1 /* PIM support for VxLAN BUM flooding
3 * Copyright (C) 2019 Cumulus Networks, Inc.
5 * This file is part of FRR.
7 * FRR is free software; you can redistribute it and/or modify it
8 * under the terms of the GNU General Public License as published by the
9 * Free Software Foundation; either version 2, or (at your option) any
12 * FRR is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License for more details.
16 * This program is free software; you can redistribute it and/or modify
17 * it under the terms of the GNU General Public License as published by
18 * the Free Software Foundation; either version 2 of the License, or
19 * (at your option) any later version.
31 #include "pim_iface.h"
32 #include "pim_memory.h"
34 #include "pim_register.h"
36 #include "pim_upstream.h"
37 #include "pim_ifchannel.h"
39 #include "pim_zebra.h"
40 #include "pim_vxlan.h"
43 /* pim-vxlan global info */
44 struct pim_vxlan vxlan_info
, *pim_vxlan_p
= &vxlan_info
;
46 static void pim_vxlan_work_timer_setup(bool start
);
47 static void pim_vxlan_set_peerlink_rif(struct pim_instance
*pim
,
48 struct interface
*ifp
);
50 /*************************** vxlan work list **********************************
51 * A work list is maintained for staggered generation of pim null register
52 * messages for vxlan SG entries that are in a reg_join state.
54 * A max of 500 NULL registers are generated at one shot. If paused reg
55 * generation continues on the next second and so on till all register
56 * messages have been sent out. And the process is restarted every 60s.
58 * purpose of this null register generation is to setup the SPT and maintain
59 * independent of the presence of overlay BUM traffic.
60 ****************************************************************************/
61 static void pim_vxlan_do_reg_work(void)
63 struct listnode
*listnode
;
65 struct pim_vxlan_sg
*vxlan_sg
;
70 if (sec_count
> PIM_VXLAN_NULL_REG_INTERVAL
) {
72 listnode
= vxlan_info
.next_work
?
73 vxlan_info
.next_work
:
74 vxlan_info
.work_list
->head
;
75 if (PIM_DEBUG_VXLAN
&& listnode
)
76 zlog_debug("vxlan SG work %s",
77 vxlan_info
.next_work
? "continues" : "starts");
79 listnode
= vxlan_info
.next_work
;
82 for (; listnode
; listnode
= listnode
->next
) {
83 vxlan_sg
= (struct pim_vxlan_sg
*)listnode
->data
;
84 if (vxlan_sg
->up
&& (vxlan_sg
->up
->reg_state
== PIM_REG_JOIN
)) {
86 zlog_debug("vxlan SG %s periodic NULL register",
90 * If we are on the work queue *and* the rpf
91 * has been lost on the vxlan_sg->up let's
92 * make sure that we don't send it.
94 if (vxlan_sg
->up
->rpf
.source_nexthop
.interface
) {
95 pim_null_register_send(vxlan_sg
->up
);
100 if (work_cnt
> vxlan_info
.max_work_cnt
) {
101 vxlan_info
.next_work
= listnode
->next
;
103 zlog_debug("vxlan SG %d work items proc and pause",
111 zlog_debug("vxlan SG %d work items proc", work_cnt
);
113 vxlan_info
.next_work
= NULL
;
116 /* Staggered work related info is initialized when the first work comes
119 static void pim_vxlan_init_work(void)
121 if (vxlan_info
.flags
& PIM_VXLANF_WORK_INITED
)
124 vxlan_info
.max_work_cnt
= PIM_VXLAN_WORK_MAX
;
125 vxlan_info
.flags
|= PIM_VXLANF_WORK_INITED
;
126 vxlan_info
.work_list
= list_new();
127 pim_vxlan_work_timer_setup(true/* start */);
130 static void pim_vxlan_add_work(struct pim_vxlan_sg
*vxlan_sg
)
132 if (vxlan_sg
->flags
& PIM_VXLAN_SGF_DEL_IN_PROG
) {
134 zlog_debug("vxlan SG %s skip work list; del-in-prog",
139 pim_vxlan_init_work();
141 /* already a part of the work list */
142 if (vxlan_sg
->work_node
)
146 zlog_debug("vxlan SG %s work list add",
148 vxlan_sg
->work_node
= listnode_add(vxlan_info
.work_list
, vxlan_sg
);
149 /* XXX: adjust max_work_cnt if needed */
152 static void pim_vxlan_del_work(struct pim_vxlan_sg
*vxlan_sg
)
154 if (!vxlan_sg
->work_node
)
158 zlog_debug("vxlan SG %s work list del",
161 if (vxlan_sg
->work_node
== vxlan_info
.next_work
)
162 vxlan_info
.next_work
= vxlan_sg
->work_node
->next
;
164 list_delete_node(vxlan_info
.work_list
, vxlan_sg
->work_node
);
165 vxlan_sg
->work_node
= NULL
;
168 void pim_vxlan_update_sg_reg_state(struct pim_instance
*pim
,
169 struct pim_upstream
*up
, bool reg_join
)
171 struct pim_vxlan_sg
*vxlan_sg
;
173 vxlan_sg
= pim_vxlan_sg_find(pim
, &up
->sg
);
177 /* add the vxlan sg entry to a work list for periodic reg joins.
178 * the entry will stay in the list as long as the register state is
182 pim_vxlan_add_work(vxlan_sg
);
184 pim_vxlan_del_work(vxlan_sg
);
187 static void pim_vxlan_work_timer_cb(struct thread
*t
)
189 pim_vxlan_do_reg_work();
190 pim_vxlan_work_timer_setup(true /* start */);
193 /* global 1second timer used for periodic processing */
194 static void pim_vxlan_work_timer_setup(bool start
)
196 THREAD_OFF(vxlan_info
.work_timer
);
198 thread_add_timer(router
->master
, pim_vxlan_work_timer_cb
, NULL
,
199 PIM_VXLAN_WORK_TIME
, &vxlan_info
.work_timer
);
202 /**************************** vxlan origination mroutes ***********************
203 * For every (local-vtep-ip, bum-mcast-grp) registered by evpn an origination
204 * mroute is setup by pimd. The purpose of this mroute is to forward vxlan
205 * encapsulated BUM (broadcast, unknown-unicast and unknown-multicast packets
206 * over the underlay.)
208 * Sample mroute (single VTEP):
209 * (27.0.0.7, 239.1.1.100) Iif: lo Oifs: uplink-1
211 * Sample mroute (anycast VTEP):
212 * (36.0.0.9, 239.1.1.100) Iif: peerlink-3.4094\
213 * Oifs: peerlink-3.4094 uplink-1
214 ***************************************************************************/
215 static void pim_vxlan_orig_mr_up_del(struct pim_vxlan_sg
*vxlan_sg
)
217 struct pim_upstream
*up
= vxlan_sg
->up
;
223 zlog_debug("vxlan SG %s orig mroute-up del",
228 if (up
->flags
& PIM_UPSTREAM_FLAG_MASK_SRC_VXLAN_ORIG
) {
229 /* clear out all the vxlan properties */
230 up
->flags
&= ~(PIM_UPSTREAM_FLAG_MASK_SRC_VXLAN_ORIG
|
231 PIM_UPSTREAM_FLAG_MASK_STATIC_IIF
|
232 PIM_UPSTREAM_FLAG_MASK_DISABLE_KAT_EXPIRY
|
233 PIM_UPSTREAM_FLAG_MASK_FORCE_PIMREG
|
234 PIM_UPSTREAM_FLAG_MASK_NO_PIMREG_DATA
|
235 PIM_UPSTREAM_FLAG_MASK_ALLOW_IIF_IN_OIL
);
237 /* We bring things to a grinding halt by force expirying
238 * the kat. Doing this will also remove the reference we
239 * created as a "vxlan" source and delete the upstream entry
240 * if there are no other references.
242 if (PIM_UPSTREAM_FLAG_TEST_SRC_STREAM(up
->flags
)) {
243 THREAD_OFF(up
->t_ka_timer
);
244 up
= pim_upstream_keep_alive_timer_proc(up
);
246 /* this is really unexpected as we force vxlan
247 * origination mroutes active sources but just in
250 up
= pim_upstream_del(vxlan_sg
->pim
, up
, __func__
);
252 /* if there are other references register the source
256 enum pim_rpf_result r
;
258 r
= pim_rpf_update(vxlan_sg
->pim
, up
, NULL
, __func__
);
259 if (r
== PIM_RPF_FAILURE
) {
262 "vxlan SG %s rpf_update failure",
269 static void pim_vxlan_orig_mr_up_iif_update(struct pim_vxlan_sg
*vxlan_sg
)
271 /* update MFC with the new IIF */
272 pim_upstream_fill_static_iif(vxlan_sg
->up
, vxlan_sg
->iif
);
273 pim_upstream_mroute_iif_update(vxlan_sg
->up
->channel_oil
, __func__
);
276 zlog_debug("vxlan SG %s orig mroute-up updated with iif %s",
278 vxlan_sg
->iif
?vxlan_sg
->iif
->name
:"-");
282 /* For every VxLAN BUM multicast group we setup a SG-up that has the following
283 * "forced properties" -
284 * 1. Directly connected on a DR interface i.e. we must act as an FHR
285 * 2. We prime the pump i.e. no multicast data is needed to register this
286 * source with the FHR. To do that we send periodic null registers if
287 * the SG entry is in a register-join state. We also prevent expiry of
289 * 3. As this SG is setup without data there is no need to register encapsulate
290 * data traffic. This encapsulation is explicitly skipped for the following
292 * a) Many levels of encapsulation are needed creating MTU disc challenges.
293 * Overlay BUM is encapsulated in a vxlan/UDP/IP header and then
294 * encapsulated again in a pim-register header.
295 * b) On a vxlan-aa setup both switches rx a copy of each BUM packet. if
296 * they both reg encapsulated traffic the RP will accept the duplicates
297 * as there are no RPF checks for this encapsulated data.
298 * a), b) can be workarounded if needed, but there is really no need because
299 * of (2) i.e. the pump is primed without data.
301 static void pim_vxlan_orig_mr_up_add(struct pim_vxlan_sg
*vxlan_sg
)
303 struct pim_upstream
*up
;
304 struct pim_interface
*term_ifp
;
307 struct pim_instance
*pim
= vxlan_sg
->pim
;
315 zlog_debug("vxlan SG %s orig mroute-up add with iif %s",
317 vxlan_sg
->iif
?vxlan_sg
->iif
->name
:"-");
319 PIM_UPSTREAM_FLAG_SET_SRC_VXLAN_ORIG(flags
);
320 /* pin the IIF to lo or peerlink-subinterface and disable NHT */
321 PIM_UPSTREAM_FLAG_SET_STATIC_IIF(flags
);
322 /* Fake traffic by setting SRC_STREAM and starting KAT */
323 /* We intentionally skip updating ref count for SRC_STREAM/FHR.
324 * Setting SRC_VXLAN should have already created a reference
325 * preventing the entry from being deleted
327 PIM_UPSTREAM_FLAG_SET_FHR(flags
);
328 PIM_UPSTREAM_FLAG_SET_SRC_STREAM(flags
);
329 /* Force pimreg even if non-DR. This is needed on a MLAG setup for
332 PIM_UPSTREAM_FLAG_SET_FORCE_PIMREG(flags
);
333 /* prevent KAT expiry. we want the MDT setup even if there is no BUM
336 PIM_UPSTREAM_FLAG_SET_DISABLE_KAT_EXPIRY(flags
);
337 /* SPT for vxlan BUM groups is primed and maintained via NULL
338 * registers so there is no need to reg-encapsulate
339 * vxlan-encapsulated overlay data traffic
341 PIM_UPSTREAM_FLAG_SET_NO_PIMREG_DATA(flags
);
342 /* On a MLAG setup we force a copy to the MLAG peer while also
343 * accepting traffic from the peer. To do this we set peerlink-rif as
344 * the IIF and also add it to the OIL
346 PIM_UPSTREAM_FLAG_SET_ALLOW_IIF_IN_OIL(flags
);
348 /* XXX: todo: defer pim_upstream add if pim is not enabled on the iif */
349 up
= pim_upstream_find(vxlan_sg
->pim
, &vxlan_sg
->sg
);
351 /* if the iif is set to something other than the vxlan_sg->iif
352 * we must dereg the old nexthop and force to new "static"
355 if (!PIM_UPSTREAM_FLAG_TEST_STATIC_IIF(up
->flags
)) {
356 pim_addr_to_prefix(&nht_p
, up
->upstream_addr
);
357 pim_delete_tracked_nexthop(vxlan_sg
->pim
, &nht_p
, up
,
360 /* We are acting FHR; clear out use_rpt setting if any */
361 pim_upstream_update_use_rpt(up
, false /*update_mroute*/);
362 pim_upstream_ref(up
, flags
, __func__
);
364 term_ifp
= pim_vxlan_get_term_ifp(pim
);
365 /* mute termination device on origination mroutes */
367 pim_channel_update_oif_mute(up
->channel_oil
,
369 pim_vxlan_orig_mr_up_iif_update(vxlan_sg
);
370 /* mute pimreg on origination mroutes */
372 pim_channel_update_oif_mute(up
->channel_oil
,
373 pim
->regiface
->info
);
375 up
= pim_upstream_add(vxlan_sg
->pim
, &vxlan_sg
->sg
,
376 vxlan_sg
->iif
, flags
, __func__
, NULL
);
382 zlog_debug("vxlan SG %s orig mroute-up add failed",
387 pim_upstream_keep_alive_timer_start(up
, vxlan_sg
->pim
->keep_alive_time
);
389 /* register the source with the RP */
390 switch (up
->reg_state
) {
393 pim_register_join(up
);
394 pim_null_register_send(up
);
398 /* if the pim upstream entry is already in reg-join state
399 * send null_register right away and add to the register
402 pim_null_register_send(up
);
403 pim_vxlan_update_sg_reg_state(pim
, up
, true);
406 case PIM_REG_JOIN_PENDING
:
411 /* update the inherited OIL */
412 pim_upstream_inherited_olist(vxlan_sg
->pim
, up
);
413 if (!up
->channel_oil
->installed
)
414 pim_upstream_mroute_add(up
->channel_oil
, __func__
);
417 static void pim_vxlan_orig_mr_oif_add(struct pim_vxlan_sg
*vxlan_sg
)
419 if (!vxlan_sg
->up
|| !vxlan_sg
->orig_oif
)
423 zlog_debug("vxlan SG %s oif %s add",
424 vxlan_sg
->sg_str
, vxlan_sg
->orig_oif
->name
);
426 vxlan_sg
->flags
|= PIM_VXLAN_SGF_OIF_INSTALLED
;
427 pim_channel_add_oif(vxlan_sg
->up
->channel_oil
,
428 vxlan_sg
->orig_oif
, PIM_OIF_FLAG_PROTO_VXLAN
,
432 static void pim_vxlan_orig_mr_oif_del(struct pim_vxlan_sg
*vxlan_sg
)
434 struct interface
*orig_oif
;
436 orig_oif
= vxlan_sg
->orig_oif
;
437 vxlan_sg
->orig_oif
= NULL
;
439 if (!(vxlan_sg
->flags
& PIM_VXLAN_SGF_OIF_INSTALLED
))
443 zlog_debug("vxlan SG %s oif %s del",
444 vxlan_sg
->sg_str
, orig_oif
->name
);
446 vxlan_sg
->flags
&= ~PIM_VXLAN_SGF_OIF_INSTALLED
;
447 pim_channel_del_oif(vxlan_sg
->up
->channel_oil
,
448 orig_oif
, PIM_OIF_FLAG_PROTO_VXLAN
, __func__
);
451 static inline struct interface
*pim_vxlan_orig_mr_oif_get(
452 struct pim_instance
*pim
)
454 return (vxlan_mlag
.flags
& PIM_VXLAN_MLAGF_ENABLED
) ?
455 pim
->vxlan
.peerlink_rif
: NULL
;
458 /* Single VTEPs: IIF for the vxlan-origination-mroutes is lo or vrf-dev (if
459 * the mroute is in a non-default vrf).
460 * Anycast VTEPs: IIF is the MLAG ISL/peerlink.
462 static inline struct interface
*pim_vxlan_orig_mr_iif_get(
463 struct pim_instance
*pim
)
465 return ((vxlan_mlag
.flags
& PIM_VXLAN_MLAGF_ENABLED
) &&
466 pim
->vxlan
.peerlink_rif
) ?
467 pim
->vxlan
.peerlink_rif
: pim
->vxlan
.default_iif
;
470 static bool pim_vxlan_orig_mr_add_is_ok(struct pim_vxlan_sg
*vxlan_sg
)
472 struct pim_interface
*pim_ifp
;
474 vxlan_sg
->iif
= pim_vxlan_orig_mr_iif_get(vxlan_sg
->pim
);
478 pim_ifp
= (struct pim_interface
*)vxlan_sg
->iif
->info
;
479 if (!pim_ifp
|| (pim_ifp
->mroute_vif_index
< 0))
485 static void pim_vxlan_orig_mr_install(struct pim_vxlan_sg
*vxlan_sg
)
487 pim_vxlan_orig_mr_up_add(vxlan_sg
);
489 vxlan_sg
->orig_oif
= pim_vxlan_orig_mr_oif_get(vxlan_sg
->pim
);
490 pim_vxlan_orig_mr_oif_add(vxlan_sg
);
493 static void pim_vxlan_orig_mr_add(struct pim_vxlan_sg
*vxlan_sg
)
495 if (!pim_vxlan_orig_mr_add_is_ok(vxlan_sg
))
499 zlog_debug("vxlan SG %s orig-mr add", vxlan_sg
->sg_str
);
501 pim_vxlan_orig_mr_install(vxlan_sg
);
504 static void pim_vxlan_orig_mr_del(struct pim_vxlan_sg
*vxlan_sg
)
507 zlog_debug("vxlan SG %s orig-mr del", vxlan_sg
->sg_str
);
509 pim_vxlan_orig_mr_oif_del(vxlan_sg
);
510 pim_vxlan_orig_mr_up_del(vxlan_sg
);
513 static void pim_vxlan_orig_mr_iif_update(struct hash_bucket
*bucket
, void *arg
)
515 struct interface
*ifp
;
516 struct pim_vxlan_sg
*vxlan_sg
= (struct pim_vxlan_sg
*)bucket
->data
;
517 struct interface
*old_iif
= vxlan_sg
->iif
;
519 if (!pim_vxlan_is_orig_mroute(vxlan_sg
))
522 ifp
= pim_vxlan_orig_mr_iif_get(vxlan_sg
->pim
);
524 zlog_debug("vxlan SG %s iif changed from %s to %s",
526 old_iif
? old_iif
->name
: "-",
527 ifp
? ifp
->name
: "-");
529 if (pim_vxlan_orig_mr_add_is_ok(vxlan_sg
)) {
531 /* upstream exists but iif changed */
532 pim_vxlan_orig_mr_up_iif_update(vxlan_sg
);
535 pim_vxlan_orig_mr_install(vxlan_sg
);
538 pim_vxlan_orig_mr_del(vxlan_sg
);
542 /**************************** vxlan termination mroutes ***********************
543 * For every bum-mcast-grp registered by evpn a *G termination
544 * mroute is setup by pimd. The purpose of this mroute is to pull down vxlan
545 * packets with the bum-mcast-grp dip from the underlay and terminate the
546 * tunnel. This is done by including the vxlan termination device (ipmr-lo) in
547 * its OIL. The vxlan de-capsulated packets are subject to subsequent overlay
551 * (0.0.0.0, 239.1.1.100) Iif: uplink-1 Oifs: ipmr-lo, uplink-1
552 *****************************************************************************/
553 struct pim_interface
*pim_vxlan_get_term_ifp(struct pim_instance
*pim
)
555 return pim
->vxlan
.term_if
?
556 (struct pim_interface
*)pim
->vxlan
.term_if
->info
: NULL
;
559 static void pim_vxlan_term_mr_oif_add(struct pim_vxlan_sg
*vxlan_sg
)
561 if (vxlan_sg
->flags
& PIM_VXLAN_SGF_OIF_INSTALLED
)
565 zlog_debug("vxlan SG %s term-oif %s add",
566 vxlan_sg
->sg_str
, vxlan_sg
->term_oif
->name
);
568 if (pim_ifchannel_local_membership_add(vxlan_sg
->term_oif
,
569 &vxlan_sg
->sg
, true /*is_vxlan */)) {
570 vxlan_sg
->flags
|= PIM_VXLAN_SGF_OIF_INSTALLED
;
571 /* update the inherited OIL */
572 /* XXX - I don't see the inherited OIL updated when a local
573 * member is added. And that probably needs to be fixed. Till
574 * that happens we do a force update on the inherited OIL
577 pim_upstream_inherited_olist(vxlan_sg
->pim
, vxlan_sg
->up
);
579 zlog_warn("vxlan SG %s term-oif %s add failed",
580 vxlan_sg
->sg_str
, vxlan_sg
->term_oif
->name
);
584 static void pim_vxlan_term_mr_oif_del(struct pim_vxlan_sg
*vxlan_sg
)
586 if (!(vxlan_sg
->flags
& PIM_VXLAN_SGF_OIF_INSTALLED
))
590 zlog_debug("vxlan SG %s oif %s del",
591 vxlan_sg
->sg_str
, vxlan_sg
->term_oif
->name
);
593 vxlan_sg
->flags
&= ~PIM_VXLAN_SGF_OIF_INSTALLED
;
594 pim_ifchannel_local_membership_del(vxlan_sg
->term_oif
, &vxlan_sg
->sg
);
595 /* update the inherited OIL */
596 /* XXX - I don't see the inherited OIL updated when a local member
597 * is deleted. And that probably needs to be fixed. Till that happens
598 * we do a force update on the inherited OIL here.
600 pim_upstream_inherited_olist(vxlan_sg
->pim
, vxlan_sg
->up
);
603 static void pim_vxlan_update_sg_entry_mlag(struct pim_instance
*pim
,
604 struct pim_upstream
*up
, bool inherit
)
608 if (inherit
&& up
->parent
&&
609 PIM_UPSTREAM_FLAG_TEST_MLAG_VXLAN(up
->parent
->flags
) &&
610 PIM_UPSTREAM_FLAG_TEST_MLAG_NON_DF(up
->parent
->flags
))
613 pim_mlag_up_df_role_update(pim
, up
, is_df
, "inherit_xg_df");
616 /* We run MLAG DF election only on mroutes that have the termination
617 * device ipmr-lo in the immediate OIL. This is only (*, G) entries at the
618 * moment. For (S, G) entries that (with ipmr-lo in the inherited OIL) we
619 * inherit the DF role from the (*, G) entry.
621 void pim_vxlan_inherit_mlag_flags(struct pim_instance
*pim
,
622 struct pim_upstream
*up
, bool inherit
)
624 struct listnode
*listnode
;
625 struct pim_upstream
*child
;
627 for (ALL_LIST_ELEMENTS_RO(up
->sources
, listnode
,
629 pim_vxlan_update_sg_entry_mlag(pim
,
630 child
, true /* inherit */);
634 static void pim_vxlan_term_mr_up_add(struct pim_vxlan_sg
*vxlan_sg
)
636 struct pim_upstream
*up
;
645 zlog_debug("vxlan SG %s term mroute-up add",
648 PIM_UPSTREAM_FLAG_SET_SRC_VXLAN_TERM(flags
);
649 /* enable MLAG designated-forwarder election on termination mroutes */
650 PIM_UPSTREAM_FLAG_SET_MLAG_VXLAN(flags
);
652 up
= pim_upstream_add(vxlan_sg
->pim
, &vxlan_sg
->sg
, NULL
/* iif */,
653 flags
, __func__
, NULL
);
657 zlog_warn("vxlan SG %s term mroute-up add failed",
662 /* update existing SG entries with the parent's MLAG flag */
663 pim_vxlan_inherit_mlag_flags(vxlan_sg
->pim
, up
, true /*enable*/);
666 static void pim_vxlan_term_mr_up_del(struct pim_vxlan_sg
*vxlan_sg
)
668 struct pim_upstream
*up
= vxlan_sg
->up
;
674 zlog_debug("vxlan SG %s term mroute-up del",
677 if (up
->flags
& PIM_UPSTREAM_FLAG_MASK_SRC_VXLAN_TERM
) {
678 /* update SG entries that are inheriting from this XG entry */
679 pim_vxlan_inherit_mlag_flags(vxlan_sg
->pim
, up
,
681 /* clear out all the vxlan related flags */
682 up
->flags
&= ~(PIM_UPSTREAM_FLAG_MASK_SRC_VXLAN_TERM
|
683 PIM_UPSTREAM_FLAG_MASK_MLAG_VXLAN
);
684 pim_mlag_up_local_del(vxlan_sg
->pim
, up
);
685 pim_upstream_del(vxlan_sg
->pim
, up
, __func__
);
689 static void pim_vxlan_term_mr_add(struct pim_vxlan_sg
*vxlan_sg
)
692 zlog_debug("vxlan SG %s term mroute add", vxlan_sg
->sg_str
);
694 vxlan_sg
->term_oif
= vxlan_sg
->pim
->vxlan
.term_if
;
695 if (!vxlan_sg
->term_oif
)
696 /* defer termination mroute till we have a termination device */
699 pim_vxlan_term_mr_up_add(vxlan_sg
);
700 /* set up local membership for the term-oif */
701 pim_vxlan_term_mr_oif_add(vxlan_sg
);
704 static void pim_vxlan_term_mr_del(struct pim_vxlan_sg
*vxlan_sg
)
707 zlog_debug("vxlan SG %s term mroute del", vxlan_sg
->sg_str
);
709 /* remove local membership associated with the term oif */
710 pim_vxlan_term_mr_oif_del(vxlan_sg
);
711 /* remove references to the upstream entry */
712 pim_vxlan_term_mr_up_del(vxlan_sg
);
715 /************************** vxlan SG cache management ************************/
716 static unsigned int pim_vxlan_sg_hash_key_make(const void *p
)
718 const struct pim_vxlan_sg
*vxlan_sg
= p
;
720 return pim_sgaddr_hash(vxlan_sg
->sg
, 0);
723 static bool pim_vxlan_sg_hash_eq(const void *p1
, const void *p2
)
725 const struct pim_vxlan_sg
*sg1
= p1
;
726 const struct pim_vxlan_sg
*sg2
= p2
;
728 return !pim_sgaddr_cmp(sg1
->sg
, sg2
->sg
);
731 static struct pim_vxlan_sg
*pim_vxlan_sg_new(struct pim_instance
*pim
,
734 struct pim_vxlan_sg
*vxlan_sg
;
736 vxlan_sg
= XCALLOC(MTYPE_PIM_VXLAN_SG
, sizeof(*vxlan_sg
));
740 snprintfrr(vxlan_sg
->sg_str
, sizeof(vxlan_sg
->sg_str
), "%pSG", sg
);
743 zlog_debug("vxlan SG %s alloc", vxlan_sg
->sg_str
);
745 vxlan_sg
= hash_get(pim
->vxlan
.sg_hash
, vxlan_sg
, hash_alloc_intern
);
747 /* we register with the MLAG daemon in the first VxLAN SG and never
748 * de-register during that life of the pimd
750 if (pim
->vxlan
.sg_hash
->count
== 1) {
751 vxlan_mlag
.flags
|= PIM_VXLAN_MLAGF_DO_REG
;
758 struct pim_vxlan_sg
*pim_vxlan_sg_find(struct pim_instance
*pim
, pim_sgaddr
*sg
)
760 struct pim_vxlan_sg lookup
;
763 return hash_lookup(pim
->vxlan
.sg_hash
, &lookup
);
766 struct pim_vxlan_sg
*pim_vxlan_sg_add(struct pim_instance
*pim
, pim_sgaddr
*sg
)
768 struct pim_vxlan_sg
*vxlan_sg
;
770 vxlan_sg
= pim_vxlan_sg_find(pim
, sg
);
774 vxlan_sg
= pim_vxlan_sg_new(pim
, sg
);
776 if (pim_vxlan_is_orig_mroute(vxlan_sg
))
777 pim_vxlan_orig_mr_add(vxlan_sg
);
779 pim_vxlan_term_mr_add(vxlan_sg
);
784 static void pim_vxlan_sg_del_item(struct pim_vxlan_sg
*vxlan_sg
)
786 vxlan_sg
->flags
|= PIM_VXLAN_SGF_DEL_IN_PROG
;
788 pim_vxlan_del_work(vxlan_sg
);
790 if (pim_vxlan_is_orig_mroute(vxlan_sg
))
791 pim_vxlan_orig_mr_del(vxlan_sg
);
793 pim_vxlan_term_mr_del(vxlan_sg
);
796 zlog_debug("vxlan SG %s free", vxlan_sg
->sg_str
);
798 XFREE(MTYPE_PIM_VXLAN_SG
, vxlan_sg
);
801 void pim_vxlan_sg_del(struct pim_instance
*pim
, pim_sgaddr
*sg
)
803 struct pim_vxlan_sg
*vxlan_sg
;
805 vxlan_sg
= pim_vxlan_sg_find(pim
, sg
);
809 hash_release(pim
->vxlan
.sg_hash
, vxlan_sg
);
810 pim_vxlan_sg_del_item(vxlan_sg
);
813 /******************************* MLAG handling *******************************/
814 bool pim_vxlan_do_mlag_reg(void)
816 return (vxlan_mlag
.flags
& PIM_VXLAN_MLAGF_DO_REG
);
819 /* The peerlink sub-interface is added as an OIF to the origination-mroute.
820 * This is done to send a copy of the multicast-vxlan encapsulated traffic
821 * to the MLAG peer which may mroute it over the underlay if there are any
822 * interested receivers.
824 static void pim_vxlan_sg_peerlink_oif_update(struct hash_bucket
*bucket
,
827 struct interface
*new_oif
= (struct interface
*)arg
;
828 struct pim_vxlan_sg
*vxlan_sg
= (struct pim_vxlan_sg
*)bucket
->data
;
830 if (!pim_vxlan_is_orig_mroute(vxlan_sg
))
833 if (vxlan_sg
->orig_oif
== new_oif
)
836 pim_vxlan_orig_mr_oif_del(vxlan_sg
);
838 vxlan_sg
->orig_oif
= new_oif
;
839 pim_vxlan_orig_mr_oif_add(vxlan_sg
);
842 /* In the case of anycast VTEPs the VTEP-PIP must be used as the
845 bool pim_vxlan_get_register_src(struct pim_instance
*pim
,
846 struct pim_upstream
*up
, struct in_addr
*src_p
)
848 if (!(vxlan_mlag
.flags
& PIM_VXLAN_MLAGF_ENABLED
))
851 /* if address is not available suppress the pim-register */
852 if (vxlan_mlag
.reg_addr
.s_addr
== INADDR_ANY
)
855 *src_p
= vxlan_mlag
.reg_addr
;
859 void pim_vxlan_mlag_update(bool enable
, bool peer_state
, uint32_t role
,
860 struct interface
*peerlink_rif
,
861 struct in_addr
*reg_addr
)
863 struct pim_instance
*pim
;
864 char addr_buf
[INET_ADDRSTRLEN
];
865 struct pim_interface
*pim_ifp
= NULL
;
867 if (PIM_DEBUG_VXLAN
) {
868 inet_ntop(AF_INET
, reg_addr
,
869 addr_buf
, INET_ADDRSTRLEN
);
870 zlog_debug("vxlan MLAG update %s state %s role %d rif %s addr %s",
871 enable
? "enable" : "disable",
872 peer_state
? "up" : "down",
874 peerlink_rif
? peerlink_rif
->name
: "-",
878 /* XXX: for now vxlan termination is only possible in the default VRF
879 * when that changes this will need to change to iterate all VRFs
881 pim
= pim_get_pim_instance(VRF_DEFAULT
);
884 vxlan_mlag
.flags
|= PIM_VXLAN_MLAGF_ENABLED
;
886 vxlan_mlag
.flags
&= ~PIM_VXLAN_MLAGF_ENABLED
;
888 if (vxlan_mlag
.peerlink_rif
!= peerlink_rif
)
889 vxlan_mlag
.peerlink_rif
= peerlink_rif
;
891 vxlan_mlag
.reg_addr
= *reg_addr
;
892 vxlan_mlag
.peer_state
= peer_state
;
893 vxlan_mlag
.role
= role
;
895 /* process changes */
896 if (vxlan_mlag
.peerlink_rif
)
897 pim_ifp
= (struct pim_interface
*)vxlan_mlag
.peerlink_rif
->info
;
898 if ((vxlan_mlag
.flags
& PIM_VXLAN_MLAGF_ENABLED
) &&
899 pim_ifp
&& (pim_ifp
->mroute_vif_index
> 0))
900 pim_vxlan_set_peerlink_rif(pim
, peerlink_rif
);
902 pim_vxlan_set_peerlink_rif(pim
, NULL
);
905 /****************************** misc callbacks *******************************/
906 static void pim_vxlan_set_default_iif(struct pim_instance
*pim
,
907 struct interface
*ifp
)
909 struct interface
*old_iif
;
911 if (pim
->vxlan
.default_iif
== ifp
)
914 old_iif
= pim
->vxlan
.default_iif
;
916 zlog_debug("%s: vxlan default iif changed from %s to %s",
917 __func__
, old_iif
? old_iif
->name
: "-",
918 ifp
? ifp
->name
: "-");
920 old_iif
= pim_vxlan_orig_mr_iif_get(pim
);
921 pim
->vxlan
.default_iif
= ifp
;
922 ifp
= pim_vxlan_orig_mr_iif_get(pim
);
927 zlog_debug("%s: vxlan orig iif changed from %s to %s", __func__
,
928 old_iif
? old_iif
->name
: "-",
929 ifp
? ifp
->name
: "-");
931 /* add/del upstream entries for the existing vxlan SG when the
932 * interface becomes available
934 if (pim
->vxlan
.sg_hash
)
935 hash_iterate(pim
->vxlan
.sg_hash
,
936 pim_vxlan_orig_mr_iif_update
, NULL
);
939 static void pim_vxlan_up_cost_update(struct pim_instance
*pim
,
940 struct pim_upstream
*up
,
941 struct interface
*old_peerlink_rif
)
943 if (!PIM_UPSTREAM_FLAG_TEST_MLAG_VXLAN(up
->flags
))
946 if (up
->rpf
.source_nexthop
.interface
&&
947 ((up
->rpf
.source_nexthop
.interface
==
948 pim
->vxlan
.peerlink_rif
) ||
949 (up
->rpf
.source_nexthop
.interface
==
950 old_peerlink_rif
))) {
952 zlog_debug("RPF cost adjust for %s on peerlink-rif (old: %s, new: %s) change",
955 old_peerlink_rif
->name
: "-",
956 pim
->vxlan
.peerlink_rif
?
957 pim
->vxlan
.peerlink_rif
->name
: "-");
958 pim_mlag_up_local_add(pim
, up
);
962 static void pim_vxlan_term_mr_cost_update(struct hash_bucket
*bucket
, void *arg
)
964 struct interface
*old_peerlink_rif
= (struct interface
*)arg
;
965 struct pim_vxlan_sg
*vxlan_sg
= (struct pim_vxlan_sg
*)bucket
->data
;
966 struct pim_upstream
*up
;
967 struct listnode
*listnode
;
968 struct pim_upstream
*child
;
970 if (pim_vxlan_is_orig_mroute(vxlan_sg
))
973 /* Lookup all XG and SG entries with RPF-interface peerlink_rif */
978 pim_vxlan_up_cost_update(vxlan_sg
->pim
, up
,
981 for (ALL_LIST_ELEMENTS_RO(up
->sources
, listnode
,
983 pim_vxlan_up_cost_update(vxlan_sg
->pim
, child
,
987 static void pim_vxlan_sg_peerlink_rif_update(struct hash_bucket
*bucket
,
990 pim_vxlan_orig_mr_iif_update(bucket
, NULL
);
991 pim_vxlan_term_mr_cost_update(bucket
, arg
);
994 static void pim_vxlan_set_peerlink_rif(struct pim_instance
*pim
,
995 struct interface
*ifp
)
997 struct interface
*old_iif
;
998 struct interface
*new_iif
;
999 struct interface
*old_oif
;
1000 struct interface
*new_oif
;
1002 if (pim
->vxlan
.peerlink_rif
== ifp
)
1005 old_iif
= pim
->vxlan
.peerlink_rif
;
1006 if (PIM_DEBUG_VXLAN
)
1007 zlog_debug("%s: vxlan peerlink_rif changed from %s to %s",
1008 __func__
, old_iif
? old_iif
->name
: "-",
1009 ifp
? ifp
->name
: "-");
1011 old_iif
= pim_vxlan_orig_mr_iif_get(pim
);
1012 old_oif
= pim_vxlan_orig_mr_oif_get(pim
);
1013 pim
->vxlan
.peerlink_rif
= ifp
;
1015 new_iif
= pim_vxlan_orig_mr_iif_get(pim
);
1016 if (old_iif
!= new_iif
) {
1017 if (PIM_DEBUG_VXLAN
)
1018 zlog_debug("%s: vxlan orig iif changed from %s to %s",
1019 __func__
, old_iif
? old_iif
->name
: "-",
1020 new_iif
? new_iif
->name
: "-");
1022 /* add/del upstream entries for the existing vxlan SG when the
1023 * interface becomes available
1025 if (pim
->vxlan
.sg_hash
)
1026 hash_iterate(pim
->vxlan
.sg_hash
,
1027 pim_vxlan_sg_peerlink_rif_update
,
1031 new_oif
= pim_vxlan_orig_mr_oif_get(pim
);
1032 if (old_oif
!= new_oif
) {
1033 if (PIM_DEBUG_VXLAN
)
1034 zlog_debug("%s: vxlan orig oif changed from %s to %s",
1035 __func__
, old_oif
? old_oif
->name
: "-",
1036 new_oif
? new_oif
->name
: "-");
1037 if (pim
->vxlan
.sg_hash
)
1038 hash_iterate(pim
->vxlan
.sg_hash
,
1039 pim_vxlan_sg_peerlink_oif_update
,
1044 static void pim_vxlan_term_mr_oif_update(struct hash_bucket
*bucket
, void *arg
)
1046 struct interface
*ifp
= (struct interface
*)arg
;
1047 struct pim_vxlan_sg
*vxlan_sg
= (struct pim_vxlan_sg
*)bucket
->data
;
1049 if (pim_vxlan_is_orig_mroute(vxlan_sg
))
1052 if (vxlan_sg
->term_oif
== ifp
)
1055 if (PIM_DEBUG_VXLAN
)
1056 zlog_debug("vxlan SG %s term oif changed from %s to %s",
1058 vxlan_sg
->term_oif
? vxlan_sg
->term_oif
->name
: "-",
1059 ifp
? ifp
->name
: "-");
1061 pim_vxlan_term_mr_del(vxlan_sg
);
1062 vxlan_sg
->term_oif
= ifp
;
1063 pim_vxlan_term_mr_add(vxlan_sg
);
1066 static void pim_vxlan_term_oif_update(struct pim_instance
*pim
,
1067 struct interface
*ifp
)
1069 if (pim
->vxlan
.term_if
== ifp
)
1072 if (PIM_DEBUG_VXLAN
)
1073 zlog_debug("vxlan term oif changed from %s to %s",
1074 pim
->vxlan
.term_if
? pim
->vxlan
.term_if
->name
: "-",
1075 ifp
? ifp
->name
: "-");
1077 pim
->vxlan
.term_if
= ifp
;
1078 if (pim
->vxlan
.sg_hash
)
1079 hash_iterate(pim
->vxlan
.sg_hash
,
1080 pim_vxlan_term_mr_oif_update
, ifp
);
1083 void pim_vxlan_add_vif(struct interface
*ifp
)
1085 struct pim_interface
*pim_ifp
= ifp
->info
;
1086 struct pim_instance
*pim
= pim_ifp
->pim
;
1088 if (pim
->vrf
->vrf_id
!= VRF_DEFAULT
)
1091 if (if_is_loopback(ifp
))
1092 pim_vxlan_set_default_iif(pim
, ifp
);
1094 if (vxlan_mlag
.flags
& PIM_VXLAN_MLAGF_ENABLED
&&
1095 (ifp
== vxlan_mlag
.peerlink_rif
))
1096 pim_vxlan_set_peerlink_rif(pim
, ifp
);
1098 if (pim
->vxlan
.term_if_cfg
== ifp
)
1099 pim_vxlan_term_oif_update(pim
, ifp
);
1102 void pim_vxlan_del_vif(struct interface
*ifp
)
1104 struct pim_interface
*pim_ifp
= ifp
->info
;
1105 struct pim_instance
*pim
= pim_ifp
->pim
;
1107 if (pim
->vrf
->vrf_id
!= VRF_DEFAULT
)
1110 if (pim
->vxlan
.default_iif
== ifp
)
1111 pim_vxlan_set_default_iif(pim
, NULL
);
1113 if (pim
->vxlan
.peerlink_rif
== ifp
)
1114 pim_vxlan_set_peerlink_rif(pim
, NULL
);
1116 if (pim
->vxlan
.term_if
== ifp
)
1117 pim_vxlan_term_oif_update(pim
, NULL
);
1120 /* enable pim implicitly on the termination device add */
1121 void pim_vxlan_add_term_dev(struct pim_instance
*pim
,
1122 struct interface
*ifp
)
1124 struct pim_interface
*pim_ifp
;
1126 if (pim
->vxlan
.term_if_cfg
== ifp
)
1129 if (PIM_DEBUG_VXLAN
)
1130 zlog_debug("vxlan term oif cfg changed from %s to %s",
1131 pim
->vxlan
.term_if_cfg
?
1132 pim
->vxlan
.term_if_cfg
->name
: "-",
1135 pim
->vxlan
.term_if_cfg
= ifp
;
1137 /* enable pim on the term ifp */
1138 pim_ifp
= (struct pim_interface
*)ifp
->info
;
1140 PIM_IF_DO_PIM(pim_ifp
->options
);
1141 /* ifp is already oper up; activate it as a term dev */
1142 if (pim_ifp
->mroute_vif_index
>= 0)
1143 pim_vxlan_term_oif_update(pim
, ifp
);
1145 /* ensure that pimreg exists before using the newly created
1146 * vxlan termination device
1148 pim_if_create_pimreg(pim
);
1149 (void)pim_if_new(ifp
, false /*igmp*/, true /*pim*/,
1150 false /*pimreg*/, true /*vxlan_term*/);
1154 /* disable pim implicitly, if needed, on the termination device deletion */
1155 void pim_vxlan_del_term_dev(struct pim_instance
*pim
)
1157 struct interface
*ifp
= pim
->vxlan
.term_if_cfg
;
1158 struct pim_interface
*pim_ifp
;
1160 if (PIM_DEBUG_VXLAN
)
1161 zlog_debug("vxlan term oif cfg changed from %s to -",
1164 pim
->vxlan
.term_if_cfg
= NULL
;
1166 pim_ifp
= (struct pim_interface
*)ifp
->info
;
1168 PIM_IF_DONT_PIM(pim_ifp
->options
);
1169 if (!PIM_IF_TEST_IGMP(pim_ifp
->options
))
1174 void pim_vxlan_init(struct pim_instance
*pim
)
1178 snprintf(hash_name
, sizeof(hash_name
),
1179 "PIM %s vxlan SG hash", pim
->vrf
->name
);
1180 pim
->vxlan
.sg_hash
= hash_create(pim_vxlan_sg_hash_key_make
,
1181 pim_vxlan_sg_hash_eq
, hash_name
);
1184 void pim_vxlan_exit(struct pim_instance
*pim
)
1186 if (pim
->vxlan
.sg_hash
) {
1187 hash_clean(pim
->vxlan
.sg_hash
,
1188 (void (*)(void *))pim_vxlan_sg_del_item
);
1189 hash_free(pim
->vxlan
.sg_hash
);
1190 pim
->vxlan
.sg_hash
= NULL
;
1194 void pim_vxlan_terminate(void)
1196 pim_vxlan_work_timer_setup(false);