diff options
Diffstat (limited to 'pimd/pim_vxlan.c')
| -rw-r--r-- | pimd/pim_vxlan.c | 1218 |
1 files changed, 1218 insertions, 0 deletions
diff --git a/pimd/pim_vxlan.c b/pimd/pim_vxlan.c new file mode 100644 index 0000000..9650da8 --- /dev/null +++ b/pimd/pim_vxlan.c @@ -0,0 +1,1218 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* PIM support for VxLAN BUM flooding + * + * Copyright (C) 2019 Cumulus Networks, Inc. + */ + +#include <zebra.h> + +#include <hash.h> +#include <jhash.h> +#include <log.h> +#include <prefix.h> +#include <vrf.h> + +#include "pimd.h" +#include "pim_iface.h" +#include "pim_memory.h" +#include "pim_oil.h" +#include "pim_register.h" +#include "pim_str.h" +#include "pim_upstream.h" +#include "pim_ifchannel.h" +#include "pim_nht.h" +#include "pim_zebra.h" +#include "pim_vxlan.h" +#include "pim_mlag.h" + +/* pim-vxlan global info */ +struct pim_vxlan vxlan_info, *pim_vxlan_p = &vxlan_info; + +static void pim_vxlan_work_timer_setup(bool start); +static void pim_vxlan_set_peerlink_rif(struct pim_instance *pim, + struct interface *ifp); + +/* + * The rp info has gone from no path to having a + * path. Let's immediately send out the null pim register + * as that else we will be sitting for up to 60 seconds waiting + * for it too pop. Which is not cool. + */ +void pim_vxlan_rp_info_is_alive(struct pim_instance *pim, + struct pim_rpf *rpg_changed) +{ + struct listnode *listnode; + struct pim_vxlan_sg *vxlan_sg; + struct pim_rpf *rpg; + + /* + * No vxlan here, move along, nothing to see + */ + if (!vxlan_info.work_list) + return; + + for (listnode = vxlan_info.work_list->head; listnode; + listnode = listnode->next) { + vxlan_sg = listgetdata(listnode); + + rpg = RP(pim, vxlan_sg->up->sg.grp); + + /* + * If the rp is the same we should send + */ + if (rpg == rpg_changed) { + zlog_debug("VXLAN RP INFO is alive sending"); + pim_null_register_send(vxlan_sg->up); + } + } +} + +/*************************** vxlan work list ********************************** + * A work list is maintained for staggered generation of pim null register + * messages for vxlan SG entries that are in a reg_join state. + * + * A max of 500 NULL registers are generated at one shot. If paused reg + * generation continues on the next second and so on till all register + * messages have been sent out. And the process is restarted every 60s. + * + * purpose of this null register generation is to setup the SPT and maintain + * independent of the presence of overlay BUM traffic. + ****************************************************************************/ +static void pim_vxlan_do_reg_work(void) +{ + struct listnode *listnode; + int work_cnt = 0; + struct pim_vxlan_sg *vxlan_sg; + static int sec_count; + + ++sec_count; + + if (sec_count > PIM_VXLAN_NULL_REG_INTERVAL) { + sec_count = 0; + listnode = vxlan_info.next_work ? + vxlan_info.next_work : + vxlan_info.work_list->head; + if (PIM_DEBUG_VXLAN && listnode) + zlog_debug("vxlan SG work %s", + vxlan_info.next_work ? "continues" : "starts"); + } else { + listnode = vxlan_info.next_work; + } + + for (; listnode; listnode = listnode->next) { + vxlan_sg = (struct pim_vxlan_sg *)listnode->data; + + if (vxlan_sg->up && (vxlan_sg->up->reg_state == PIM_REG_JOIN)) { + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s periodic NULL register", + vxlan_sg->sg_str); + + /* + * If we are on the work queue *and* the rpf + * has been lost on the vxlan_sg->up let's + * make sure that we don't send it. + */ + if (vxlan_sg->up->rpf.source_nexthop.interface) { + pim_null_register_send(vxlan_sg->up); + ++work_cnt; + } + } + + if (work_cnt > vxlan_info.max_work_cnt) { + vxlan_info.next_work = listnode->next; + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %d work items proc and pause", + work_cnt); + return; + } + } + + if (work_cnt) { + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %d work items proc", work_cnt); + } + vxlan_info.next_work = NULL; +} + +/* Staggered work related info is initialized when the first work comes + * along + */ +static void pim_vxlan_init_work(void) +{ + if (vxlan_info.flags & PIM_VXLANF_WORK_INITED) + return; + + vxlan_info.max_work_cnt = PIM_VXLAN_WORK_MAX; + vxlan_info.flags |= PIM_VXLANF_WORK_INITED; + vxlan_info.work_list = list_new(); + pim_vxlan_work_timer_setup(true/* start */); +} + +static void pim_vxlan_add_work(struct pim_vxlan_sg *vxlan_sg) +{ + if (vxlan_sg->flags & PIM_VXLAN_SGF_DEL_IN_PROG) { + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s skip work list; del-in-prog", + vxlan_sg->sg_str); + return; + } + + pim_vxlan_init_work(); + + /* already a part of the work list */ + if (vxlan_sg->work_node) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s work list add", + vxlan_sg->sg_str); + vxlan_sg->work_node = listnode_add(vxlan_info.work_list, vxlan_sg); + /* XXX: adjust max_work_cnt if needed */ +} + +static void pim_vxlan_del_work(struct pim_vxlan_sg *vxlan_sg) +{ + if (!vxlan_sg->work_node) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s work list del", + vxlan_sg->sg_str); + + if (vxlan_sg->work_node == vxlan_info.next_work) + vxlan_info.next_work = vxlan_sg->work_node->next; + + list_delete_node(vxlan_info.work_list, vxlan_sg->work_node); + vxlan_sg->work_node = NULL; +} + +void pim_vxlan_update_sg_reg_state(struct pim_instance *pim, + struct pim_upstream *up, bool reg_join) +{ + struct pim_vxlan_sg *vxlan_sg; + + vxlan_sg = pim_vxlan_sg_find(pim, &up->sg); + if (!vxlan_sg) + return; + + /* add the vxlan sg entry to a work list for periodic reg joins. + * the entry will stay in the list as long as the register state is + * PIM_REG_JOIN + */ + if (reg_join) + pim_vxlan_add_work(vxlan_sg); + else + pim_vxlan_del_work(vxlan_sg); +} + +static void pim_vxlan_work_timer_cb(struct event *t) +{ + pim_vxlan_do_reg_work(); + pim_vxlan_work_timer_setup(true /* start */); +} + +/* global 1second timer used for periodic processing */ +static void pim_vxlan_work_timer_setup(bool start) +{ + EVENT_OFF(vxlan_info.work_timer); + if (start) + event_add_timer(router->master, pim_vxlan_work_timer_cb, NULL, + PIM_VXLAN_WORK_TIME, &vxlan_info.work_timer); +} + +/**************************** vxlan origination mroutes *********************** + * For every (local-vtep-ip, bum-mcast-grp) registered by evpn an origination + * mroute is setup by pimd. The purpose of this mroute is to forward vxlan + * encapsulated BUM (broadcast, unknown-unicast and unknown-multicast packets + * over the underlay.) + * + * Sample mroute (single VTEP): + * (27.0.0.7, 239.1.1.100) Iif: lo Oifs: uplink-1 + * + * Sample mroute (anycast VTEP): + * (36.0.0.9, 239.1.1.100) Iif: peerlink-3.4094\ + * Oifs: peerlink-3.4094 uplink-1 + ***************************************************************************/ +static void pim_vxlan_orig_mr_up_del(struct pim_vxlan_sg *vxlan_sg) +{ + struct pim_upstream *up = vxlan_sg->up; + + if (!up) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s orig mroute-up del", + vxlan_sg->sg_str); + + vxlan_sg->up = NULL; + + if (up->flags & PIM_UPSTREAM_FLAG_MASK_SRC_VXLAN_ORIG) { + /* clear out all the vxlan properties */ + up->flags &= ~(PIM_UPSTREAM_FLAG_MASK_SRC_VXLAN_ORIG | + PIM_UPSTREAM_FLAG_MASK_STATIC_IIF | + PIM_UPSTREAM_FLAG_MASK_DISABLE_KAT_EXPIRY | + PIM_UPSTREAM_FLAG_MASK_FORCE_PIMREG | + PIM_UPSTREAM_FLAG_MASK_NO_PIMREG_DATA | + PIM_UPSTREAM_FLAG_MASK_ALLOW_IIF_IN_OIL); + + /* We bring things to a grinding halt by force expirying + * the kat. Doing this will also remove the reference we + * created as a "vxlan" source and delete the upstream entry + * if there are no other references. + */ + if (PIM_UPSTREAM_FLAG_TEST_SRC_STREAM(up->flags)) { + EVENT_OFF(up->t_ka_timer); + up = pim_upstream_keep_alive_timer_proc(up); + } else { + /* this is really unexpected as we force vxlan + * origination mroutes active sources but just in + * case + */ + up = pim_upstream_del(vxlan_sg->pim, up, __func__); + } + /* if there are other references register the source + * for nht + */ + if (up) { + enum pim_rpf_result r; + + r = pim_rpf_update(vxlan_sg->pim, up, NULL, __func__); + if (r == PIM_RPF_FAILURE) { + if (PIM_DEBUG_VXLAN) + zlog_debug( + "vxlan SG %s rpf_update failure", + vxlan_sg->sg_str); + } + } + } +} + +static void pim_vxlan_orig_mr_up_iif_update(struct pim_vxlan_sg *vxlan_sg) +{ + /* update MFC with the new IIF */ + pim_upstream_fill_static_iif(vxlan_sg->up, vxlan_sg->iif); + pim_upstream_mroute_iif_update(vxlan_sg->up->channel_oil, __func__); + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s orig mroute-up updated with iif %s", + vxlan_sg->sg_str, + vxlan_sg->iif?vxlan_sg->iif->name:"-"); + +} + +/* For every VxLAN BUM multicast group we setup a SG-up that has the following + * "forced properties" - + * 1. Directly connected on a DR interface i.e. we must act as an FHR + * 2. We prime the pump i.e. no multicast data is needed to register this + * source with the FHR. To do that we send periodic null registers if + * the SG entry is in a register-join state. We also prevent expiry of + * KAT. + * 3. As this SG is setup without data there is no need to register encapsulate + * data traffic. This encapsulation is explicitly skipped for the following + * reasons - + * a) Many levels of encapsulation are needed creating MTU disc challenges. + * Overlay BUM is encapsulated in a vxlan/UDP/IP header and then + * encapsulated again in a pim-register header. + * b) On a vxlan-aa setup both switches rx a copy of each BUM packet. if + * they both reg encapsulated traffic the RP will accept the duplicates + * as there are no RPF checks for this encapsulated data. + * a), b) can be workarounded if needed, but there is really no need because + * of (2) i.e. the pump is primed without data. + */ +static void pim_vxlan_orig_mr_up_add(struct pim_vxlan_sg *vxlan_sg) +{ + struct pim_upstream *up; + struct pim_interface *term_ifp; + int flags = 0; + struct pim_instance *pim = vxlan_sg->pim; + + if (vxlan_sg->up) { + /* nothing to do */ + return; + } + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s orig mroute-up add with iif %s", + vxlan_sg->sg_str, + vxlan_sg->iif?vxlan_sg->iif->name:"-"); + + PIM_UPSTREAM_FLAG_SET_SRC_VXLAN_ORIG(flags); + /* pin the IIF to lo or peerlink-subinterface and disable NHT */ + PIM_UPSTREAM_FLAG_SET_STATIC_IIF(flags); + /* Fake traffic by setting SRC_STREAM and starting KAT */ + /* We intentionally skip updating ref count for SRC_STREAM/FHR. + * Setting SRC_VXLAN should have already created a reference + * preventing the entry from being deleted + */ + PIM_UPSTREAM_FLAG_SET_FHR(flags); + PIM_UPSTREAM_FLAG_SET_SRC_STREAM(flags); + /* Force pimreg even if non-DR. This is needed on a MLAG setup for + * VxLAN AA + */ + PIM_UPSTREAM_FLAG_SET_FORCE_PIMREG(flags); + /* prevent KAT expiry. we want the MDT setup even if there is no BUM + * traffic + */ + PIM_UPSTREAM_FLAG_SET_DISABLE_KAT_EXPIRY(flags); + /* SPT for vxlan BUM groups is primed and maintained via NULL + * registers so there is no need to reg-encapsulate + * vxlan-encapsulated overlay data traffic + */ + PIM_UPSTREAM_FLAG_SET_NO_PIMREG_DATA(flags); + /* On a MLAG setup we force a copy to the MLAG peer while also + * accepting traffic from the peer. To do this we set peerlink-rif as + * the IIF and also add it to the OIL + */ + PIM_UPSTREAM_FLAG_SET_ALLOW_IIF_IN_OIL(flags); + + /* XXX: todo: defer pim_upstream add if pim is not enabled on the iif */ + up = pim_upstream_find(vxlan_sg->pim, &vxlan_sg->sg); + if (up) { + /* if the iif is set to something other than the vxlan_sg->iif + * we must dereg the old nexthop and force to new "static" + * iif + */ + if (!PIM_UPSTREAM_FLAG_TEST_STATIC_IIF(up->flags)) { + pim_delete_tracked_nexthop(vxlan_sg->pim, + up->upstream_addr, up, NULL); + } + /* We are acting FHR; clear out use_rpt setting if any */ + pim_upstream_update_use_rpt(up, false /*update_mroute*/); + pim_upstream_ref(up, flags, __func__); + vxlan_sg->up = up; + term_ifp = pim_vxlan_get_term_ifp(pim); + /* mute termination device on origination mroutes */ + if (term_ifp) + pim_channel_update_oif_mute(up->channel_oil, + term_ifp); + pim_vxlan_orig_mr_up_iif_update(vxlan_sg); + /* mute pimreg on origination mroutes */ + if (pim->regiface) + pim_channel_update_oif_mute(up->channel_oil, + pim->regiface->info); + } else { + up = pim_upstream_add(vxlan_sg->pim, &vxlan_sg->sg, + vxlan_sg->iif, flags, __func__, NULL); + vxlan_sg->up = up; + } + + if (!up) { + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s orig mroute-up add failed", + vxlan_sg->sg_str); + return; + } + + pim_upstream_keep_alive_timer_start(up, vxlan_sg->pim->keep_alive_time); + + /* register the source with the RP */ + switch (up->reg_state) { + + case PIM_REG_NOINFO: + pim_register_join(up); + pim_null_register_send(up); + break; + + case PIM_REG_JOIN: + /* if the pim upstream entry is already in reg-join state + * send null_register right away and add to the register + * worklist + */ + pim_null_register_send(up); + pim_vxlan_update_sg_reg_state(pim, up, true); + break; + + case PIM_REG_JOIN_PENDING: + case PIM_REG_PRUNE: + break; + } + + /* update the inherited OIL */ + pim_upstream_inherited_olist(vxlan_sg->pim, up); + if (!up->channel_oil->installed) + pim_upstream_mroute_add(up->channel_oil, __func__); +} + +static void pim_vxlan_orig_mr_oif_add(struct pim_vxlan_sg *vxlan_sg) +{ + if (!vxlan_sg->up || !vxlan_sg->orig_oif) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s oif %s add", + vxlan_sg->sg_str, vxlan_sg->orig_oif->name); + + vxlan_sg->flags |= PIM_VXLAN_SGF_OIF_INSTALLED; + pim_channel_add_oif(vxlan_sg->up->channel_oil, + vxlan_sg->orig_oif, PIM_OIF_FLAG_PROTO_VXLAN, + __func__); +} + +static void pim_vxlan_orig_mr_oif_del(struct pim_vxlan_sg *vxlan_sg) +{ + struct interface *orig_oif; + + orig_oif = vxlan_sg->orig_oif; + vxlan_sg->orig_oif = NULL; + + if (!(vxlan_sg->flags & PIM_VXLAN_SGF_OIF_INSTALLED)) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s oif %s del", + vxlan_sg->sg_str, orig_oif->name); + + vxlan_sg->flags &= ~PIM_VXLAN_SGF_OIF_INSTALLED; + pim_channel_del_oif(vxlan_sg->up->channel_oil, + orig_oif, PIM_OIF_FLAG_PROTO_VXLAN, __func__); +} + +static inline struct interface *pim_vxlan_orig_mr_oif_get( + struct pim_instance *pim) +{ + return (vxlan_mlag.flags & PIM_VXLAN_MLAGF_ENABLED) ? + pim->vxlan.peerlink_rif : NULL; +} + +/* Single VTEPs: IIF for the vxlan-origination-mroutes is lo or vrf-dev (if + * the mroute is in a non-default vrf). + * Anycast VTEPs: IIF is the MLAG ISL/peerlink. + */ +static inline struct interface *pim_vxlan_orig_mr_iif_get( + struct pim_instance *pim) +{ + return ((vxlan_mlag.flags & PIM_VXLAN_MLAGF_ENABLED) && + pim->vxlan.peerlink_rif) ? + pim->vxlan.peerlink_rif : pim->vxlan.default_iif; +} + +static bool pim_vxlan_orig_mr_add_is_ok(struct pim_vxlan_sg *vxlan_sg) +{ + struct pim_interface *pim_ifp; + + vxlan_sg->iif = pim_vxlan_orig_mr_iif_get(vxlan_sg->pim); + if (!vxlan_sg->iif) + return false; + + pim_ifp = (struct pim_interface *)vxlan_sg->iif->info; + if (!pim_ifp || (pim_ifp->mroute_vif_index < 0)) + return false; + + return true; +} + +static void pim_vxlan_orig_mr_install(struct pim_vxlan_sg *vxlan_sg) +{ + pim_vxlan_orig_mr_up_add(vxlan_sg); + + vxlan_sg->orig_oif = pim_vxlan_orig_mr_oif_get(vxlan_sg->pim); + pim_vxlan_orig_mr_oif_add(vxlan_sg); +} + +static void pim_vxlan_orig_mr_add(struct pim_vxlan_sg *vxlan_sg) +{ + if (!pim_vxlan_orig_mr_add_is_ok(vxlan_sg)) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s orig-mr add", vxlan_sg->sg_str); + + pim_vxlan_orig_mr_install(vxlan_sg); +} + +static void pim_vxlan_orig_mr_del(struct pim_vxlan_sg *vxlan_sg) +{ + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s orig-mr del", vxlan_sg->sg_str); + + pim_vxlan_orig_mr_oif_del(vxlan_sg); + pim_vxlan_orig_mr_up_del(vxlan_sg); +} + +static void pim_vxlan_orig_mr_iif_update(struct hash_bucket *bucket, void *arg) +{ + struct interface *ifp; + struct pim_vxlan_sg *vxlan_sg = (struct pim_vxlan_sg *)bucket->data; + struct interface *old_iif = vxlan_sg->iif; + + if (!pim_vxlan_is_orig_mroute(vxlan_sg)) + return; + + ifp = pim_vxlan_orig_mr_iif_get(vxlan_sg->pim); + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s iif changed from %s to %s", + vxlan_sg->sg_str, + old_iif ? old_iif->name : "-", + ifp ? ifp->name : "-"); + + if (pim_vxlan_orig_mr_add_is_ok(vxlan_sg)) { + if (vxlan_sg->up) { + /* upstream exists but iif changed */ + pim_vxlan_orig_mr_up_iif_update(vxlan_sg); + } else { + /* install mroute */ + pim_vxlan_orig_mr_install(vxlan_sg); + } + } else { + pim_vxlan_orig_mr_del(vxlan_sg); + } +} + +/**************************** vxlan termination mroutes *********************** + * For every bum-mcast-grp registered by evpn a *G termination + * mroute is setup by pimd. The purpose of this mroute is to pull down vxlan + * packets with the bum-mcast-grp dip from the underlay and terminate the + * tunnel. This is done by including the vxlan termination device (ipmr-lo) in + * its OIL. The vxlan de-capsulated packets are subject to subsequent overlay + * bridging. + * + * Sample mroute: + * (0.0.0.0, 239.1.1.100) Iif: uplink-1 Oifs: ipmr-lo, uplink-1 + *****************************************************************************/ +struct pim_interface *pim_vxlan_get_term_ifp(struct pim_instance *pim) +{ + return pim->vxlan.term_if ? + (struct pim_interface *)pim->vxlan.term_if->info : NULL; +} + +static void pim_vxlan_term_mr_oif_add(struct pim_vxlan_sg *vxlan_sg) +{ + if (vxlan_sg->flags & PIM_VXLAN_SGF_OIF_INSTALLED) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s term-oif %s add", + vxlan_sg->sg_str, vxlan_sg->term_oif->name); + + if (pim_ifchannel_local_membership_add(vxlan_sg->term_oif, + &vxlan_sg->sg, true /*is_vxlan */)) { + vxlan_sg->flags |= PIM_VXLAN_SGF_OIF_INSTALLED; + /* update the inherited OIL */ + /* XXX - I don't see the inherited OIL updated when a local + * member is added. And that probably needs to be fixed. Till + * that happens we do a force update on the inherited OIL + * here. + */ + pim_upstream_inherited_olist(vxlan_sg->pim, vxlan_sg->up); + } else { + zlog_warn("vxlan SG %s term-oif %s add failed", + vxlan_sg->sg_str, vxlan_sg->term_oif->name); + } +} + +static void pim_vxlan_term_mr_oif_del(struct pim_vxlan_sg *vxlan_sg) +{ + if (!(vxlan_sg->flags & PIM_VXLAN_SGF_OIF_INSTALLED)) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s oif %s del", + vxlan_sg->sg_str, vxlan_sg->term_oif->name); + + vxlan_sg->flags &= ~PIM_VXLAN_SGF_OIF_INSTALLED; + pim_ifchannel_local_membership_del(vxlan_sg->term_oif, &vxlan_sg->sg); + /* update the inherited OIL */ + /* XXX - I don't see the inherited OIL updated when a local member + * is deleted. And that probably needs to be fixed. Till that happens + * we do a force update on the inherited OIL here. + */ + pim_upstream_inherited_olist(vxlan_sg->pim, vxlan_sg->up); +} + +static void pim_vxlan_update_sg_entry_mlag(struct pim_instance *pim, + struct pim_upstream *up, bool inherit) +{ + bool is_df = true; + + if (inherit && up->parent && + PIM_UPSTREAM_FLAG_TEST_MLAG_VXLAN(up->parent->flags) && + PIM_UPSTREAM_FLAG_TEST_MLAG_NON_DF(up->parent->flags)) + is_df = false; + + pim_mlag_up_df_role_update(pim, up, is_df, "inherit_xg_df"); +} + +/* We run MLAG DF election only on mroutes that have the termination + * device ipmr-lo in the immediate OIL. This is only (*, G) entries at the + * moment. For (S, G) entries that (with ipmr-lo in the inherited OIL) we + * inherit the DF role from the (*, G) entry. + */ +void pim_vxlan_inherit_mlag_flags(struct pim_instance *pim, + struct pim_upstream *up, bool inherit) +{ + struct listnode *listnode; + struct pim_upstream *child; + + for (ALL_LIST_ELEMENTS_RO(up->sources, listnode, + child)) { + pim_vxlan_update_sg_entry_mlag(pim, + child, true /* inherit */); + } +} + +static void pim_vxlan_term_mr_up_add(struct pim_vxlan_sg *vxlan_sg) +{ + struct pim_upstream *up; + int flags = 0; + + if (vxlan_sg->up) { + /* nothing to do */ + return; + } + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s term mroute-up add", + vxlan_sg->sg_str); + + PIM_UPSTREAM_FLAG_SET_SRC_VXLAN_TERM(flags); + /* enable MLAG designated-forwarder election on termination mroutes */ + PIM_UPSTREAM_FLAG_SET_MLAG_VXLAN(flags); + + up = pim_upstream_add(vxlan_sg->pim, &vxlan_sg->sg, NULL /* iif */, + flags, __func__, NULL); + vxlan_sg->up = up; + + if (!up) { + zlog_warn("vxlan SG %s term mroute-up add failed", + vxlan_sg->sg_str); + return; + } + + /* update existing SG entries with the parent's MLAG flag */ + pim_vxlan_inherit_mlag_flags(vxlan_sg->pim, up, true /*enable*/); +} + +static void pim_vxlan_term_mr_up_del(struct pim_vxlan_sg *vxlan_sg) +{ + struct pim_upstream *up = vxlan_sg->up; + + if (!up) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s term mroute-up del", + vxlan_sg->sg_str); + vxlan_sg->up = NULL; + if (up->flags & PIM_UPSTREAM_FLAG_MASK_SRC_VXLAN_TERM) { + /* update SG entries that are inheriting from this XG entry */ + pim_vxlan_inherit_mlag_flags(vxlan_sg->pim, up, + false /*enable*/); + /* clear out all the vxlan related flags */ + up->flags &= ~(PIM_UPSTREAM_FLAG_MASK_SRC_VXLAN_TERM | + PIM_UPSTREAM_FLAG_MASK_MLAG_VXLAN); + pim_mlag_up_local_del(vxlan_sg->pim, up); + pim_upstream_del(vxlan_sg->pim, up, __func__); + } +} + +static void pim_vxlan_term_mr_add(struct pim_vxlan_sg *vxlan_sg) +{ + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s term mroute add", vxlan_sg->sg_str); + + vxlan_sg->term_oif = vxlan_sg->pim->vxlan.term_if; + if (!vxlan_sg->term_oif) + /* defer termination mroute till we have a termination device */ + return; + + pim_vxlan_term_mr_up_add(vxlan_sg); + /* set up local membership for the term-oif */ + pim_vxlan_term_mr_oif_add(vxlan_sg); +} + +static void pim_vxlan_term_mr_del(struct pim_vxlan_sg *vxlan_sg) +{ + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s term mroute del", vxlan_sg->sg_str); + + /* remove local membership associated with the term oif */ + pim_vxlan_term_mr_oif_del(vxlan_sg); + /* remove references to the upstream entry */ + pim_vxlan_term_mr_up_del(vxlan_sg); +} + +/************************** vxlan SG cache management ************************/ +static unsigned int pim_vxlan_sg_hash_key_make(const void *p) +{ + const struct pim_vxlan_sg *vxlan_sg = p; + + return pim_sgaddr_hash(vxlan_sg->sg, 0); +} + +static bool pim_vxlan_sg_hash_eq(const void *p1, const void *p2) +{ + const struct pim_vxlan_sg *sg1 = p1; + const struct pim_vxlan_sg *sg2 = p2; + + return !pim_sgaddr_cmp(sg1->sg, sg2->sg); +} + +static struct pim_vxlan_sg *pim_vxlan_sg_new(struct pim_instance *pim, + pim_sgaddr *sg) +{ + struct pim_vxlan_sg *vxlan_sg; + + vxlan_sg = XCALLOC(MTYPE_PIM_VXLAN_SG, sizeof(*vxlan_sg)); + + vxlan_sg->pim = pim; + vxlan_sg->sg = *sg; + snprintfrr(vxlan_sg->sg_str, sizeof(vxlan_sg->sg_str), "%pSG", sg); + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s alloc", vxlan_sg->sg_str); + + vxlan_sg = hash_get(pim->vxlan.sg_hash, vxlan_sg, hash_alloc_intern); + + /* we register with the MLAG daemon in the first VxLAN SG and never + * de-register during that life of the pimd + */ + if (pim->vxlan.sg_hash->count == 1) { + vxlan_mlag.flags |= PIM_VXLAN_MLAGF_DO_REG; + pim_mlag_register(); + } + + return vxlan_sg; +} + +struct pim_vxlan_sg *pim_vxlan_sg_find(struct pim_instance *pim, pim_sgaddr *sg) +{ + struct pim_vxlan_sg lookup; + + lookup.sg = *sg; + return hash_lookup(pim->vxlan.sg_hash, &lookup); +} + +struct pim_vxlan_sg *pim_vxlan_sg_add(struct pim_instance *pim, pim_sgaddr *sg) +{ + struct pim_vxlan_sg *vxlan_sg; + + vxlan_sg = pim_vxlan_sg_find(pim, sg); + if (vxlan_sg) + return vxlan_sg; + + vxlan_sg = pim_vxlan_sg_new(pim, sg); + + if (pim_vxlan_is_orig_mroute(vxlan_sg)) + pim_vxlan_orig_mr_add(vxlan_sg); + else + pim_vxlan_term_mr_add(vxlan_sg); + + return vxlan_sg; +} + +static void pim_vxlan_sg_del_item(struct pim_vxlan_sg *vxlan_sg) +{ + vxlan_sg->flags |= PIM_VXLAN_SGF_DEL_IN_PROG; + + pim_vxlan_del_work(vxlan_sg); + + if (pim_vxlan_is_orig_mroute(vxlan_sg)) + pim_vxlan_orig_mr_del(vxlan_sg); + else + pim_vxlan_term_mr_del(vxlan_sg); + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s free", vxlan_sg->sg_str); + + XFREE(MTYPE_PIM_VXLAN_SG, vxlan_sg); +} + +void pim_vxlan_sg_del(struct pim_instance *pim, pim_sgaddr *sg) +{ + struct pim_vxlan_sg *vxlan_sg; + + vxlan_sg = pim_vxlan_sg_find(pim, sg); + if (!vxlan_sg) + return; + + hash_release(pim->vxlan.sg_hash, vxlan_sg); + pim_vxlan_sg_del_item(vxlan_sg); +} + +/******************************* MLAG handling *******************************/ +bool pim_vxlan_do_mlag_reg(void) +{ + return (vxlan_mlag.flags & PIM_VXLAN_MLAGF_DO_REG); +} + +/* The peerlink sub-interface is added as an OIF to the origination-mroute. + * This is done to send a copy of the multicast-vxlan encapsulated traffic + * to the MLAG peer which may mroute it over the underlay if there are any + * interested receivers. + */ +static void pim_vxlan_sg_peerlink_oif_update(struct hash_bucket *bucket, + void *arg) +{ + struct interface *new_oif = (struct interface *)arg; + struct pim_vxlan_sg *vxlan_sg = (struct pim_vxlan_sg *)bucket->data; + + if (!pim_vxlan_is_orig_mroute(vxlan_sg)) + return; + + if (vxlan_sg->orig_oif == new_oif) + return; + + pim_vxlan_orig_mr_oif_del(vxlan_sg); + + vxlan_sg->orig_oif = new_oif; + pim_vxlan_orig_mr_oif_add(vxlan_sg); +} + +/* In the case of anycast VTEPs the VTEP-PIP must be used as the + * register source. + */ +bool pim_vxlan_get_register_src(struct pim_instance *pim, + struct pim_upstream *up, struct in_addr *src_p) +{ + if (!(vxlan_mlag.flags & PIM_VXLAN_MLAGF_ENABLED)) + return true; + + /* if address is not available suppress the pim-register */ + if (vxlan_mlag.reg_addr.s_addr == INADDR_ANY) + return false; + + *src_p = vxlan_mlag.reg_addr; + return true; +} + +void pim_vxlan_mlag_update(bool enable, bool peer_state, uint32_t role, + struct interface *peerlink_rif, + struct in_addr *reg_addr) +{ + struct pim_instance *pim; + char addr_buf[INET_ADDRSTRLEN]; + struct pim_interface *pim_ifp = NULL; + + if (PIM_DEBUG_VXLAN) { + inet_ntop(AF_INET, reg_addr, + addr_buf, INET_ADDRSTRLEN); + zlog_debug("vxlan MLAG update %s state %s role %d rif %s addr %s", + enable ? "enable" : "disable", + peer_state ? "up" : "down", + role, + peerlink_rif ? peerlink_rif->name : "-", + addr_buf); + } + + /* XXX: for now vxlan termination is only possible in the default VRF + * when that changes this will need to change to iterate all VRFs + */ + pim = pim_get_pim_instance(VRF_DEFAULT); + + if (!pim) { + if (PIM_DEBUG_VXLAN) + zlog_debug("%s: Unable to find pim instance", __func__); + return; + } + + if (enable) + vxlan_mlag.flags |= PIM_VXLAN_MLAGF_ENABLED; + else + vxlan_mlag.flags &= ~PIM_VXLAN_MLAGF_ENABLED; + + if (vxlan_mlag.peerlink_rif != peerlink_rif) + vxlan_mlag.peerlink_rif = peerlink_rif; + + vxlan_mlag.reg_addr = *reg_addr; + vxlan_mlag.peer_state = peer_state; + vxlan_mlag.role = role; + + /* process changes */ + if (vxlan_mlag.peerlink_rif) + pim_ifp = (struct pim_interface *)vxlan_mlag.peerlink_rif->info; + if ((vxlan_mlag.flags & PIM_VXLAN_MLAGF_ENABLED) && + pim_ifp && (pim_ifp->mroute_vif_index > 0)) + pim_vxlan_set_peerlink_rif(pim, peerlink_rif); + else + pim_vxlan_set_peerlink_rif(pim, NULL); +} + +/****************************** misc callbacks *******************************/ +static void pim_vxlan_set_default_iif(struct pim_instance *pim, + struct interface *ifp) +{ + struct interface *old_iif; + + if (pim->vxlan.default_iif == ifp) + return; + + old_iif = pim->vxlan.default_iif; + if (PIM_DEBUG_VXLAN) + zlog_debug("%s: vxlan default iif changed from %s to %s", + __func__, old_iif ? old_iif->name : "-", + ifp ? ifp->name : "-"); + + old_iif = pim_vxlan_orig_mr_iif_get(pim); + pim->vxlan.default_iif = ifp; + ifp = pim_vxlan_orig_mr_iif_get(pim); + if (old_iif == ifp) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("%s: vxlan orig iif changed from %s to %s", __func__, + old_iif ? old_iif->name : "-", + ifp ? ifp->name : "-"); + + /* add/del upstream entries for the existing vxlan SG when the + * interface becomes available + */ + if (pim->vxlan.sg_hash) + hash_iterate(pim->vxlan.sg_hash, + pim_vxlan_orig_mr_iif_update, NULL); +} + +static void pim_vxlan_up_cost_update(struct pim_instance *pim, + struct pim_upstream *up, + struct interface *old_peerlink_rif) +{ + if (!PIM_UPSTREAM_FLAG_TEST_MLAG_VXLAN(up->flags)) + return; + + if (up->rpf.source_nexthop.interface && + ((up->rpf.source_nexthop.interface == + pim->vxlan.peerlink_rif) || + (up->rpf.source_nexthop.interface == + old_peerlink_rif))) { + if (PIM_DEBUG_VXLAN) + zlog_debug("RPF cost adjust for %s on peerlink-rif (old: %s, new: %s) change", + up->sg_str, + old_peerlink_rif ? + old_peerlink_rif->name : "-", + pim->vxlan.peerlink_rif ? + pim->vxlan.peerlink_rif->name : "-"); + pim_mlag_up_local_add(pim, up); + } +} + +static void pim_vxlan_term_mr_cost_update(struct hash_bucket *bucket, void *arg) +{ + struct interface *old_peerlink_rif = (struct interface *)arg; + struct pim_vxlan_sg *vxlan_sg = (struct pim_vxlan_sg *)bucket->data; + struct pim_upstream *up; + struct listnode *listnode; + struct pim_upstream *child; + + if (pim_vxlan_is_orig_mroute(vxlan_sg)) + return; + + /* Lookup all XG and SG entries with RPF-interface peerlink_rif */ + up = vxlan_sg->up; + if (!up) + return; + + pim_vxlan_up_cost_update(vxlan_sg->pim, up, + old_peerlink_rif); + + for (ALL_LIST_ELEMENTS_RO(up->sources, listnode, + child)) + pim_vxlan_up_cost_update(vxlan_sg->pim, child, + old_peerlink_rif); +} + +static void pim_vxlan_sg_peerlink_rif_update(struct hash_bucket *bucket, + void *arg) +{ + pim_vxlan_orig_mr_iif_update(bucket, NULL); + pim_vxlan_term_mr_cost_update(bucket, arg); +} + +static void pim_vxlan_set_peerlink_rif(struct pim_instance *pim, + struct interface *ifp) +{ + struct interface *old_iif; + struct interface *new_iif; + struct interface *old_oif; + struct interface *new_oif; + + if (pim->vxlan.peerlink_rif == ifp) + return; + + old_iif = pim->vxlan.peerlink_rif; + if (PIM_DEBUG_VXLAN) + zlog_debug("%s: vxlan peerlink_rif changed from %s to %s", + __func__, old_iif ? old_iif->name : "-", + ifp ? ifp->name : "-"); + + old_iif = pim_vxlan_orig_mr_iif_get(pim); + old_oif = pim_vxlan_orig_mr_oif_get(pim); + pim->vxlan.peerlink_rif = ifp; + + new_iif = pim_vxlan_orig_mr_iif_get(pim); + if (old_iif != new_iif) { + if (PIM_DEBUG_VXLAN) + zlog_debug("%s: vxlan orig iif changed from %s to %s", + __func__, old_iif ? old_iif->name : "-", + new_iif ? new_iif->name : "-"); + + /* add/del upstream entries for the existing vxlan SG when the + * interface becomes available + */ + if (pim->vxlan.sg_hash) + hash_iterate(pim->vxlan.sg_hash, + pim_vxlan_sg_peerlink_rif_update, + old_iif); + } + + new_oif = pim_vxlan_orig_mr_oif_get(pim); + if (old_oif != new_oif) { + if (PIM_DEBUG_VXLAN) + zlog_debug("%s: vxlan orig oif changed from %s to %s", + __func__, old_oif ? old_oif->name : "-", + new_oif ? new_oif->name : "-"); + if (pim->vxlan.sg_hash) + hash_iterate(pim->vxlan.sg_hash, + pim_vxlan_sg_peerlink_oif_update, + new_oif); + } +} + +static void pim_vxlan_term_mr_oif_update(struct hash_bucket *bucket, void *arg) +{ + struct interface *ifp = (struct interface *)arg; + struct pim_vxlan_sg *vxlan_sg = (struct pim_vxlan_sg *)bucket->data; + + if (pim_vxlan_is_orig_mroute(vxlan_sg)) + return; + + if (vxlan_sg->term_oif == ifp) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan SG %s term oif changed from %s to %s", + vxlan_sg->sg_str, + vxlan_sg->term_oif ? vxlan_sg->term_oif->name : "-", + ifp ? ifp->name : "-"); + + pim_vxlan_term_mr_del(vxlan_sg); + vxlan_sg->term_oif = ifp; + pim_vxlan_term_mr_add(vxlan_sg); +} + +static void pim_vxlan_term_oif_update(struct pim_instance *pim, + struct interface *ifp) +{ + if (pim->vxlan.term_if == ifp) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan term oif changed from %s to %s", + pim->vxlan.term_if ? pim->vxlan.term_if->name : "-", + ifp ? ifp->name : "-"); + + pim->vxlan.term_if = ifp; + if (pim->vxlan.sg_hash) + hash_iterate(pim->vxlan.sg_hash, + pim_vxlan_term_mr_oif_update, ifp); +} + +void pim_vxlan_add_vif(struct interface *ifp) +{ + struct pim_interface *pim_ifp = ifp->info; + struct pim_instance *pim = pim_ifp->pim; + + if (pim->vrf->vrf_id != VRF_DEFAULT) + return; + + if (if_is_loopback(ifp)) + pim_vxlan_set_default_iif(pim, ifp); + + if (vxlan_mlag.flags & PIM_VXLAN_MLAGF_ENABLED && + (ifp == vxlan_mlag.peerlink_rif)) + pim_vxlan_set_peerlink_rif(pim, ifp); + + if (pim->vxlan.term_if_cfg == ifp) + pim_vxlan_term_oif_update(pim, ifp); +} + +void pim_vxlan_del_vif(struct interface *ifp) +{ + struct pim_interface *pim_ifp = ifp->info; + struct pim_instance *pim = pim_ifp->pim; + + if (pim->vrf->vrf_id != VRF_DEFAULT) + return; + + if (pim->vxlan.default_iif == ifp) + pim_vxlan_set_default_iif(pim, NULL); + + if (pim->vxlan.peerlink_rif == ifp) + pim_vxlan_set_peerlink_rif(pim, NULL); + + if (pim->vxlan.term_if == ifp) + pim_vxlan_term_oif_update(pim, NULL); +} + +/* enable pim implicitly on the termination device add */ +void pim_vxlan_add_term_dev(struct pim_instance *pim, + struct interface *ifp) +{ + struct pim_interface *pim_ifp; + + if (pim->vxlan.term_if_cfg == ifp) + return; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan term oif cfg changed from %s to %s", + pim->vxlan.term_if_cfg ? + pim->vxlan.term_if_cfg->name : "-", + ifp->name); + + pim->vxlan.term_if_cfg = ifp; + + /* enable pim on the term ifp */ + pim_ifp = (struct pim_interface *)ifp->info; + if (pim_ifp) { + pim_ifp->pim_enable = true; + /* ifp is already oper up; activate it as a term dev */ + if (pim_ifp->mroute_vif_index >= 0) + pim_vxlan_term_oif_update(pim, ifp); + } else { + /* ensure that pimreg exists before using the newly created + * vxlan termination device + */ + pim_if_create_pimreg(pim); + (void)pim_if_new(ifp, false /*igmp*/, true /*pim*/, + false /*pimreg*/, true /*vxlan_term*/); + } +} + +/* disable pim implicitly, if needed, on the termination device deletion */ +void pim_vxlan_del_term_dev(struct pim_instance *pim) +{ + struct interface *ifp = pim->vxlan.term_if_cfg; + struct pim_interface *pim_ifp; + + if (PIM_DEBUG_VXLAN) + zlog_debug("vxlan term oif cfg changed from %s to -", + ifp->name); + + pim->vxlan.term_if_cfg = NULL; + + pim_ifp = (struct pim_interface *)ifp->info; + if (pim_ifp) { + pim_ifp->pim_enable = false; + if (!pim_ifp->gm_enable) + pim_if_delete(ifp); + } +} + +void pim_vxlan_init(struct pim_instance *pim) +{ + char hash_name[64]; + + snprintf(hash_name, sizeof(hash_name), + "PIM %s vxlan SG hash", pim->vrf->name); + pim->vxlan.sg_hash = hash_create(pim_vxlan_sg_hash_key_make, + pim_vxlan_sg_hash_eq, hash_name); +} + +void pim_vxlan_exit(struct pim_instance *pim) +{ + hash_clean_and_free(&pim->vxlan.sg_hash, + (void (*)(void *))pim_vxlan_sg_del_item); +} + +void pim_vxlan_terminate(void) +{ + pim_vxlan_work_timer_setup(false); +} |