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);
/*************************** vxlan work list **********************************
* A work list is maintained for staggered generation of pim null register
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 */);
+ pim_vxlan_work_timer_setup(true/* start */);
}
static void pim_vxlan_add_work(struct pim_vxlan_sg *vxlan_sg)
* for nht
*/
if (up)
- pim_rpf_update(vxlan_sg->pim, up, NULL, 1 /* is_new */);
+ pim_rpf_update(vxlan_sg->pim, up, NULL);
}
}
nht_p.prefixlen = IPV4_MAX_BITLEN;
nht_p.u.prefix4 = up->upstream_addr;
pim_delete_tracked_nexthop(vxlan_sg->pim,
- &nht_p, up, NULL);
+ &nht_p, up, NULL, false);
}
pim_upstream_ref(up, flags, __PRETTY_FUNCTION__);
vxlan_sg->up = up;
pim_vxlan_orig_mr_up_del(vxlan_sg);
}
+static void pim_vxlan_orig_mr_iif_update(struct hash_backet *backet, void *arg)
+{
+ struct interface *ifp = (struct interface *)arg;
+ struct pim_vxlan_sg *vxlan_sg = (struct pim_vxlan_sg *)backet->data;
+ struct interface *old_iif = vxlan_sg->iif;
+
+ if (!pim_vxlan_is_orig_mroute(vxlan_sg))
+ return;
+
+ 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)) {
+ vxlan_sg->flags |= PIM_VXLAN_SGF_OIF_INSTALLED;
+ } 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);
+}
+
+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,
+ __PRETTY_FUNCTION__, NULL);
+ vxlan_sg->up = up;
+
+ if (!up) {
+ zlog_warn("vxlan SG %s term mroute-up add failed",
+ vxlan_sg->sg_str);
+ }
+}
+
+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) {
+ /* clear out all the vxlan related flags */
+ up->flags &= ~(PIM_UPSTREAM_FLAG_MASK_SRC_VXLAN_TERM |
+ PIM_UPSTREAM_FLAG_MASK_MLAG_VXLAN);
+
+ pim_upstream_del(vxlan_sg->pim, up,
+ __PRETTY_FUNCTION__);
+ }
+}
+
+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(void *p)
+static unsigned int pim_vxlan_sg_hash_key_make(const void *p)
{
- struct pim_vxlan_sg *vxlan_sg = p;
+ const struct pim_vxlan_sg *vxlan_sg = p;
return (jhash_2words(vxlan_sg->sg.src.s_addr,
vxlan_sg->sg.grp.s_addr, 0));
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;
}
if (pim_vxlan_is_orig_mroute(vxlan_sg))
pim_vxlan_orig_mr_del(vxlan_sg);
+ else
+ pim_vxlan_term_mr_del(vxlan_sg);
hash_release(vxlan_sg->pim->vxlan.sg_hash, vxlan_sg);
XFREE(MTYPE_PIM_VXLAN_SG, vxlan_sg);
}
+/******************************* MLAG handling *******************************/
+/* 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_update(struct hash_backet *backet, void *arg)
+{
+ struct interface *new_oif = (struct interface *)arg;
+ struct pim_vxlan_sg *vxlan_sg = (struct pim_vxlan_sg *)backet->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;
+ struct interface *old_oif;
+ struct interface *new_oif;
+ 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);
+
+ old_oif = pim_vxlan_orig_mr_oif_get(pim);
+
+ 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);
+
+ new_oif = pim_vxlan_orig_mr_oif_get(pim);
+ if (old_oif != new_oif)
+ hash_iterate(pim->vxlan.sg_hash, pim_vxlan_sg_peerlink_update,
+ new_oif);
+}
+
+/****************************** misc callbacks *******************************/
+void pim_vxlan_config_write(struct vty *vty, char *spaces, int *writes)
+{
+ char addr_buf[INET_ADDRSTRLEN];
+
+ if ((vxlan_mlag.flags & PIM_VXLAN_MLAGF_ENABLED) &&
+ vxlan_mlag.peerlink_rif) {
+
+ inet_ntop(AF_INET, &vxlan_mlag.reg_addr,
+ addr_buf, sizeof(addr_buf));
+ vty_out(vty,
+ "%sip pim mlag %s role %s state %s addr %s\n",
+ spaces,
+ vxlan_mlag.peerlink_rif->name,
+ (vxlan_mlag.role == PIM_VXLAN_MLAG_ROLE_PRIMARY) ?
+ "primary":"secondary",
+ vxlan_mlag.peer_state ? "up" : "down",
+ addr_buf);
+ *writes += 1;
+ }
+}
+
+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",
+ __PRETTY_FUNCTION__,
+ 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",
+ __PRETTY_FUNCTION__, 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, ifp);
+}
+
+static void pim_vxlan_set_peerlink_rif(struct pim_instance *pim,
+ struct interface *ifp)
+{
+ struct interface *old_iif;
+
+ 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",
+ __PRETTY_FUNCTION__, old_iif ? old_iif->name : "-",
+ ifp ? ifp->name : "-");
+
+ old_iif = pim_vxlan_orig_mr_iif_get(pim);
+ pim->vxlan.peerlink_rif = 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",
+ __PRETTY_FUNCTION__, 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, 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_id != VRF_DEFAULT)
+ return;
+
+ if (if_is_loopback_or_vrf(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);
+}
+
+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_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);
+}
+
+static void pim_vxlan_term_mr_oif_update(struct hash_backet *backet, void *arg)
+{
+ struct interface *ifp = (struct interface *)arg;
+ struct pim_vxlan_sg *vxlan_sg = (struct pim_vxlan_sg *)backet->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);
+}
+
+void pim_vxlan_add_term_dev(struct pim_instance *pim,
+ struct interface *ifp)
+{
+ struct pim_interface *pim_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->name);
+
+ /* enable pim on the term ifp */
+ pim_ifp = (struct pim_interface *)ifp->info;
+ if (pim_ifp) {
+ PIM_IF_DO_PIM(pim_ifp->options);
+ } else {
+ pim_ifp = pim_if_new(ifp, false /*igmp*/, true /*pim*/,
+ false /*pimreg*/, true /*vxlan_term*/);
+ /* ensure that pimreg existss before using the newly created
+ * vxlan termination device
+ */
+ pim_if_create_pimreg(pim);
+ }
+
+ pim->vxlan.term_if = ifp;
+
+ if (pim->vxlan.sg_hash)
+ hash_iterate(pim_ifp->pim->vxlan.sg_hash,
+ pim_vxlan_term_mr_oif_update, ifp);
+}
+
+void pim_vxlan_del_term_dev(struct pim_instance *pim)
+{
+ struct interface *ifp = pim->vxlan.term_if;
+ struct pim_interface *pim_ifp;
+
+ if (PIM_DEBUG_VXLAN)
+ zlog_debug("vxlan term oif changed from %s to -", ifp->name);
+
+ pim->vxlan.term_if = NULL;
+
+ if (pim->vxlan.sg_hash)
+ hash_iterate(pim->vxlan.sg_hash,
+ pim_vxlan_term_mr_oif_update, NULL);
+
+ pim_ifp = (struct pim_interface *)ifp->info;
+ if (pim_ifp) {
+ PIM_IF_DONT_PIM(pim_ifp->options);
+ if (!PIM_IF_TEST_IGMP(pim_ifp->options))
+ pim_if_delete(ifp);
+ }
+
+}
+
void pim_vxlan_init(struct pim_instance *pim)
{
char hash_name[64];