_table
*rt
;
1894 struct agg_node
*rn
;
1895 struct rfapi_next_hop_entry
*biglist
= NULL
;
1896 struct rfapi_next_hop_entry
*nhl
;
1897 struct rfapi_next_hop_entry
*tail
= NULL
;
1901 it
= rfapiMacImportTableGet(bgp
, logical_net_id
);
1902 rt
= it
->imported_vpn
[AFI_L2VPN
];
1904 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
1906 nhl
= rfapiEthRouteNode2NextHopList(
1907 rn
, rprefix
, lifetime
, exclude_vnaddr
, rfd_rib_table
,
1908 pfx_target_original
);
1910 tail
= biglist
= nhl
;
1917 while (tail
->next
) {
1924 vnc_zlog_debug_verbose("%s: returning %d routes", __func__
, count
);
1929 * Insert a new bpi to the imported route table node,
1930 * keeping the list of BPIs sorted best route first
1932 static void rfapiBgpInfoAttachSorted(struct agg_node
*rn
,
1933 struct bgp_path_info
*info_new
, afi_t afi
,
1937 struct bgp_path_info
*prev
;
1938 struct bgp_path_info
*next
;
1939 char pfx_buf
[PREFIX2STR_BUFFER
];
1942 bgp
= bgp_get_default(); /* assume 1 instance for now */
1944 if (VNC_DEBUG(IMPORT_BI_ATTACH
)) {
1945 vnc_zlog_debug_verbose("%s: info_new->peer=%p", __func__
,
1947 vnc_zlog_debug_verbose("%s: info_new->peer->su_remote=%p",
1948 __func__
, info_new
->peer
->su_remote
);
1951 for (prev
= NULL
, next
= rn
->info
; next
;
1952 prev
= next
, next
= next
->next
) {
1953 enum bgp_path_selection_reason reason
;
1956 || (!CHECK_FLAG(info_new
->flags
, BGP_PATH_REMOVED
)
1957 && CHECK_FLAG(next
->flags
, BGP_PATH_REMOVED
))
1958 || bgp_path_info_cmp_compatible(bgp
, info_new
, next
,
1961 == -1) { /* -1 if 1st is better */
1965 vnc_zlog_debug_verbose("%s: prev=%p, next=%p", __func__
, prev
, next
);
1967 prev
->next
= info_new
;
1969 rn
->info
= info_new
;
1971 info_new
->prev
= prev
;
1972 info_new
->next
= next
;
1974 next
->prev
= info_new
;
1975 bgp_attr_intern(info_new
->attr
);
1978 static void rfapiBgpInfoDetach(struct agg_node
*rn
, struct bgp_path_info
*bpi
)
1981 * Remove the route (doubly-linked)
1983 // bgp_attr_unintern (&bpi->attr);
1985 bpi
->next
->prev
= bpi
->prev
;
1987 bpi
->prev
->next
= bpi
->next
;
1989 rn
->info
= bpi
->next
;
1993 * For L3-indexed import tables
1995 static int rfapi_bi_peer_rd_cmp(const void *b1
, const void *b2
)
1997 const struct bgp_path_info
*bpi1
= b1
;
1998 const struct bgp_path_info
*bpi2
= b2
;
2003 if (bpi1
->peer
< bpi2
->peer
)
2005 if (bpi1
->peer
> bpi2
->peer
)
2011 return vnc_prefix_cmp(
2012 (const struct prefix
*)&bpi1
->extra
->vnc
.import
.rd
,
2013 (const struct prefix
*)&bpi2
->extra
->vnc
.import
.rd
);
2017 * For L2-indexed import tables
2018 * The BPIs in these tables should ALWAYS have an aux_prefix set because
2019 * they arrive via IPv4 or IPv6 advertisements.
2021 static int rfapi_bi_peer_rd_aux_cmp(const void *b1
, const void *b2
)
2023 const struct bgp_path_info
*bpi1
= b1
;
2024 const struct bgp_path_info
*bpi2
= b2
;
2030 if (bpi1
->peer
< bpi2
->peer
)
2032 if (bpi1
->peer
> bpi2
->peer
)
2038 rc
= vnc_prefix_cmp((struct prefix
*)&bpi1
->extra
->vnc
.import
.rd
,
2039 (struct prefix
*)&bpi2
->extra
->vnc
.import
.rd
);
2045 * L2 import tables can have multiple entries with the
2046 * same MAC address, same RD, but different L3 addresses.
2048 * Use presence of aux_prefix with AF=ethernet and prefixlen=1
2049 * as magic value to signify explicit wildcarding of the aux_prefix.
2050 * This magic value will not appear in bona fide bpi entries in
2051 * the import table, but is allowed in the "fake" bpi used to
2052 * probe the table when searching. (We have to test both b1 and b2
2053 * because there is no guarantee of the order the test key and
2054 * the real key will be passed)
2056 if ((bpi1
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2057 && (bpi1
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))
2058 || (bpi2
->extra
->vnc
.import
.aux_prefix
.family
== AF_ETHERNET
2059 && (bpi2
->extra
->vnc
.import
.aux_prefix
.prefixlen
== 1))) {
2062 * wildcard aux address specified
2067 return vnc_prefix_cmp(&bpi1
->extra
->vnc
.import
.aux_prefix
,
2068 &bpi2
->extra
->vnc
.import
.aux_prefix
);
2073 * Index on RD and Peer
2075 static void rfapiItBiIndexAdd(struct agg_node
*rn
, /* Import table VPN node */
2076 struct bgp_path_info
*bpi
) /* new BPI */
2078 struct skiplist
*sl
;
2079 const struct prefix
*p
;
2085 vnc_zlog_debug_verbose("%s: bpi %p, peer %p, rd %pRDP", __func__
, bpi
,
2086 bpi
->peer
, &bpi
->extra
->vnc
.import
.rd
);
2088 sl
= RFAPI_RDINDEX_W_ALLOC(rn
);
2090 p
= agg_node_get_prefix(rn
);
2091 if (AF_ETHERNET
== p
->family
) {
2092 sl
= skiplist_new(0, rfapi_bi_peer_rd_aux_cmp
, NULL
);
2094 sl
= skiplist_new(0, rfapi_bi_peer_rd_cmp
, NULL
);
2096 RFAPI_IT_EXTRA_GET(rn
)->u
.vpn
.idx_rd
= sl
;
2097 agg_lock_node(rn
); /* for skiplist */
2099 assert(!skiplist_insert(sl
, (void *)bpi
, (void *)bpi
));
2100 agg_lock_node(rn
); /* for skiplist entry */
2102 /* NB: BPIs in import tables are not refcounted */
2105 static void rfapiItBiIndexDump(struct agg_node
*rn
)
2107 struct skiplist
*sl
;
2108 void *cursor
= NULL
;
2109 struct bgp_path_info
*k
;
2110 struct bgp_path_info
*v
;
2113 sl
= RFAPI_RDINDEX(rn
);
2117 for (rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
); !rc
;
2118 rc
= skiplist_next(sl
, (void **)&k
, (void **)&v
, &cursor
)) {
2120 char buf
[RD_ADDRSTRLEN
];
2121 char buf_aux_pfx
[PREFIX_STRLEN
];
2124 &k
->extra
->vnc
.import
.rd
, buf
, sizeof(buf
),
2125 bgp_get_asnotation(k
->peer
? k
->peer
->bgp
: NULL
));
2126 if (k
->extra
->vnc
.import
.aux_prefix
.family
) {
2127 prefix2str(&k
->extra
->vnc
.import
.aux_prefix
,
2128 buf_aux_pfx
, sizeof(buf_aux_pfx
));
2130 strlcpy(buf_aux_pfx
, "(none)", sizeof(buf_aux_pfx
));
2132 vnc_zlog_debug_verbose("bpi %p, peer %p, rd %s, aux_prefix %s",
2133 k
, k
->peer
, buf
, buf_aux_pfx
);
2137 static struct bgp_path_info
*rfapiItBiIndexSearch(
2138 struct agg_node
*rn
, /* Import table VPN node */
2139 struct prefix_rd
*prd
, struct peer
*peer
,
2140 const struct prefix
*aux_prefix
) /* optional L3 addr for L2 ITs */
2142 struct skiplist
*sl
;
2144 struct bgp_path_info bpi_fake
= {0};
2145 struct bgp_path_info_extra bpi_extra
= {0};
2146 struct bgp_path_info
*bpi_result
;
2148 sl
= RFAPI_RDINDEX(rn
);
2152 #ifdef DEBUG_BI_SEARCH
2154 char buf_aux_pfx
[PREFIX_STRLEN
];
2157 prefix2str(aux_prefix
, buf_aux_pfx
,
2158 sizeof(buf_aux_pfx
));
2160 strlcpy(buf_aux_pfx
, "(nil)", sizeof(buf_aux_pfx
));
2162 vnc_zlog_debug_verbose(
2163 "%s want prd=%pRDP, peer=%p, aux_prefix=%s", __func__
,
2164 prd
, peer
, buf_aux_pfx
);
2165 rfapiItBiIndexDump(rn
);
2169 /* threshold is a WAG */
2170 if (sl
->count
< 3) {
2171 #ifdef DEBUG_BI_SEARCH
2172 vnc_zlog_debug_verbose("%s: short list algorithm", __func__
);
2174 /* if short list, linear search might be faster */
2175 for (bpi_result
= rn
->info
; bpi_result
;
2176 bpi_result
= bpi_result
->next
) {
2177 #ifdef DEBUG_BI_SEARCH
2178 vnc_zlog_debug_verbose(
2179 "%s: bpi has prd=%pRDP, peer=%p", __func__
,
2180 &bpi_result
->extra
->vnc
.import
.rd
,
2183 if (peer
== bpi_result
->peer
2184 && !prefix_cmp((struct prefix
*)&bpi_result
->extra
2186 (struct prefix
*)prd
)) {
2188 #ifdef DEBUG_BI_SEARCH
2189 vnc_zlog_debug_verbose(
2190 "%s: peer and RD same, doing aux_prefix check",
2196 &bpi_result
->extra
->vnc
.import
2199 #ifdef DEBUG_BI_SEARCH
2200 vnc_zlog_debug_verbose("%s: match",
2210 bpi_fake
.peer
= peer
;
2211 bpi_fake
.extra
= &bpi_extra
;
2212 bpi_fake
.extra
->vnc
.import
.rd
= *prd
;
2214 bpi_fake
.extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
2217 bpi_fake
.extra
->vnc
.import
.aux_prefix
.family
= AF_ETHERNET
;
2218 bpi_fake
.extra
->vnc
.import
.aux_prefix
.prefixlen
= 1;
2221 rc
= skiplist_search(sl
, (void *)&bpi_fake
, (void *)&bpi_result
);
2224 #ifdef DEBUG_BI_SEARCH
2225 vnc_zlog_debug_verbose("%s: no match", __func__
);
2230 #ifdef DEBUG_BI_SEARCH
2231 vnc_zlog_debug_verbose("%s: matched bpi=%p", __func__
, bpi_result
);
2237 static void rfapiItBiIndexDel(struct agg_node
*rn
, /* Import table VPN node */
2238 struct bgp_path_info
*bpi
) /* old BPI */
2240 struct skiplist
*sl
;
2243 vnc_zlog_debug_verbose("%s: bpi %p, peer %p, rd %pRDP", __func__
, bpi
,
2244 bpi
->peer
, &bpi
->extra
->vnc
.import
.rd
);
2246 sl
= RFAPI_RDINDEX(rn
);
2249 rc
= skiplist_delete(sl
, (void *)(bpi
), (void *)bpi
);
2251 rfapiItBiIndexDump(rn
);
2255 agg_unlock_node(rn
); /* for skiplist entry */
2257 /* NB: BPIs in import tables are not refcounted */
2261 * Add a backreference at the ENCAP node to the VPN route that
2265 rfapiMonitorEncapAdd(struct rfapi_import_table
*import_table
,
2266 struct prefix
*p
, /* VN address */
2267 struct agg_node
*vpn_rn
, /* VPN node */
2268 struct bgp_path_info
*vpn_bpi
) /* VPN bpi/route */
2270 afi_t afi
= family2afi(p
->family
);
2271 struct agg_node
*rn
;
2272 struct rfapi_monitor_encap
*m
;
2275 rn
= agg_node_get(import_table
->imported_encap
[afi
], p
); /* locks rn */
2278 m
= XCALLOC(MTYPE_RFAPI_MONITOR_ENCAP
,
2279 sizeof(struct rfapi_monitor_encap
));
2285 /* insert to encap node's list */
2286 m
->next
= RFAPI_MONITOR_ENCAP(rn
);
2289 RFAPI_MONITOR_ENCAP_W_ALLOC(rn
) = m
;
2291 /* for easy lookup when deleting vpn route */
2292 vpn_bpi
->extra
->vnc
.import
.hme
= m
;
2294 vnc_zlog_debug_verbose(
2295 "%s: it=%p, vpn_bpi=%p, afi=%d, encap rn=%p, setting vpn_bpi->extra->vnc.import.hme=%p",
2296 __func__
, import_table
, vpn_bpi
, afi
, rn
, m
);
2298 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
2299 bgp_attr_intern(vpn_bpi
->attr
);
2302 static void rfapiMonitorEncapDelete(struct bgp_path_info
*vpn_bpi
)
2305 * Remove encap monitor
2307 vnc_zlog_debug_verbose("%s: vpn_bpi=%p", __func__
, vpn_bpi
);
2308 if (vpn_bpi
->extra
) {
2309 struct rfapi_monitor_encap
*hme
=
2310 vpn_bpi
->extra
->vnc
.import
.hme
;
2314 vnc_zlog_debug_verbose("%s: hme=%p", __func__
, hme
);
2316 /* Refcount checking takes too long here */
2317 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 0);
2319 hme
->next
->prev
= hme
->prev
;
2321 hme
->prev
->next
= hme
->next
;
2323 RFAPI_MONITOR_ENCAP_W_ALLOC(hme
->rn
) =
2325 /* Refcount checking takes too long here */
2326 // RFAPI_CHECK_REFCOUNT(hme->rn, SAFI_ENCAP, 1);
2328 /* see if the struct rfapi_it_extra is empty and can be
2330 rfapiMonitorExtraPrune(SAFI_ENCAP
, hme
->rn
);
2332 agg_unlock_node(hme
->rn
); /* decr ref count */
2333 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, hme
);
2334 vpn_bpi
->extra
->vnc
.import
.hme
= NULL
;
2340 * quagga lib/thread.h says this must return int even though
2341 * it doesn't do anything with the return value
2343 static void rfapiWithdrawTimerVPN(struct thread
*t
)
2345 struct rfapi_withdraw
*wcb
= THREAD_ARG(t
);
2346 struct bgp_path_info
*bpi
= wcb
->info
;
2347 struct bgp
*bgp
= bgp_get_default();
2348 const struct prefix
*p
;
2349 struct rfapi_monitor_vpn
*moved
;
2353 vnc_zlog_debug_verbose(
2354 "%s: NULL BGP pointer, assume shutdown race condition!!!",
2358 if (CHECK_FLAG(bgp
->flags
, BGP_FLAG_DELETE_IN_PROGRESS
)) {
2359 vnc_zlog_debug_verbose(
2360 "%s: BGP delete in progress, assume shutdown race condition!!!",
2366 assert(wcb
->import_table
);
2369 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, wcb
->lockoffset
);
2371 vnc_zlog_debug_verbose("%s: removing bpi %p at prefix %pRN", __func__
,
2375 * Remove the route (doubly-linked)
2377 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_VALID
)
2378 && VALID_INTERIOR_TYPE(bpi
->type
))
2379 RFAPI_MONITOR_EXTERIOR(wcb
->node
)->valid_interior_count
--;
2381 p
= agg_node_get_prefix(wcb
->node
);
2382 afi
= family2afi(p
->family
);
2383 wcb
->import_table
->holddown_count
[afi
] -= 1; /* keep count consistent */
2384 rfapiItBiIndexDel(wcb
->node
, bpi
);
2385 rfapiBgpInfoDetach(wcb
->node
, bpi
); /* with removed bpi */
2387 vnc_import_bgp_exterior_del_route_interior(bgp
, wcb
->import_table
,
2392 * If VNC is configured to send response remove messages, AND
2393 * if the removed route had a UN address, do response removal
2396 if (!(bgp
->rfapi_cfg
->flags
2397 & BGP_VNC_CONFIG_RESPONSE_REMOVAL_DISABLE
)) {
2399 int has_valid_duplicate
= 0;
2400 struct bgp_path_info
*bpii
;
2403 * First check if there are any OTHER routes at this node
2404 * that have the same nexthop and a valid UN address. If
2405 * there are (e.g., from other peers), then the route isn't
2406 * really gone, so skip sending a response removal message.
2408 for (bpii
= wcb
->node
->info
; bpii
; bpii
= bpii
->next
) {
2409 if (rfapiVpnBiSamePtUn(bpi
, bpii
)) {
2410 has_valid_duplicate
= 1;
2415 vnc_zlog_debug_verbose("%s: has_valid_duplicate=%d", __func__
,
2416 has_valid_duplicate
);
2418 if (!has_valid_duplicate
) {
2419 rfapiRibPendingDeleteRoute(bgp
, wcb
->import_table
, afi
,
2424 rfapiMonitorEncapDelete(bpi
);
2427 * If there are no VPN monitors at this VPN Node A,
2430 if (!RFAPI_MONITOR_VPN(wcb
->node
)) {
2431 vnc_zlog_debug_verbose("%s: no VPN monitors at this node",
2437 * rfapiMonitorMoveShorter only moves monitors if there are
2438 * no remaining valid routes at the current node
2440 moved
= rfapiMonitorMoveShorter(wcb
->node
, 1);
2443 rfapiMonitorMovedUp(wcb
->import_table
, wcb
->node
, moved
->node
,
2451 rfapiBgpInfoFree(bpi
);
2455 * If route count at this node has gone to 0, withdraw exported prefix
2457 if (!wcb
->node
->info
) {
2458 /* see if the struct rfapi_it_extra is empty and can be freed */
2459 rfapiMonitorExtraPrune(SAFI_MPLS_VPN
, wcb
->node
);
2460 vnc_direct_bgp_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2461 vnc_zebra_del_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2464 * nexthop change event
2465 * vnc_direct_bgp_add_prefix() will recompute the VN addr
2468 vnc_direct_bgp_add_prefix(bgp
, wcb
->import_table
, wcb
->node
);
2471 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_MPLS_VPN
, 1 + wcb
->lockoffset
);
2472 agg_unlock_node(wcb
->node
); /* decr ref count */
2473 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2477 * This works for multiprotocol extension, but not for plain ol'
2478 * unicast IPv4 because that nexthop is stored in attr->nexthop
2480 void rfapiNexthop2Prefix(struct attr
*attr
, struct prefix
*p
)
2485 memset(p
, 0, sizeof(struct prefix
));
2487 switch (p
->family
= BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2489 p
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2490 p
->prefixlen
= IPV4_MAX_BITLEN
;
2494 p
->u
.prefix6
= attr
->mp_nexthop_global
;
2495 p
->prefixlen
= IPV6_MAX_BITLEN
;
2499 vnc_zlog_debug_verbose("%s: Family is unknown = %d", __func__
,
2504 void rfapiUnicastNexthop2Prefix(afi_t afi
, struct attr
*attr
, struct prefix
*p
)
2506 if (afi
== AFI_IP
) {
2507 p
->family
= AF_INET
;
2508 p
->prefixlen
= IPV4_MAX_BITLEN
;
2509 p
->u
.prefix4
= attr
->nexthop
;
2511 rfapiNexthop2Prefix(attr
, p
);
2515 static int rfapiAttrNexthopAddrDifferent(struct prefix
*p1
, struct prefix
*p2
)
2518 vnc_zlog_debug_verbose("%s: p1 or p2 is NULL", __func__
);
2523 * Are address families the same?
2525 if (p1
->family
!= p2
->family
) {
2529 switch (p1
->family
) {
2531 if (IPV4_ADDR_SAME(&p1
->u
.prefix4
, &p2
->u
.prefix4
))
2536 if (IPV6_ADDR_SAME(&p1
->u
.prefix6
, &p2
->u
.prefix6
))
2547 static void rfapiCopyUnEncap2VPN(struct bgp_path_info
*encap_bpi
,
2548 struct bgp_path_info
*vpn_bpi
)
2550 if (!vpn_bpi
|| !vpn_bpi
->extra
) {
2551 zlog_warn("%s: no vpn bpi attr/extra, can't copy UN address",
2556 switch (BGP_MP_NEXTHOP_FAMILY(encap_bpi
->attr
->mp_nexthop_len
)) {
2560 * instrumentation to debug segfault of 091127
2562 vnc_zlog_debug_verbose("%s: vpn_bpi=%p", __func__
, vpn_bpi
);
2563 vnc_zlog_debug_verbose("%s: vpn_bpi->extra=%p", __func__
,
2566 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_INET
;
2567 vpn_bpi
->extra
->vnc
.import
.un
.addr4
=
2568 encap_bpi
->attr
->mp_nexthop_global_in
;
2572 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_INET6
;
2573 vpn_bpi
->extra
->vnc
.import
.un
.addr6
=
2574 encap_bpi
->attr
->mp_nexthop_global
;
2578 zlog_warn("%s: invalid encap nexthop length: %d", __func__
,
2579 encap_bpi
->attr
->mp_nexthop_len
);
2580 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_UNSPEC
;
2586 * returns 0 on success, nonzero on error
2589 rfapiWithdrawEncapUpdateCachedUn(struct rfapi_import_table
*import_table
,
2590 struct bgp_path_info
*encap_bpi
,
2591 struct agg_node
*vpn_rn
,
2592 struct bgp_path_info
*vpn_bpi
)
2597 * clear cached UN address
2599 if (!vpn_bpi
|| !vpn_bpi
->extra
) {
2601 "%s: missing VPN bpi/extra, can't clear UN addr",
2605 vpn_bpi
->extra
->vnc
.import
.un_family
= AF_UNSPEC
;
2606 memset(&vpn_bpi
->extra
->vnc
.import
.un
, 0,
2607 sizeof(vpn_bpi
->extra
->vnc
.import
.un
));
2608 if (CHECK_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
)) {
2609 if (rfapiGetVncTunnelUnAddr(vpn_bpi
->attr
, NULL
)) {
2610 UNSET_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
);
2611 if (VALID_INTERIOR_TYPE(vpn_bpi
->type
))
2612 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2613 ->valid_interior_count
--;
2614 /* signal interior route withdrawal to
2615 * import-exterior */
2616 vnc_import_bgp_exterior_del_route_interior(
2617 bgp_get_default(), import_table
, vpn_rn
,
2624 zlog_warn("%s: missing VPN bpi, can't clear UN addr",
2628 rfapiCopyUnEncap2VPN(encap_bpi
, vpn_bpi
);
2629 if (!CHECK_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
)) {
2630 SET_FLAG(vpn_bpi
->flags
, BGP_PATH_VALID
);
2631 if (VALID_INTERIOR_TYPE(vpn_bpi
->type
))
2632 RFAPI_MONITOR_EXTERIOR(vpn_rn
)
2633 ->valid_interior_count
++;
2634 /* signal interior route withdrawal to import-exterior
2636 vnc_import_bgp_exterior_add_route_interior(
2637 bgp_get_default(), import_table
, vpn_rn
,
2644 static void rfapiWithdrawTimerEncap(struct thread
*t
)
2646 struct rfapi_withdraw
*wcb
= THREAD_ARG(t
);
2647 struct bgp_path_info
*bpi
= wcb
->info
;
2648 int was_first_route
= 0;
2649 struct rfapi_monitor_encap
*em
;
2650 struct skiplist
*vpn_node_sl
= skiplist_new(0, NULL
, NULL
);
2654 assert(wcb
->import_table
);
2656 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 0);
2658 if (wcb
->node
->info
== bpi
)
2659 was_first_route
= 1;
2662 * Remove the route/bpi and free it
2664 rfapiBgpInfoDetach(wcb
->node
, bpi
);
2665 rfapiBgpInfoFree(bpi
);
2667 if (!was_first_route
)
2670 for (em
= RFAPI_MONITOR_ENCAP(wcb
->node
); em
; em
= em
->next
) {
2673 * Update monitoring VPN BPIs with new encap info at the
2674 * head of the encap bpi chain (which could be NULL after
2675 * removing the expiring bpi above)
2677 if (rfapiWithdrawEncapUpdateCachedUn(wcb
->import_table
,
2678 wcb
->node
->info
, em
->node
,
2683 * Build a list of unique VPN nodes referenced by these
2685 * Use a skiplist for speed.
2687 skiplist_insert(vpn_node_sl
, em
->node
, em
->node
);
2692 * for each VPN node referenced in the ENCAP monitors:
2694 struct agg_node
*rn
;
2695 while (!skiplist_first(vpn_node_sl
, (void **)&rn
, NULL
)) {
2696 if (!wcb
->node
->info
) {
2697 struct rfapi_monitor_vpn
*moved
;
2699 moved
= rfapiMonitorMoveShorter(rn
, 0);
2701 // rfapiDoRouteCallback(wcb->import_table,
2702 // moved->node, moved);
2703 rfapiMonitorMovedUp(wcb
->import_table
, rn
,
2704 moved
->node
, moved
);
2707 // rfapiDoRouteCallback(wcb->import_table, rn, NULL);
2708 rfapiMonitorItNodeChanged(wcb
->import_table
, rn
, NULL
);
2710 skiplist_delete_first(vpn_node_sl
);
2714 RFAPI_CHECK_REFCOUNT(wcb
->node
, SAFI_ENCAP
, 1);
2715 agg_unlock_node(wcb
->node
); /* decr ref count */
2716 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
2717 skiplist_free(vpn_node_sl
);
2722 * Works for both VPN and ENCAP routes; timer_service_func is different
2726 rfapiBiStartWithdrawTimer(struct rfapi_import_table
*import_table
,
2727 struct agg_node
*rn
, struct bgp_path_info
*bpi
,
2728 afi_t afi
, safi_t safi
,
2729 void (*timer_service_func
)(struct thread
*))
2732 struct rfapi_withdraw
*wcb
;
2734 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
2736 * Already on the path to being withdrawn,
2737 * should already have a timer set up to
2740 vnc_zlog_debug_verbose(
2741 "%s: already being withdrawn, do nothing", __func__
);
2745 rfapiGetVncLifetime(bpi
->attr
, &lifetime
);
2746 vnc_zlog_debug_verbose("%s: VNC lifetime is %u", __func__
, lifetime
);
2749 * withdrawn routes get to hang around for a while
2751 SET_FLAG(bpi
->flags
, BGP_PATH_REMOVED
);
2753 /* set timer to remove the route later */
2754 lifetime
= rfapiGetHolddownFromLifetime(lifetime
);
2755 vnc_zlog_debug_verbose("%s: using timeout %u", __func__
, lifetime
);
2758 * Stash import_table, node, and info for use by timer
2759 * service routine, which is supposed to free the wcb.
2761 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2764 wcb
->import_table
= import_table
;
2765 bgp_attr_intern(bpi
->attr
);
2767 if (VNC_DEBUG(VERBOSE
)) {
2768 vnc_zlog_debug_verbose(
2769 "%s: wcb values: node=%p, info=%p, import_table=%p (bpi follows)",
2770 __func__
, wcb
->node
, wcb
->info
, wcb
->import_table
);
2771 rfapiPrintBi(NULL
, bpi
);
2776 if (lifetime
> UINT32_MAX
/ 1001) {
2777 /* sub-optimal case, but will probably never happen */
2778 bpi
->extra
->vnc
.import
.timer
= NULL
;
2779 thread_add_timer(bm
->master
, timer_service_func
, wcb
, lifetime
,
2780 &bpi
->extra
->vnc
.import
.timer
);
2782 static uint32_t jitter
;
2783 uint32_t lifetime_msec
;
2786 * the goal here is to spread out the timers so they are
2787 * sortable in the skip list
2789 if (++jitter
>= 1000)
2792 lifetime_msec
= (lifetime
* 1000) + jitter
;
2794 bpi
->extra
->vnc
.import
.timer
= NULL
;
2795 thread_add_timer_msec(bm
->master
, timer_service_func
, wcb
,
2797 &bpi
->extra
->vnc
.import
.timer
);
2800 /* re-sort route list (BGP_PATH_REMOVED routes are last) */
2801 if (((struct bgp_path_info
*)rn
->info
)->next
) {
2802 rfapiBgpInfoDetach(rn
, bpi
);
2803 rfapiBgpInfoAttachSorted(rn
, bpi
, afi
, safi
);
2808 typedef void(rfapi_bi_filtered_import_f
)(struct rfapi_import_table
*table
,
2809 int action
, struct peer
*peer
,
2810 void *rfd
, const struct prefix
*prefix
,
2811 const struct prefix
*aux_prefix
,
2812 afi_t afi
, struct prefix_rd
*prd
,
2813 struct attr
*attr
, uint8_t type
,
2814 uint8_t sub_type
, uint32_t *label
);
2817 static void rfapiExpireEncapNow(struct rfapi_import_table
*it
,
2818 struct agg_node
*rn
, struct bgp_path_info
*bpi
)
2820 struct rfapi_withdraw
*wcb
;
2824 * pretend we're an expiring timer
2826 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
2829 wcb
->import_table
= it
;
2830 memset(&t
, 0, sizeof(t
));
2832 rfapiWithdrawTimerEncap(&t
); /* frees wcb */
2835 static int rfapiGetNexthop(struct attr
*attr
, struct prefix
*prefix
)
2837 switch (BGP_MP_NEXTHOP_FAMILY(attr
->mp_nexthop_len
)) {
2839 prefix
->family
= AF_INET
;
2840 prefix
->prefixlen
= IPV4_MAX_BITLEN
;
2841 prefix
->u
.prefix4
= attr
->mp_nexthop_global_in
;
2844 prefix
->family
= AF_INET6
;
2845 prefix
->prefixlen
= IPV6_MAX_BITLEN
;
2846 prefix
->u
.prefix6
= attr
->mp_nexthop_global
;
2849 vnc_zlog_debug_verbose("%s: unknown attr->mp_nexthop_len %d",
2850 __func__
, attr
->mp_nexthop_len
);
2857 * import a bgp_path_info if its route target list intersects with the
2858 * import table's route target list
2860 static void rfapiBgpInfoFilteredImportEncap(
2861 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
2862 void *rfd
, /* set for looped back routes */
2863 const struct prefix
*p
,
2864 const struct prefix
*aux_prefix
, /* Unused for encap routes */
2865 afi_t afi
, struct prefix_rd
*prd
,
2866 struct attr
*attr
, /* part of bgp_path_info */
2867 uint8_t type
, /* part of bgp_path_info */
2868 uint8_t sub_type
, /* part of bgp_path_info */
2869 uint32_t *label
) /* part of bgp_path_info */
2871 struct agg_table
*rt
= NULL
;
2872 struct agg_node
*rn
;
2873 struct bgp_path_info
*info_new
;
2874 struct bgp_path_info
*bpi
;
2875 struct bgp_path_info
*next
;
2878 struct prefix p_firstbpi_old
;
2879 struct prefix p_firstbpi_new
;
2881 const char *action_str
= NULL
;
2882 struct prefix un_prefix
;
2885 bgp
= bgp_get_default(); /* assume 1 instance for now */
2888 case FIF_ACTION_UPDATE
:
2889 action_str
= "update";
2891 case FIF_ACTION_WITHDRAW
:
2892 action_str
= "withdraw";
2894 case FIF_ACTION_KILL
:
2895 action_str
= "kill";
2902 vnc_zlog_debug_verbose(
2903 "%s: entry: %s: prefix %s/%d", __func__
, action_str
,
2904 inet_ntop(p
->family
, &p
->u
.prefix
, buf
, sizeof(buf
)),
2907 memset(&p_firstbpi_old
, 0, sizeof(p_firstbpi_old
));
2908 memset(&p_firstbpi_new
, 0, sizeof(p_firstbpi_new
));
2910 if (action
== FIF_ACTION_UPDATE
) {
2912 * Compare rt lists. If no intersection, don't import this route
2913 * On a withdraw, peer and RD are sufficient to determine if
2916 if (!attr
|| !bgp_attr_get_ecommunity(attr
)) {
2918 vnc_zlog_debug_verbose(
2919 "%s: attr, extra, or ecommunity missing, not importing",
2923 #ifdef RFAPI_REQUIRE_ENCAP_BEEC
2924 if (!rfapiEcommunitiesMatchBeec(
2925 bgp_attr_get_ecommunity(attr
))) {
2926 vnc_zlog_debug_verbose(
2927 "%s: it=%p: no match for BGP Encapsulation ecommunity",
2928 __func__
, import_table
);
2932 if (!rfapiEcommunitiesIntersect(
2933 import_table
->rt_import_list
,
2934 bgp_attr_get_ecommunity(attr
))) {
2936 vnc_zlog_debug_verbose(
2937 "%s: it=%p: no ecommunity intersection",
2938 __func__
, import_table
);
2943 * Updates must also have a nexthop address
2945 memset(&un_prefix
, 0,
2946 sizeof(un_prefix
)); /* keep valgrind happy */
2947 if (rfapiGetNexthop(attr
, &un_prefix
)) {
2948 vnc_zlog_debug_verbose("%s: missing nexthop address",
2955 * Figure out which radix tree the route would go into
2960 rt
= import_table
->imported_encap
[afi
];
2966 flog_err(EC_LIB_DEVELOPMENT
, "%s: bad afi %d", __func__
, afi
);
2971 * agg_node_lookup returns a node only if there is at least
2972 * one route attached.
2974 rn
= agg_node_lookup(rt
, p
);
2976 #ifdef DEBUG_ENCAP_MONITOR
2977 vnc_zlog_debug_verbose("%s: initial encap lookup(it=%p) rn=%p",
2978 __func__
, import_table
, rn
);
2983 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 1);
2984 agg_unlock_node(rn
); /* undo lock in agg_node_lookup */
2988 * capture nexthop of first bpi
2991 rfapiNexthop2Prefix(
2992 ((struct bgp_path_info
*)(rn
->info
))->attr
,
2996 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
2999 * Does this bgp_path_info refer to the same route
3000 * as we are trying to add?
3002 vnc_zlog_debug_verbose("%s: comparing BPI %p", __func__
,
3009 * RD of import table bpi is in
3010 * bpi->extra->vnc.import.rd RD of info_orig is in prd
3013 vnc_zlog_debug_verbose("%s: no bpi->extra",
3018 (struct prefix
*)&bpi
->extra
->vnc
.import
.rd
,
3019 (struct prefix
*)prd
)) {
3021 vnc_zlog_debug_verbose("%s: prd does not match",
3029 if (bpi
->peer
!= peer
) {
3030 vnc_zlog_debug_verbose(
3031 "%s: peer does not match", __func__
);
3035 vnc_zlog_debug_verbose("%s: found matching bpi",
3038 /* Same route. Delete this bpi, replace with new one */
3040 if (action
== FIF_ACTION_WITHDRAW
) {
3042 vnc_zlog_debug_verbose(
3043 "%s: withdrawing at prefix %pRN",
3046 rfapiBiStartWithdrawTimer(
3047 import_table
, rn
, bpi
, afi
, SAFI_ENCAP
,
3048 rfapiWithdrawTimerEncap
);
3051 vnc_zlog_debug_verbose(
3052 "%s: %s at prefix %pRN", __func__
,
3053 ((action
== FIF_ACTION_KILL
)
3059 * If this route is waiting to be deleted
3061 * a previous withdraw, we must cancel its
3064 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
3065 && bpi
->extra
->vnc
.import
.timer
) {
3066 struct rfapi_withdraw
*wcb
= THREAD_ARG(
3067 bpi
->extra
->vnc
.import
.timer
);
3069 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3071 bpi
->extra
->vnc
.import
.timer
);
3074 if (action
== FIF_ACTION_UPDATE
) {
3075 rfapiBgpInfoDetach(rn
, bpi
);
3076 rfapiBgpInfoFree(bpi
);
3080 * Kill: do export stuff when removing
3083 struct rfapi_withdraw
*wcb
;
3087 * pretend we're an expiring timer
3090 MTYPE_RFAPI_WITHDRAW
,
3091 sizeof(struct rfapi_withdraw
));
3094 wcb
->import_table
= import_table
;
3095 memset(&t
, 0, sizeof(t
));
3097 rfapiWithdrawTimerEncap(
3098 &t
); /* frees wcb */
3107 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, replacing
? 1 : 0);
3109 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
)
3113 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, NULL
);
3117 agg_lock_node(rn
); /* incr ref count for new BPI */
3119 rn
= agg_node_get(rt
, p
);
3122 vnc_zlog_debug_verbose("%s: (afi=%d, rn=%p) inserting at prefix %pRN",
3123 __func__
, afi
, rn
, rn
);
3125 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_ENCAP
);
3128 * Delete holddown routes from same NVE. See details in
3129 * rfapiBgpInfoFilteredImportVPN()
3131 for (bpi
= info_new
->next
; bpi
; bpi
= next
) {
3133 struct prefix pfx_un
;
3137 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
))
3141 * We already match the VN address (it is the prefix
3142 * of the route node)
3145 if (!rfapiGetNexthop(bpi
->attr
, &pfx_un
)
3146 && prefix_same(&pfx_un
, &un_prefix
)) {
3154 vnc_zlog_debug_verbose(
3155 "%s: removing holddown bpi matching NVE of new route",
3157 if (bpi
->extra
->vnc
.import
.timer
) {
3158 struct rfapi_withdraw
*wcb
=
3159 THREAD_ARG(bpi
->extra
->vnc
.import
.timer
);
3161 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3162 THREAD_OFF(bpi
->extra
->vnc
.import
.timer
);
3164 rfapiExpireEncapNow(import_table
, rn
, bpi
);
3167 rfapiNexthop2Prefix(((struct bgp_path_info
*)(rn
->info
))->attr
,
3171 * If the nexthop address of the selected Encap route (i.e.,
3172 * the UN address) has changed, then we must update the VPN
3173 * routes that refer to this Encap route and possibly force
3176 if (rfapiAttrNexthopAddrDifferent(&p_firstbpi_old
, &p_firstbpi_new
)) {
3178 struct rfapi_monitor_encap
*m
;
3179 struct rfapi_monitor_encap
*mnext
;
3181 struct agg_node
*referenced_vpn_prefix
;
3184 * Optimized approach: build radix tree on the fly to
3185 * hold list of VPN nodes referenced by the ENCAP monitors
3187 * The nodes in this table correspond to prefixes of VPN routes.
3188 * The "info" pointer of the node points to a chain of
3189 * struct rfapi_monitor_encap, each of which refers to a
3190 * specific VPN node.
3192 struct agg_table
*referenced_vpn_table
;
3194 referenced_vpn_table
= agg_table_init();
3197 * iterate over the set of monitors at this ENCAP node.
3199 #ifdef DEBUG_ENCAP_MONITOR
3200 vnc_zlog_debug_verbose("%s: examining monitors at rn=%p",
3203 for (m
= RFAPI_MONITOR_ENCAP(rn
); m
; m
= m
->next
) {
3204 const struct prefix
*p
;
3207 * For each referenced bpi/route, copy the ENCAP route's
3208 * nexthop to the VPN route's cached UN address field
3210 * the address family of the cached UN address field.
3212 rfapiCopyUnEncap2VPN(info_new
, m
->bpi
);
3213 if (!CHECK_FLAG(m
->bpi
->flags
, BGP_PATH_VALID
)) {
3214 SET_FLAG(m
->bpi
->flags
, BGP_PATH_VALID
);
3215 if (VALID_INTERIOR_TYPE(m
->bpi
->type
))
3216 RFAPI_MONITOR_EXTERIOR(m
->node
)
3217 ->valid_interior_count
++;
3218 vnc_import_bgp_exterior_add_route_interior(
3219 bgp
, import_table
, m
->node
, m
->bpi
);
3223 * Build a list of unique VPN nodes referenced by these
3226 * There could be more than one VPN node here with a
3228 * prefix. Those are currently in an unsorted linear
3232 p
= agg_node_get_prefix(m
->node
);
3233 referenced_vpn_prefix
=
3234 agg_node_get(referenced_vpn_table
, p
);
3235 assert(referenced_vpn_prefix
);
3236 for (mnext
= referenced_vpn_prefix
->info
; mnext
;
3237 mnext
= mnext
->next
) {
3239 if (mnext
->node
== m
->node
)
3245 * already have an entry for this VPN node
3247 agg_unlock_node(referenced_vpn_prefix
);
3250 MTYPE_RFAPI_MONITOR_ENCAP
,
3251 sizeof(struct rfapi_monitor_encap
));
3252 mnext
->node
= m
->node
;
3253 mnext
->next
= referenced_vpn_prefix
->info
;
3254 referenced_vpn_prefix
->info
= mnext
;
3259 * for each VPN node referenced in the ENCAP monitors:
3261 for (referenced_vpn_prefix
=
3262 agg_route_top(referenced_vpn_table
);
3263 referenced_vpn_prefix
;
3264 referenced_vpn_prefix
=
3265 agg_route_next(referenced_vpn_prefix
)) {
3267 while ((m
= referenced_vpn_prefix
->info
)) {
3271 rfapiMonitorMoveLonger(m
->node
);
3272 for (n
= m
->node
; n
; n
= agg_node_parent(n
)) {
3273 // rfapiDoRouteCallback(import_table, n,
3276 rfapiMonitorItNodeChanged(import_table
, m
->node
,
3279 referenced_vpn_prefix
->info
= m
->next
;
3280 agg_unlock_node(referenced_vpn_prefix
);
3281 XFREE(MTYPE_RFAPI_MONITOR_ENCAP
, m
);
3284 agg_table_finish(referenced_vpn_table
);
3287 RFAPI_CHECK_REFCOUNT(rn
, SAFI_ENCAP
, 0);
3290 static void rfapiExpireVpnNow(struct rfapi_import_table
*it
,
3291 struct agg_node
*rn
, struct bgp_path_info
*bpi
,
3294 struct rfapi_withdraw
*wcb
;
3298 * pretend we're an expiring timer
3300 wcb
= XCALLOC(MTYPE_RFAPI_WITHDRAW
, sizeof(struct rfapi_withdraw
));
3303 wcb
->import_table
= it
;
3304 wcb
->lockoffset
= lockoffset
;
3305 memset(&t
, 0, sizeof(t
));
3307 rfapiWithdrawTimerVPN(&t
); /* frees wcb */
3312 * import a bgp_path_info if its route target list intersects with the
3313 * import table's route target list
3315 void rfapiBgpInfoFilteredImportVPN(
3316 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3317 void *rfd
, /* set for looped back routes */
3318 const struct prefix
*p
,
3319 const struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3320 afi_t afi
, struct prefix_rd
*prd
,
3321 struct attr
*attr
, /* part of bgp_path_info */
3322 uint8_t type
, /* part of bgp_path_info */
3323 uint8_t sub_type
, /* part of bgp_path_info */
3324 uint32_t *label
) /* part of bgp_path_info */
3326 struct agg_table
*rt
= NULL
;
3327 struct agg_node
*rn
;
3329 struct bgp_path_info
*info_new
;
3330 struct bgp_path_info
*bpi
;
3331 struct bgp_path_info
*next
;
3333 struct prefix vn_prefix
;
3334 struct prefix un_prefix
;
3335 int un_prefix_valid
= 0;
3336 struct agg_node
*ern
;
3338 int original_had_routes
= 0;
3339 struct prefix original_nexthop
;
3340 const char *action_str
= NULL
;
3344 bgp
= bgp_get_default(); /* assume 1 instance for now */
3347 case FIF_ACTION_UPDATE
:
3348 action_str
= "update";
3350 case FIF_ACTION_WITHDRAW
:
3351 action_str
= "withdraw";
3353 case FIF_ACTION_KILL
:
3354 action_str
= "kill";
3361 if (import_table
== bgp
->rfapi
->it_ce
)
3364 vnc_zlog_debug_verbose("%s: entry: %s%s: prefix %s/%d: it %p, afi %s",
3365 __func__
, (is_it_ce
? "CE-IT " : ""), action_str
,
3366 rfapi_ntop(p
->family
, &p
->u
.prefix
, buf
, BUFSIZ
),
3367 p
->prefixlen
, import_table
, afi2str(afi
));
3372 * Compare rt lists. If no intersection, don't import this route
3373 * On a withdraw, peer and RD are sufficient to determine if
3376 if (action
== FIF_ACTION_UPDATE
) {
3377 if (!attr
|| !bgp_attr_get_ecommunity(attr
)) {
3379 vnc_zlog_debug_verbose(
3380 "%s: attr, extra, or ecommunity missing, not importing",
3384 if ((import_table
!= bgp
->rfapi
->it_ce
) &&
3385 !rfapiEcommunitiesIntersect(
3386 import_table
->rt_import_list
,
3387 bgp_attr_get_ecommunity(attr
))) {
3389 vnc_zlog_debug_verbose(
3390 "%s: it=%p: no ecommunity intersection",
3391 __func__
, import_table
);
3395 memset(&vn_prefix
, 0,
3396 sizeof(vn_prefix
)); /* keep valgrind happy */
3397 if (rfapiGetNexthop(attr
, &vn_prefix
)) {
3398 /* missing nexthop address would be a bad, bad thing */
3399 vnc_zlog_debug_verbose("%s: missing nexthop", __func__
);
3405 * Figure out which radix tree the route would go into
3411 rt
= import_table
->imported_vpn
[afi
];
3416 flog_err(EC_LIB_DEVELOPMENT
, "%s: bad afi %d", __func__
, afi
);
3421 memset(&original_nexthop
, 0, sizeof(original_nexthop
));
3424 * agg_node_lookup returns a node only if there is at least
3425 * one route attached.
3427 rn
= agg_node_lookup(rt
, p
);
3429 vnc_zlog_debug_verbose("%s: rn=%p", __func__
, rn
);
3433 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
3434 agg_unlock_node(rn
); /* undo lock in agg_node_lookup */
3437 original_had_routes
= 1;
3439 if (VNC_DEBUG(VERBOSE
)) {
3440 vnc_zlog_debug_verbose("%s: showing IT node on entry",
3442 rfapiShowItNode(NULL
, rn
); /* debug */
3446 * Look for same route (will have same RD and peer)
3448 bpi
= rfapiItBiIndexSearch(rn
, prd
, peer
, aux_prefix
);
3453 * This was an old test when we iterated over the
3454 * BPIs linearly. Since we're now looking up with
3455 * RD and peer, comparing types should not be
3456 * needed. Changed to assertion.
3458 * Compare types. Doing so prevents a RFP-originated
3459 * route from matching an imported route, for example.
3461 if (VNC_DEBUG(VERBOSE
) && bpi
->type
!= type
)
3462 /* should be handled by RDs, but warn for now */
3463 zlog_warn("%s: type mismatch! (bpi=%d, arg=%d)",
3464 __func__
, bpi
->type
, type
);
3466 vnc_zlog_debug_verbose("%s: found matching bpi",
3470 * In the special CE table, withdrawals occur without
3473 if (import_table
== bgp
->rfapi
->it_ce
) {
3474 vnc_direct_bgp_del_route_ce(bgp
, rn
, bpi
);
3475 if (action
== FIF_ACTION_WITHDRAW
)
3476 action
= FIF_ACTION_KILL
;
3479 if (action
== FIF_ACTION_WITHDRAW
) {
3481 int washolddown
= CHECK_FLAG(bpi
->flags
,
3484 vnc_zlog_debug_verbose(
3485 "%s: withdrawing at prefix %pRN%s",
3488 ? " (already being withdrawn)"
3493 rfapiBiStartWithdrawTimer(
3494 import_table
, rn
, bpi
, afi
,
3496 rfapiWithdrawTimerVPN
);
3498 RFAPI_UPDATE_ITABLE_COUNT(
3499 bpi
, import_table
, afi
, -1);
3500 import_table
->holddown_count
[afi
] += 1;
3504 vnc_zlog_debug_verbose(
3505 "%s: %s at prefix %pRN", __func__
,
3506 ((action
== FIF_ACTION_KILL
)
3512 * If this route is waiting to be deleted
3514 * a previous withdraw, we must cancel its
3517 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)
3518 && bpi
->extra
->vnc
.import
.timer
) {
3519 struct rfapi_withdraw
*wcb
= THREAD_ARG(
3520 bpi
->extra
->vnc
.import
.timer
);
3522 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3524 bpi
->extra
->vnc
.import
.timer
);
3526 import_table
->holddown_count
[afi
] -= 1;
3527 RFAPI_UPDATE_ITABLE_COUNT(
3528 bpi
, import_table
, afi
, 1);
3531 * decrement remote count (if route is remote)
3533 * we are going to remove it below
3535 RFAPI_UPDATE_ITABLE_COUNT(bpi
, import_table
,
3537 if (action
== FIF_ACTION_UPDATE
) {
3541 * make copy of original nexthop so we
3542 * can see if it changed
3544 rfapiGetNexthop(bpi
->attr
,
3548 * remove bpi without doing any export
3551 if (CHECK_FLAG(bpi
->flags
,
3553 && VALID_INTERIOR_TYPE(bpi
->type
))
3554 RFAPI_MONITOR_EXTERIOR(rn
)
3555 ->valid_interior_count
--;
3556 rfapiItBiIndexDel(rn
, bpi
);
3557 rfapiBgpInfoDetach(rn
, bpi
);
3558 rfapiMonitorEncapDelete(bpi
);
3559 vnc_import_bgp_exterior_del_route_interior(
3560 bgp
, import_table
, rn
, bpi
);
3561 rfapiBgpInfoFree(bpi
);
3565 * remove bpi and do export processing
3567 import_table
->holddown_count
[afi
] += 1;
3568 rfapiExpireVpnNow(import_table
, rn
, bpi
,
3576 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, replacing
? 1 : 0);
3578 if (action
== FIF_ACTION_WITHDRAW
|| action
== FIF_ACTION_KILL
) {
3584 rfapiBgpInfoCreate(attr
, peer
, rfd
, prd
, type
, sub_type
, label
);
3587 * lookup un address in encap table
3589 ern
= agg_node_match(import_table
->imported_encap
[afi
], &vn_prefix
);
3591 rfapiCopyUnEncap2VPN(ern
->info
, info_new
);
3592 agg_unlock_node(ern
); /* undo lock in route_note_match */
3594 /* Not a big deal, just means VPN route got here first */
3595 vnc_zlog_debug_verbose("%s: no encap route for vn addr %pFX",
3596 __func__
, &vn_prefix
);
3597 info_new
->extra
->vnc
.import
.un_family
= AF_UNSPEC
;
3605 * No need to increment reference count, so only "get"
3606 * if the node is not there already
3608 rn
= agg_node_get(rt
, p
);
3612 * For ethernet routes, if there is an accompanying IP address,
3613 * save it in the bpi
3615 if ((AFI_L2VPN
== afi
) && aux_prefix
) {
3617 vnc_zlog_debug_verbose("%s: setting BPI's aux_prefix",
3619 info_new
->extra
->vnc
.import
.aux_prefix
= *aux_prefix
;
3622 vnc_zlog_debug_verbose("%s: inserting bpi %p at prefix %pRN #%d",
3623 __func__
, info_new
, rn
,
3624 agg_node_get_lock_count(rn
));
3626 rfapiBgpInfoAttachSorted(rn
, info_new
, afi
, SAFI_MPLS_VPN
);
3627 rfapiItBiIndexAdd(rn
, info_new
);
3628 if (!rfapiGetUnAddrOfVpnBi(info_new
, NULL
)) {
3629 if (VALID_INTERIOR_TYPE(info_new
->type
))
3630 RFAPI_MONITOR_EXTERIOR(rn
)->valid_interior_count
++;
3631 SET_FLAG(info_new
->flags
, BGP_PATH_VALID
);
3633 RFAPI_UPDATE_ITABLE_COUNT(info_new
, import_table
, afi
, 1);
3634 vnc_import_bgp_exterior_add_route_interior(bgp
, import_table
, rn
,
3637 if (import_table
== bgp
->rfapi
->it_ce
)
3638 vnc_direct_bgp_add_route_ce(bgp
, rn
, info_new
);
3640 if (VNC_DEBUG(VERBOSE
)) {
3641 vnc_zlog_debug_verbose("%s: showing IT node", __func__
);
3642 rfapiShowItNode(NULL
, rn
); /* debug */
3645 rfapiMonitorEncapAdd(import_table
, &vn_prefix
, rn
, info_new
);
3647 if (!rfapiGetUnAddrOfVpnBi(info_new
, &un_prefix
)) {
3650 * if we have a valid UN address (either via Encap route
3651 * or via tunnel attribute), then we should attempt
3652 * to move any monitors at less-specific nodes to this node
3654 rfapiMonitorMoveLonger(rn
);
3656 un_prefix_valid
= 1;
3660 * 101129 Enhancement: if we add a route (implication: it is not
3661 * in holddown), delete all other routes from this nve at this
3662 * node that are in holddown, regardless of peer.
3664 * Reasons it's OK to do that:
3666 * - if the holddown route being deleted originally came from BGP VPN,
3667 * it is already gone from BGP (implication of holddown), so there
3668 * won't be any added inconsistency with the BGP RIB.
3670 * - once a fresh route is added at a prefix, any routes in holddown
3671 * at that prefix will not show up in RFP responses, so deleting
3672 * the holddown routes won't affect the contents of responses.
3674 * - lifetimes are supposed to be consistent, so there should not
3675 * be a case where the fresh route has a shorter lifetime than
3676 * the holddown route, so we don't expect the fresh route to
3677 * disappear and complete its holddown time before the existing
3678 * holddown routes time out. Therefore, we won't have a situation
3679 * where we expect the existing holddown routes to be hidden and
3680 * then to reappear sometime later (as holddown routes) in a
3683 * Among other things, this would enable us to skirt the problem
3684 * of local holddown routes that refer to NVE descriptors that
3685 * have already been closed (if the same NVE triggers a subsequent
3686 * rfapi_open(), the new peer is different and doesn't match the
3687 * peer of the holddown route, so the stale holddown route still
3688 * hangs around until it times out instead of just being replaced
3689 * by the fresh route).
3692 * We know that the new bpi will have been inserted before any routes
3693 * in holddown, so we can skip any that came before it
3695 for (bpi
= info_new
->next
; bpi
; bpi
= next
) {
3697 struct prefix pfx_vn
;
3698 struct prefix pfx_un
;
3700 int remote_peer_match
= 0;
3707 if (!CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
))
3711 * Must match VN address (nexthop of VPN route)
3713 if (rfapiGetNexthop(bpi
->attr
, &pfx_vn
))
3715 if (!prefix_same(&pfx_vn
, &vn_prefix
))
3718 if (un_prefix_valid
&& /* new route UN addr */
3719 !rfapiGetUnAddrOfVpnBi(bpi
, &pfx_un
)
3720 && /* old route UN addr */
3721 prefix_same(&pfx_un
, &un_prefix
)) { /* compare */
3724 if (!RFAPI_LOCAL_BI(bpi
) && !RFAPI_LOCAL_BI(info_new
)
3725 && sockunion_same(&bpi
->peer
->su
, &info_new
->peer
->su
)) {
3726 /* old & new are both remote, same peer */
3727 remote_peer_match
= 1;
3730 if (!un_match
&& !remote_peer_match
)
3733 vnc_zlog_debug_verbose(
3734 "%s: removing holddown bpi matching NVE of new route",
3736 if (bpi
->extra
->vnc
.import
.timer
) {
3737 struct rfapi_withdraw
*wcb
=
3738 THREAD_ARG(bpi
->extra
->vnc
.import
.timer
);
3740 XFREE(MTYPE_RFAPI_WITHDRAW
, wcb
);
3741 THREAD_OFF(bpi
->extra
->vnc
.import
.timer
);
3743 rfapiExpireVpnNow(import_table
, rn
, bpi
, 0);
3746 if (!original_had_routes
) {
3748 * We went from 0 usable routes to 1 usable route. Perform the
3749 * "Adding a Route" export process.
3751 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3752 vnc_zebra_add_prefix(bgp
, import_table
, rn
);
3755 * Check for nexthop change event
3756 * Note: the prefix_same() test below detects two situations:
3757 * 1. route is replaced, new route has different nexthop
3758 * 2. new route is added (original_nexthop is 0)
3760 struct prefix new_nexthop
;
3762 rfapiGetNexthop(attr
, &new_nexthop
);
3763 if (!prefix_same(&original_nexthop
, &new_nexthop
)) {
3765 * nexthop change event
3766 * vnc_direct_bgp_add_prefix() will recompute VN addr
3769 vnc_direct_bgp_add_prefix(bgp
, import_table
, rn
);
3773 if (!(bgp
->rfapi_cfg
->flags
& BGP_VNC_CONFIG_CALLBACK_DISABLE
)) {
3774 for (n
= rn
; n
; n
= agg_node_parent(n
)) {
3775 // rfapiDoRouteCallback(import_table, n, NULL);
3777 rfapiMonitorItNodeChanged(import_table
, rn
, NULL
);
3779 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 0);
3783 static void rfapiBgpInfoFilteredImportBadSafi(
3784 struct rfapi_import_table
*import_table
, int action
, struct peer
*peer
,
3785 void *rfd
, /* set for looped back routes */
3786 const struct prefix
*p
,
3787 const struct prefix
*aux_prefix
, /* AFI_L2VPN: optional IP */
3788 afi_t afi
, struct prefix_rd
*prd
,
3789 struct attr
*attr
, /* part of bgp_path_info */
3790 uint8_t type
, /* part of bgp_path_info */
3791 uint8_t sub_type
, /* part of bgp_path_info */
3792 uint32_t *label
) /* part of bgp_path_info */
3794 vnc_zlog_debug_verbose("%s: Error, bad safi", __func__
);
3797 static rfapi_bi_filtered_import_f
*
3798 rfapiBgpInfoFilteredImportFunction(safi_t safi
)
3802 return rfapiBgpInfoFilteredImportVPN
;
3805 return rfapiBgpInfoFilteredImportEncap
;
3809 case SAFI_MULTICAST
:
3811 case SAFI_LABELED_UNICAST
:
3815 flog_err(EC_LIB_DEVELOPMENT
, "%s: bad safi %d", __func__
, safi
);
3816 return rfapiBgpInfoFilteredImportBadSafi
;
3819 assert(!"Reached end of function when we were not expecting to");
3822 void rfapiProcessUpdate(struct peer
*peer
,
3823 void *rfd
, /* set when looped from RFP/RFAPI */
3824 const struct prefix
*p
, struct prefix_rd
*prd
,
3825 struct attr
*attr
, afi_t afi
, safi_t safi
, uint8_t type
,
3826 uint8_t sub_type
, uint32_t *label
)
3830 struct rfapi_import_table
*it
;
3831 int has_ip_route
= 1;
3834 bgp
= bgp_get_default(); /* assume 1 instance for now */
3841 * look at high-order byte of RD. FF means MAC
3842 * address is present (VNC L2VPN)
3844 if ((safi
== SAFI_MPLS_VPN
)
3845 && (decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
)) {
3846 struct prefix pfx_mac_buf
;
3847 struct prefix pfx_nexthop_buf
;
3851 * Set flag if prefix and nexthop are the same - don't
3852 * add the route to normal IP-based import tables
3854 if (!rfapiGetNexthop(attr
, &pfx_nexthop_buf
)) {
3855 if (!prefix_cmp(&pfx_nexthop_buf
, p
)) {
3860 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
3861 pfx_mac_buf
.family
= AF_ETHERNET
;
3862 pfx_mac_buf
.prefixlen
= 48;
3863 memcpy(&pfx_mac_buf
.u
.prefix_eth
.octet
, prd
->val
+ 2, 6);
3866 * Find rt containing LNI (Logical Network ID), which
3867 * _should_ always be present when mac address is present
3869 rc
= rfapiEcommunityGetLNI(bgp_attr_get_ecommunity(attr
), &lni
);
3871 vnc_zlog_debug_verbose(
3872 "%s: rfapiEcommunityGetLNI returned %d, lni=%d, attr=%p",
3873 __func__
, rc
, lni
, attr
);
3875 it
= rfapiMacImportTableGet(bgp
, lni
);
3877 rfapiBgpInfoFilteredImportVPN(
3878 it
, FIF_ACTION_UPDATE
, peer
, rfd
,
3879 &pfx_mac_buf
, /* prefix */
3880 p
, /* aux prefix: IP addr */
3881 AFI_L2VPN
, prd
, attr
, type
, sub_type
, label
);
3889 * Iterate over all import tables; do a filtered import
3890 * for the afi/safi combination
3892 for (it
= h
->imports
; it
; it
= it
->next
) {
3893 (*rfapiBgpInfoFilteredImportFunction(safi
))(
3894 it
, FIF_ACTION_UPDATE
, peer
, rfd
, p
, /* prefix */
3895 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3898 if (safi
== SAFI_MPLS_VPN
) {
3899 vnc_direct_bgp_rh_add_route(bgp
, afi
, p
, peer
, attr
);
3900 rfapiBgpInfoFilteredImportVPN(
3901 bgp
->rfapi
->it_ce
, FIF_ACTION_UPDATE
, peer
, rfd
,
3903 NULL
, afi
, prd
, attr
, type
, sub_type
, label
);
3908 void rfapiProcessWithdraw(struct peer
*peer
, void *rfd
, const struct prefix
*p
,
3909 struct prefix_rd
*prd
, struct attr
*attr
, afi_t afi
,
3910 safi_t safi
, uint8_t type
, int kill
)
3914 struct rfapi_import_table
*it
;
3916 bgp
= bgp_get_default(); /* assume 1 instance for now */
3923 * look at high-order byte of RD. FF means MAC
3924 * address is present (VNC L2VPN)
3926 if (h
->import_mac
!= NULL
&& safi
== SAFI_MPLS_VPN
3927 && decode_rd_type(prd
->val
) == RD_TYPE_VNC_ETH
) {
3928 struct prefix pfx_mac_buf
;
3929 void *cursor
= NULL
;
3932 memset(&pfx_mac_buf
, 0, sizeof(pfx_mac_buf
));
3933 pfx_mac_buf
.family
= AF_ETHERNET
;
3934 pfx_mac_buf
.prefixlen
= 48;
3935 memcpy(&pfx_mac_buf
.u
.prefix_eth
, prd
->val
+ 2, 6);
3938 * withdraw does not contain attrs, so we don't have
3939 * access to the route's LNI, which would ordinarily
3940 * select the specific mac-based import table. Instead,
3941 * we must iterate over all mac-based tables and rely
3942 * on the RD to match.
3944 * If this approach is too slow, add an index where
3945 * key is {RD, peer} and value is the import table
3947 for (rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
3949 rc
== 0; rc
= skiplist_next(h
->import_mac
, NULL
,
3950 (void **)&it
, &cursor
)) {
3952 #ifdef DEBUG_L2_EXTRA
3953 vnc_zlog_debug_verbose(
3954 "%s: calling rfapiBgpInfoFilteredImportVPN(it=%p, afi=AFI_L2VPN)",
3958 rfapiBgpInfoFilteredImportVPN(
3960 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
3961 peer
, rfd
, &pfx_mac_buf
, /* prefix */
3962 p
, /* aux_prefix: IP */
3963 AFI_L2VPN
, prd
, attr
, type
, 0,
3964 NULL
); /* sub_type & label unused for withdraw
3970 * XXX For the case where the withdraw involves an L2
3971 * route with no IP information, we rely on the lack
3972 * of RT-list intersection to filter out the withdraw
3973 * from the IP-based import tables below
3977 * Iterate over all import tables; do a filtered import
3978 * for the afi/safi combination
3981 for (it
= h
->imports
; it
; it
= it
->next
) {
3982 (*rfapiBgpInfoFilteredImportFunction(safi
))(
3983 it
, (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
),
3984 peer
, rfd
, p
, /* prefix */
3985 NULL
, afi
, prd
, attr
, type
, 0,
3986 NULL
); /* sub_type & label unused for withdraw */
3989 /* TBD the deletion should happen after the lifetime expires */
3990 if (safi
== SAFI_MPLS_VPN
)
3991 vnc_direct_bgp_rh_del_route(bgp
, afi
, p
, peer
);
3993 if (safi
== SAFI_MPLS_VPN
) {
3994 rfapiBgpInfoFilteredImportVPN(
3996 (kill
? FIF_ACTION_KILL
: FIF_ACTION_WITHDRAW
), peer
,
3997 rfd
, p
, /* prefix */
3998 NULL
, afi
, prd
, attr
, type
, 0,
3999 NULL
); /* sub_type & label unused for withdraw */
4004 * TBD optimized withdraw timer algorithm for case of many
4005 * routes expiring at the same time due to peer drop.
4008 * 1. Visit all BPIs in all ENCAP import tables.
4010 * a. If a bpi's peer is the failed peer, remove the bpi.
4011 * b. If the removed ENCAP bpi was first in the list of
4012 * BPIs at this ENCAP node, loop over all monitors
4015 * (1) for each ENCAP monitor, loop over all its
4016 * VPN node monitors and set their RFAPI_MON_FLAG_NEEDCALLBACK
4019 * 2. Visit all BPIs in all VPN import tables.
4020 * a. If a bpi's peer is the failed peer, remove the bpi.
4021 * b. loop over all the VPN node monitors and set their
4022 * RFAPI_MON_FLAG_NEEDCALLBACK flags
4023 * c. If there are no BPIs left at this VPN node,
4028 /* surprise, this gets called from peer_delete(), from rfapi_close() */
4029 static void rfapiProcessPeerDownRt(struct peer
*peer
,
4030 struct rfapi_import_table
*import_table
,
4031 afi_t afi
, safi_t safi
)
4033 struct agg_node
*rn
;
4034 struct bgp_path_info
*bpi
;
4035 struct agg_table
*rt
= NULL
;
4036 void (*timer_service_func
)(struct thread
*) = NULL
;
4038 assert(afi
== AFI_IP
|| afi
== AFI_IP6
);
4044 rt
= import_table
->imported_vpn
[afi
];
4045 timer_service_func
= rfapiWithdrawTimerVPN
;
4048 rt
= import_table
->imported_encap
[afi
];
4049 timer_service_func
= rfapiWithdrawTimerEncap
;
4053 case SAFI_MULTICAST
:
4055 case SAFI_LABELED_UNICAST
:
4058 /* Suppress uninitialized variable warning */
4060 timer_service_func
= NULL
;
4064 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
4065 for (bpi
= rn
->info
; bpi
; bpi
= bpi
->next
) {
4066 if (bpi
->peer
== peer
) {
4068 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
4069 /* already in holddown, skip */
4073 if (safi
== SAFI_MPLS_VPN
) {
4074 RFAPI_UPDATE_ITABLE_COUNT(
4075 bpi
, import_table
, afi
, -1);
4076 import_table
->holddown_count
[afi
] += 1;
4078 rfapiBiStartWithdrawTimer(import_table
, rn
, bpi
,
4080 timer_service_func
);
4088 * This gets called when a peer connection drops. We have to remove
4089 * all the routes from this peer.
4091 * Current approach is crude. TBD Optimize by setting fewer timers and
4092 * grouping withdrawn routes so we can generate callbacks more
4095 void rfapiProcessPeerDown(struct peer
*peer
)
4099 struct rfapi_import_table
*it
;
4102 * If this peer is a "dummy" peer structure atached to a RFAPI
4103 * nve_descriptor, we don't need to walk the import tables
4104 * because the routes are already withdrawn by rfapi_close()
4106 if (CHECK_FLAG(peer
->flags
, PEER_FLAG_IS_RFAPI_HD
))
4110 * 1. Visit all BPIs in all ENCAP import tables.
4111 * Start withdraw timer on the BPIs that match peer.
4113 * 2. Visit All BPIs in all VPN import tables.
4114 * Start withdraw timer on the BPIs that match peer.
4117 bgp
= bgp_get_default(); /* assume 1 instance for now */
4124 for (it
= h
->imports
; it
; it
= it
->next
) {
4125 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_ENCAP
);
4126 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_ENCAP
);
4127 rfapiProcessPeerDownRt(peer
, it
, AFI_IP
, SAFI_MPLS_VPN
);
4128 rfapiProcessPeerDownRt(peer
, it
, AFI_IP6
, SAFI_MPLS_VPN
);
4132 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4133 rfapiProcessPeerDownRt(peer
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4138 * Import an entire RIB (for an afi/safi) to an import table RIB,
4139 * filtered according to the import table's RT list
4141 * TBD: does this function need additions to match rfapiProcessUpdate()
4142 * for, e.g., L2 handling?
4144 static void rfapiBgpTableFilteredImport(struct bgp
*bgp
,
4145 struct rfapi_import_table
*it
,
4146 afi_t afi
, safi_t safi
)
4148 struct bgp_dest
*dest1
;
4149 struct bgp_dest
*dest2
;
4151 /* Only these SAFIs have 2-level RIBS */
4152 assert(safi
== SAFI_MPLS_VPN
|| safi
== SAFI_ENCAP
);
4155 * Now visit all the rd nodes and the nodes of all the
4156 * route tables attached to them, and import the routes
4157 * if they have matching route targets
4159 for (dest1
= bgp_table_top(bgp
->rib
[afi
][safi
]); dest1
;
4160 dest1
= bgp_route_next(dest1
)) {
4162 if (bgp_dest_has_bgp_path_info_data(dest1
)) {
4164 for (dest2
= bgp_table_top(
4165 bgp_dest_get_bgp_table_info(dest1
));
4166 dest2
; dest2
= bgp_route_next(dest2
)) {
4168 struct bgp_path_info
*bpi
;
4170 for (bpi
= bgp_dest_get_bgp_path_info(dest2
);
4171 bpi
; bpi
= bpi
->next
) {
4174 if (CHECK_FLAG(bpi
->flags
,
4179 label
= decode_label(
4180 &bpi
->extra
->label
[0]);
4181 (*rfapiBgpInfoFilteredImportFunction(
4183 it
, /* which import table */
4184 FIF_ACTION_UPDATE
, bpi
->peer
,
4186 bgp_dest_get_prefix(dest2
),
4188 (struct prefix_rd
*)
4189 bgp_dest_get_prefix(
4191 bpi
->attr
, bpi
->type
,
4192 bpi
->sub_type
, &label
);
4200 /* per-bgp-instance rfapi data */
4201 struct rfapi
*bgp_rfapi_new(struct bgp
*bgp
)
4205 struct rfapi_rfp_cfg
*cfg
= NULL
;
4206 struct rfapi_rfp_cb_methods
*cbm
= NULL
;
4208 assert(bgp
->rfapi_cfg
== NULL
);
4210 h
= XCALLOC(MTYPE_RFAPI
, sizeof(struct rfapi
));
4212 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4213 h
->un
[afi
] = agg_table_init();
4217 * initialize the ce import table
4219 h
->it_ce
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4220 sizeof(struct rfapi_import_table
));
4221 h
->it_ce
->imported_vpn
[AFI_IP
] = agg_table_init();
4222 h
->it_ce
->imported_vpn
[AFI_IP6
] = agg_table_init();
4223 h
->it_ce
->imported_encap
[AFI_IP
] = agg_table_init();
4224 h
->it_ce
->imported_encap
[AFI_IP6
] = agg_table_init();
4225 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP
, SAFI_MPLS_VPN
);
4226 rfapiBgpTableFilteredImport(bgp
, h
->it_ce
, AFI_IP6
, SAFI_MPLS_VPN
);
4229 * Set up work queue for deferred rfapi_close operations
4231 h
->deferred_close_q
=
4232 work_queue_new(bm
->master
, "rfapi deferred close");
4233 h
->deferred_close_q
->spec
.workfunc
= rfapi_deferred_close_workfunc
;
4234 h
->deferred_close_q
->spec
.data
= h
;
4236 h
->rfp
= rfp_start(bm
->master
, &cfg
, &cbm
);
4237 bgp
->rfapi_cfg
= bgp_rfapi_cfg_new(cfg
);
4239 h
->rfp_methods
= *cbm
;
4244 void bgp_rfapi_destroy(struct bgp
*bgp
, struct rfapi
*h
)
4248 if (bgp
== NULL
|| h
== NULL
)
4251 if (h
->resolve_nve_nexthop
) {
4252 skiplist_free(h
->resolve_nve_nexthop
);
4253 h
->resolve_nve_nexthop
= NULL
;
4256 agg_table_finish(h
->it_ce
->imported_vpn
[AFI_IP
]);
4257 agg_table_finish(h
->it_ce
->imported_vpn
[AFI_IP6
]);
4258 agg_table_finish(h
->it_ce
->imported_encap
[AFI_IP
]);
4259 agg_table_finish(h
->it_ce
->imported_encap
[AFI_IP6
]);
4261 if (h
->import_mac
) {
4262 struct rfapi_import_table
*it
;
4267 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4269 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4272 rfapiImportTableFlush(it
);
4273 XFREE(MTYPE_RFAPI_IMPORTTABLE
, it
);
4275 skiplist_free(h
->import_mac
);
4276 h
->import_mac
= NULL
;
4279 work_queue_free_and_null(&h
->deferred_close_q
);
4284 for (afi
= AFI_IP
; afi
< AFI_MAX
; afi
++) {
4285 agg_table_finish(h
->un
[afi
]);
4288 XFREE(MTYPE_RFAPI_IMPORTTABLE
, h
->it_ce
);
4289 XFREE(MTYPE_RFAPI
, h
);
4292 struct rfapi_import_table
*
4293 rfapiImportTableRefAdd(struct bgp
*bgp
, struct ecommunity
*rt_import_list
,
4294 struct rfapi_nve_group_cfg
*rfg
)
4297 struct rfapi_import_table
*it
;
4303 for (it
= h
->imports
; it
; it
= it
->next
) {
4304 if (ecommunity_cmp(it
->rt_import_list
, rt_import_list
))
4308 vnc_zlog_debug_verbose("%s: matched it=%p", __func__
, it
);
4311 it
= XCALLOC(MTYPE_RFAPI_IMPORTTABLE
,
4312 sizeof(struct rfapi_import_table
));
4313 it
->next
= h
->imports
;
4316 it
->rt_import_list
= ecommunity_dup(rt_import_list
);
4318 it
->monitor_exterior_orphans
=
4319 skiplist_new(0, NULL
, prefix_free_lists
);
4322 * fill import route tables from RIBs
4324 * Potential area for optimization. If this occurs when
4325 * tables are large (e.g., the operator adds a nve group
4326 * with a new RT list to a running system), it could take
4330 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4332 it
->imported_vpn
[afi
] = agg_table_init();
4333 it
->imported_encap
[afi
] = agg_table_init();
4335 rfapiBgpTableFilteredImport(bgp
, it
, afi
,
4337 rfapiBgpTableFilteredImport(bgp
, it
, afi
, SAFI_ENCAP
);
4339 vnc_import_bgp_exterior_redist_enable_it(bgp
, afi
, it
);
4349 * skiplist element free function
4351 static void delete_rem_pfx_na_free(void *na
)
4353 uint32_t *pCounter
= ((struct rfapi_nve_addr
*)na
)->info
;
4356 XFREE(MTYPE_RFAPI_NVE_ADDR
, na
);
4360 * Common deleter for IP and MAC import tables
4362 static void rfapiDeleteRemotePrefixesIt(
4363 struct bgp
*bgp
, struct rfapi_import_table
*it
, struct prefix
*un
,
4364 struct prefix
*vn
, struct prefix
*p
, int delete_active
,
4365 int delete_holddown
, uint32_t *pARcount
, uint32_t *pAHcount
,
4366 uint32_t *pHRcount
, uint32_t *pHHcount
,
4367 struct skiplist
*uniq_active_nves
, struct skiplist
*uniq_holddown_nves
)
4371 #ifdef DEBUG_L2_EXTRA
4373 char buf_pfx
[PREFIX_STRLEN
];
4376 prefix2str(p
, buf_pfx
, sizeof(buf_pfx
));
4382 vnc_zlog_debug_verbose(
4383 "%s: entry, p=%s, delete_active=%d, delete_holddown=%d",
4384 __func__
, buf_pfx
, delete_active
, delete_holddown
);
4388 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4390 struct agg_table
*rt
;
4391 struct agg_node
*rn
;
4393 if (p
&& (family2afi(p
->family
) != afi
)) {
4397 rt
= it
->imported_vpn
[afi
];
4401 vnc_zlog_debug_verbose("%s: scanning rt for afi=%d", __func__
,
4404 for (rn
= agg_route_top(rt
); rn
; rn
= agg_route_next(rn
)) {
4405 struct bgp_path_info
*bpi
;
4406 struct bgp_path_info
*next
;
4407 const struct prefix
*rn_p
= agg_node_get_prefix(rn
);
4409 if (p
&& VNC_DEBUG(IMPORT_DEL_REMOTE
))
4410 vnc_zlog_debug_any("%s: want %pFX, have %pRN",
4413 if (p
&& prefix_cmp(p
, rn_p
))
4416 vnc_zlog_debug_verbose("%s: rn pfx=%pRN", __func__
, rn
);
4418 /* TBD is this valid for afi == AFI_L2VPN? */
4419 RFAPI_CHECK_REFCOUNT(rn
, SAFI_MPLS_VPN
, 1);
4421 for (bpi
= rn
->info
; bpi
; bpi
= next
) {
4430 vnc_zlog_debug_verbose("%s: examining bpi %p",
4433 if (!rfapiGetNexthop(bpi
->attr
, &qpt
))
4438 || !prefix_match(vn
, &qpt
)) {
4439 #ifdef DEBUG_L2_EXTRA
4440 vnc_zlog_debug_verbose(
4441 "%s: continue at vn && !qpt_valid || !prefix_match(vn, &qpt)",
4448 if (!rfapiGetUnAddrOfVpnBi(bpi
, &qct
))
4453 || !prefix_match(un
, &qct
)) {
4454 #ifdef DEBUG_L2_EXTRA
4455 vnc_zlog_debug_verbose(
4456 "%s: continue at un && !qct_valid || !prefix_match(un, &qct)",
4468 * If this route is waiting to be deleted
4470 * a previous withdraw, we must cancel its
4473 if (CHECK_FLAG(bpi
->flags
, BGP_PATH_REMOVED
)) {
4474 if (!delete_holddown
)
4476 if (bpi
->extra
->vnc
.import
.timer
) {
4477 struct rfapi_withdraw
*wcb
=
4484 ->holddown_count
[afi
] -=
4486 RFAPI_UPDATE_ITABLE_COUNT(
4487 bpi
, wcb
->import_table
,
4489 XFREE(MTYPE_RFAPI_WITHDRAW
,
4492 bpi
->extra
->vnc
.import
4501 vnc_zlog_debug_verbose(
4502 "%s: deleting bpi %p (qct_valid=%d, qpt_valid=%d, delete_holddown=%d, delete_active=%d)",
4503 __func__
, bpi
, qct_valid
, qpt_valid
,
4504 delete_holddown
, delete_active
);
4510 if (qct_valid
&& qpt_valid
) {
4512 struct rfapi_nve_addr na
;
4513 struct rfapi_nve_addr
*nap
;
4515 memset(&na
, 0, sizeof(na
));
4516 assert(!rfapiQprefix2Raddr(&qct
,
4518 assert(!rfapiQprefix2Raddr(&qpt
,
4521 if (skiplist_search(
4524 : uniq_holddown_nves
),
4525 &na
, (void **)&nap
)) {
4529 MTYPE_RFAPI_NVE_ADDR
,
4533 nap
->info
= is_active
4539 : uniq_holddown_nves
),
4542 rfapiNveAddr2Str(nap
, line
,
4547 vnc_direct_bgp_rh_del_route(bgp
, afi
, rn_p
,
4550 RFAPI_UPDATE_ITABLE_COUNT(bpi
, it
, afi
, -1);
4551 it
->holddown_count
[afi
] += 1;
4552 rfapiExpireVpnNow(it
, rn
, bpi
, 1);
4554 vnc_zlog_debug_verbose(
4555 "%s: incrementing count (is_active=%d)",
4556 __func__
, is_active
);
4569 * For use by the "clear vnc prefixes" command
4571 /*------------------------------------------
4572 * rfapiDeleteRemotePrefixes
4574 * UI helper: For use by the "clear vnc prefixes" command
4577 * un if set, tunnel must match this prefix
4578 * vn if set, nexthop prefix must match this prefix
4579 * p if set, prefix must match this prefix
4580 * it if set, only look in this import table
4583 * pARcount number of active routes deleted
4584 * pAHcount number of active nves deleted
4585 * pHRcount number of holddown routes deleted
4586 * pHHcount number of holddown nves deleted
4590 --------------------------------------------*/
4591 void rfapiDeleteRemotePrefixes(struct prefix
*un
, struct prefix
*vn
,
4593 struct rfapi_import_table
*arg_it
,
4594 int delete_active
, int delete_holddown
,
4595 uint32_t *pARcount
, uint32_t *pAHcount
,
4596 uint32_t *pHRcount
, uint32_t *pHHcount
)
4600 struct rfapi_import_table
*it
;
4601 uint32_t deleted_holddown_route_count
= 0;
4602 uint32_t deleted_active_route_count
= 0;
4603 uint32_t deleted_holddown_nve_count
= 0;
4604 uint32_t deleted_active_nve_count
= 0;
4605 struct skiplist
*uniq_holddown_nves
;
4606 struct skiplist
*uniq_active_nves
;
4610 bgp
= bgp_get_default(); /* assume 1 instance for now */
4611 /* If no bgp instantiated yet, no vnc prefixes exist */
4618 uniq_holddown_nves
=
4619 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4621 skiplist_new(0, rfapi_nve_addr_cmp
, delete_rem_pfx_na_free
);
4624 * Iterate over all import tables; do a filtered import
4625 * for the afi/safi combination
4634 vnc_zlog_debug_verbose(
4635 "%s: calling rfapiDeleteRemotePrefixesIt() on (IP) import %p",
4638 rfapiDeleteRemotePrefixesIt(
4639 bgp
, it
, un
, vn
, p
, delete_active
, delete_holddown
,
4640 &deleted_active_route_count
, &deleted_active_nve_count
,
4641 &deleted_holddown_route_count
,
4642 &deleted_holddown_nve_count
, uniq_active_nves
,
4643 uniq_holddown_nves
);
4652 * Now iterate over L2 import tables
4654 if (h
->import_mac
&& !(p
&& (p
->family
!= AF_ETHERNET
))) {
4656 void *cursor
= NULL
;
4660 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4662 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4665 vnc_zlog_debug_verbose(
4666 "%s: calling rfapiDeleteRemotePrefixesIt() on import_mac %p",
4669 rfapiDeleteRemotePrefixesIt(
4670 bgp
, it
, un
, vn
, p
, delete_active
,
4671 delete_holddown
, &deleted_active_route_count
,
4672 &deleted_active_nve_count
,
4673 &deleted_holddown_route_count
,
4674 &deleted_holddown_nve_count
, uniq_active_nves
,
4675 uniq_holddown_nves
);
4680 * our custom element freeing function above counts as it deletes
4682 skiplist_free(uniq_holddown_nves
);
4683 skiplist_free(uniq_active_nves
);
4686 *pARcount
= deleted_active_route_count
;
4688 *pAHcount
= deleted_active_nve_count
;
4690 *pHRcount
= deleted_holddown_route_count
;
4692 *pHHcount
= deleted_holddown_nve_count
;
4697 /*------------------------------------------
4698 * rfapiCountRemoteRoutes
4700 * UI helper: count VRF routes from BGP side
4705 * pALRcount count of active local routes
4706 * pARRcount count of active remote routes
4707 * pHRcount count of holddown routes
4708 * pIRcount count of direct imported routes
4712 --------------------------------------------*/
4713 void rfapiCountAllItRoutes(int *pALRcount
, /* active local routes */
4714 int *pARRcount
, /* active remote routes */
4715 int *pHRcount
, /* holddown routes */
4716 int *pIRcount
) /* imported routes */
4720 struct rfapi_import_table
*it
;
4723 int total_active_local
= 0;
4724 int total_active_remote
= 0;
4725 int total_holddown
= 0;
4726 int total_imported
= 0;
4728 bgp
= bgp_get_default(); /* assume 1 instance for now */
4735 * Iterate over all import tables; do a filtered import
4736 * for the afi/safi combination
4739 for (it
= h
->imports
; it
; it
= it
->next
) {
4741 for (afi
= AFI_IP
; afi
< AFI_MAX
; ++afi
) {
4743 total_active_local
+= it
->local_count
[afi
];
4744 total_active_remote
+= it
->remote_count
[afi
];
4745 total_holddown
+= it
->holddown_count
[afi
];
4746 total_imported
+= it
->imported_count
[afi
];
4753 if (h
->import_mac
) {
4755 rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4757 !rc
; rc
= skiplist_next(h
->import_mac
, NULL
, (void **)&it
,
4760 total_active_local
+= it
->local_count
[AFI_L2VPN
];
4761 total_active_remote
+= it
->remote_count
[AFI_L2VPN
];
4762 total_holddown
+= it
->holddown_count
[AFI_L2VPN
];
4763 total_imported
+= it
->imported_count
[AFI_L2VPN
];
4769 *pALRcount
= total_active_local
;
4772 *pARRcount
= total_active_remote
;
4775 *pHRcount
= total_holddown
;
4778 *pIRcount
= total_imported
;
4782 /*------------------------------------------
4783 * rfapiGetHolddownFromLifetime
4785 * calculate holddown value based on lifetime
4791 * Holddown value based on lifetime, holddown_factor,
4792 * and RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
4794 --------------------------------------------*/
4795 /* hold down time maxes out at RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY */
4796 uint32_t rfapiGetHolddownFromLifetime(uint32_t lifetime
)
4801 bgp
= bgp_get_default();
4802 if (bgp
&& bgp
->rfapi_cfg
)
4803 factor
= bgp
->rfapi_cfg
->rfp_cfg
.holddown_factor
;
4805 factor
= RFAPI_RFP_CFG_DEFAULT_HOLDDOWN_FACTOR
;
4807 if (factor
< 100 || lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4808 lifetime
= lifetime
* factor
/ 100;
4809 if (lifetime
< RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
)
4812 return RFAPI_LIFETIME_INFINITE_WITHDRAW_DELAY
;