diff options
Diffstat (limited to 'pimd/pim_mroute.c')
| -rw-r--r-- | pimd/pim_mroute.c | 1365 |
1 files changed, 1365 insertions, 0 deletions
diff --git a/pimd/pim_mroute.c b/pimd/pim_mroute.c new file mode 100644 index 0000000..24b251b --- /dev/null +++ b/pimd/pim_mroute.c @@ -0,0 +1,1365 @@ +/* + * PIM for Quagga + * Copyright (C) 2008 Everton da Silva Marques + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; see the file COPYING; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <zebra.h> +#include "log.h" +#include "privs.h" +#include "if.h" +#include "prefix.h" +#include "vty.h" +#include "plist.h" +#include "sockopt.h" +#include "lib_errors.h" +#include "lib/network.h" + +#include "pimd.h" +#include "pim_rpf.h" +#include "pim_mroute.h" +#include "pim_oil.h" +#include "pim_str.h" +#include "pim_time.h" +#include "pim_iface.h" +#include "pim_macro.h" +#include "pim_rp.h" +#include "pim_oil.h" +#include "pim_register.h" +#include "pim_ifchannel.h" +#include "pim_zlookup.h" +#include "pim_ssm.h" +#include "pim_sock.h" +#include "pim_vxlan.h" +#include "pim_msg.h" + +static void mroute_read_on(struct pim_instance *pim); +static int pim_upstream_mroute_update(struct channel_oil *c_oil, + const char *name); + +int pim_mroute_set(struct pim_instance *pim, int enable) +{ + int err; + int opt, data; + socklen_t data_len = sizeof(data); + + /* + * We need to create the VRF table for the pim mroute_socket + */ + if (pim->vrf->vrf_id != VRF_DEFAULT) { + frr_with_privs (&pimd_privs) { + + data = pim->vrf->data.l.table_id; + err = setsockopt(pim->mroute_socket, PIM_IPPROTO, + MRT_TABLE, &data, data_len); + if (err) { + zlog_warn( + "%s %s: failure: setsockopt(fd=%d,PIM_IPPROTO, MRT_TABLE=%d): errno=%d: %s", + __FILE__, __func__, pim->mroute_socket, + data, errno, safe_strerror(errno)); + return -1; + } + } + } + + frr_with_privs (&pimd_privs) { + opt = enable ? MRT_INIT : MRT_DONE; + /* + * *BSD *cares* about what value we pass down + * here + */ + data = 1; + err = setsockopt(pim->mroute_socket, PIM_IPPROTO, opt, &data, + data_len); + if (err) { + zlog_warn( + "%s %s: failure: setsockopt(fd=%d,PIM_IPPROTO,%s=%d): errno=%d: %s", + __FILE__, __func__, pim->mroute_socket, + enable ? "MRT_INIT" : "MRT_DONE", data, errno, + safe_strerror(errno)); + return -1; + } + } + +#if defined(HAVE_IP_PKTINFO) + if (enable) { + /* Linux and Solaris IP_PKTINFO */ + data = 1; + if (setsockopt(pim->mroute_socket, PIM_IPPROTO, IP_PKTINFO, + &data, data_len)) { + zlog_warn( + "Could not set IP_PKTINFO on socket fd=%d: errno=%d: %s", + pim->mroute_socket, errno, + safe_strerror(errno)); + } + } +#endif + +#if PIM_IPV == 6 + if (enable) { + /* Linux and Solaris IPV6_PKTINFO */ + data = 1; + if (setsockopt(pim->mroute_socket, PIM_IPPROTO, + IPV6_RECVPKTINFO, &data, data_len)) { + zlog_warn( + "Could not set IPV6_RECVPKTINFO on socket fd=%d: errno=%d: %s", + pim->mroute_socket, errno, + safe_strerror(errno)); + } + } +#endif + setsockopt_so_recvbuf(pim->mroute_socket, 1024 * 1024 * 8); + + if (set_nonblocking(pim->mroute_socket) < 0) { + zlog_warn( + "Could not set non blocking on socket fd=%d: errno=%d: %s", + pim->mroute_socket, errno, safe_strerror(errno)); + return -1; + } + + if (enable) { +#if defined linux + int upcalls = GMMSG_WRVIFWHOLE; + opt = MRT_PIM; + + err = setsockopt(pim->mroute_socket, PIM_IPPROTO, opt, &upcalls, + sizeof(upcalls)); + if (err) { + zlog_warn( + "Failure to register for VIFWHOLE and WRONGVIF upcalls %d %s", + errno, safe_strerror(errno)); + return -1; + } +#else + zlog_warn( + "PIM-SM will not work properly on this platform, until the ability to receive the WRVIFWHOLE upcall"); +#endif + } + + return 0; +} + +static const char *const gmmsgtype2str[GMMSG_WRVIFWHOLE + 1] = { + "<unknown_upcall?>", "NOCACHE", "WRONGVIF", "WHOLEPKT", "WRVIFWHOLE"}; + + +int pim_mroute_msg_nocache(int fd, struct interface *ifp, const kernmsg *msg) +{ + struct pim_interface *pim_ifp = ifp->info; + struct pim_upstream *up; + struct pim_rpf *rpg; + pim_sgaddr sg; + bool desync = false; + + memset(&sg, 0, sizeof(sg)); + sg.src = msg->msg_im_src; + sg.grp = msg->msg_im_dst; + + if (!pim_ifp) { + if (PIM_DEBUG_MROUTE) + zlog_debug( + "%s: PIM not enabled on interface, dropping packet to %pSG", + ifp->name, &sg); + return 0; + } + + rpg = RP(pim_ifp->pim, msg->msg_im_dst); + /* + * If the incoming interface is unknown OR + * the Interface type is SSM we don't need to + * do anything here + */ + if (!rpg) { + if (PIM_DEBUG_MROUTE) + zlog_debug("%s: no RPF for packet to %pSG", ifp->name, + &sg); + return 0; + } + if (pim_rpf_addr_is_inaddr_any(rpg)) { + if (PIM_DEBUG_MROUTE) + zlog_debug("%s: null RPF for packet to %pSG", ifp->name, + &sg); + return 0; + } + + /* + * If we've received a multicast packet that isn't connected to + * us + */ + if (!pim_if_connected_to_source(ifp, msg->msg_im_src)) { + if (PIM_DEBUG_MROUTE) + zlog_debug( + "%s: incoming packet to %pSG from non-connected source", + ifp->name, &sg); + return 0; + } + + if (!(PIM_I_am_DR(pim_ifp))) { + /* unlike the other debug messages, this one is further in the + * "normal operation" category and thus under _DETAIL + */ + if (PIM_DEBUG_MROUTE_DETAIL) + zlog_debug( + "%s: not DR on interface, not forwarding traffic for %pSG", + ifp->name, &sg); + + /* + * We are not the DR, but we are still receiving packets + * Let's blackhole those packets for the moment + * As that they will be coming up to the cpu + * and causing us to consider them. + * + * This *will* create a dangling channel_oil + * that I see no way to get rid of. Just noting + * this for future reference. + */ + up = pim_upstream_find_or_add( + &sg, ifp, PIM_UPSTREAM_FLAG_MASK_SRC_NOCACHE, __func__); + pim_upstream_mroute_add(up->channel_oil, __func__); + + return 0; + } + + up = pim_upstream_find_or_add(&sg, ifp, PIM_UPSTREAM_FLAG_MASK_FHR, + __func__); + if (up->channel_oil->installed) { + zlog_warn( + "%s: NOCACHE for %pSG, MFC entry disappeared - reinstalling", + ifp->name, &sg); + desync = true; + } + + /* + * I moved this debug till after the actual add because + * I want to take advantage of the up->sg_str being filled in. + */ + if (PIM_DEBUG_MROUTE) { + zlog_debug("%s: Adding a Route %s for WHOLEPKT consumption", + __func__, up->sg_str); + } + + PIM_UPSTREAM_FLAG_SET_SRC_STREAM(up->flags); + pim_upstream_keep_alive_timer_start(up, pim_ifp->pim->keep_alive_time); + + up->channel_oil->cc.pktcnt++; + // resolve mfcc_parent prior to mroute_add in channel_add_oif + if (up->rpf.source_nexthop.interface && + *oil_parent(up->channel_oil) >= MAXVIFS) { + pim_upstream_mroute_iif_update(up->channel_oil, __func__); + } + pim_register_join(up); + /* if we have receiver, inherit from parent */ + pim_upstream_inherited_olist_decide(pim_ifp->pim, up); + + /* we just got NOCACHE from the kernel, so... MFC is not in the + * kernel for some reason or another. Try installing again. + */ + if (desync) + pim_upstream_mroute_update(up->channel_oil, __func__); + return 0; +} + +int pim_mroute_msg_wholepkt(int fd, struct interface *ifp, const char *buf, + size_t len) +{ + struct pim_interface *pim_ifp; + pim_sgaddr sg; + struct pim_rpf *rpg; + const ipv_hdr *ip_hdr; + struct pim_upstream *up; + + pim_ifp = ifp->info; + + ip_hdr = (const ipv_hdr *)buf; + + memset(&sg, 0, sizeof(sg)); + sg.src = IPV_SRC(ip_hdr); + sg.grp = IPV_DST(ip_hdr); + + up = pim_upstream_find(pim_ifp->pim, &sg); + if (!up) { + pim_sgaddr star = sg; + star.src = PIMADDR_ANY; + + up = pim_upstream_find(pim_ifp->pim, &star); + + if (up && PIM_UPSTREAM_FLAG_TEST_CAN_BE_LHR(up->flags)) { + up = pim_upstream_add(pim_ifp->pim, &sg, ifp, + PIM_UPSTREAM_FLAG_MASK_SRC_LHR, + __func__, NULL); + if (!up) { + if (PIM_DEBUG_MROUTE) + zlog_debug( + "%s: Unable to create upstream information for %pSG", + __func__, &sg); + return 0; + } + pim_upstream_keep_alive_timer_start( + up, pim_ifp->pim->keep_alive_time); + pim_upstream_inherited_olist(pim_ifp->pim, up); + pim_upstream_update_join_desired(pim_ifp->pim, up); + + if (PIM_DEBUG_MROUTE) + zlog_debug("%s: Creating %s upstream on LHR", + __func__, up->sg_str); + return 0; + } + if (PIM_DEBUG_MROUTE_DETAIL) { + zlog_debug( + "%s: Unable to find upstream channel WHOLEPKT%pSG", + __func__, &sg); + } + return 0; + } + + if (!up->rpf.source_nexthop.interface) { + if (PIM_DEBUG_PIM_TRACE) + zlog_debug("%s: up %s RPF is not present", __func__, + up->sg_str); + return 0; + } + + pim_ifp = up->rpf.source_nexthop.interface->info; + + rpg = pim_ifp ? RP(pim_ifp->pim, sg.grp) : NULL; + + if ((pim_rpf_addr_is_inaddr_any(rpg)) || (!pim_ifp) || + (!(PIM_I_am_DR(pim_ifp)))) { + if (PIM_DEBUG_MROUTE) { + zlog_debug("%s: Failed Check send packet", __func__); + } + return 0; + } + + /* + * If we've received a register suppress + */ + if (!up->t_rs_timer) { + if (pim_is_grp_ssm(pim_ifp->pim, sg.grp)) { + if (PIM_DEBUG_PIM_REG) + zlog_debug( + "%pSG register forward skipped as group is SSM", + &sg); + return 0; + } + + if (!PIM_UPSTREAM_FLAG_TEST_FHR(up->flags)) { + if (PIM_DEBUG_PIM_REG) + zlog_debug( + "%s register forward skipped, not FHR", + up->sg_str); + return 0; + } + + pim_register_send((uint8_t *)buf + sizeof(ipv_hdr), + len - sizeof(ipv_hdr), + pim_ifp->primary_address, rpg, 0, up); + } + return 0; +} + +int pim_mroute_msg_wrongvif(int fd, struct interface *ifp, const kernmsg *msg) +{ + struct pim_ifchannel *ch; + struct pim_interface *pim_ifp; + pim_sgaddr sg; + + memset(&sg, 0, sizeof(sg)); + sg.src = msg->msg_im_src; + sg.grp = msg->msg_im_dst; + + /* + Send Assert(S,G) on iif as response to WRONGVIF kernel upcall. + + RFC 4601 4.8.2. PIM-SSM-Only Routers + + iif is the incoming interface of the packet. + if (iif is in inherited_olist(S,G)) { + send Assert(S,G) on iif + } + */ + + if (!ifp) { + if (PIM_DEBUG_MROUTE) + zlog_debug( + "%s: WRONGVIF (S,G)=%pSG could not find input interface for input_vif_index=%d", + __func__, &sg, msg->msg_im_vif); + return -1; + } + + pim_ifp = ifp->info; + if (!pim_ifp) { + if (PIM_DEBUG_MROUTE) + zlog_debug( + "%s: WRONGVIF (S,G)=%pSG multicast not enabled on interface %s", + __func__, &sg, ifp->name); + return -2; + } + + ch = pim_ifchannel_find(ifp, &sg); + if (!ch) { + pim_sgaddr star_g = sg; + if (PIM_DEBUG_MROUTE) + zlog_debug( + "%s: WRONGVIF (S,G)=%pSG could not find channel on interface %s", + __func__, &sg, ifp->name); + + star_g.src = PIMADDR_ANY; + ch = pim_ifchannel_find(ifp, &star_g); + if (!ch) { + if (PIM_DEBUG_MROUTE) + zlog_debug( + "%s: WRONGVIF (*,G)=%pSG could not find channel on interface %s", + __func__, &star_g, ifp->name); + return -3; + } + } + + /* + RFC 4601: 4.6.1. (S,G) Assert Message State Machine + + Transitions from NoInfo State + + An (S,G) data packet arrives on interface I, AND + CouldAssert(S,G,I)==TRUE An (S,G) data packet arrived on an + downstream interface that is in our (S,G) outgoing interface + list. We optimistically assume that we will be the assert + winner for this (S,G), and so we transition to the "I am Assert + Winner" state and perform Actions A1 (below), which will + initiate the assert negotiation for (S,G). + */ + + if (ch->ifassert_state != PIM_IFASSERT_NOINFO) { + if (PIM_DEBUG_MROUTE) { + zlog_debug( + "%s: WRONGVIF (S,G)=%s channel is not on Assert NoInfo state for interface %s", + __func__, ch->sg_str, ifp->name); + } + return -4; + } + + if (!PIM_IF_FLAG_TEST_COULD_ASSERT(ch->flags)) { + if (PIM_DEBUG_MROUTE) { + zlog_debug( + "%s: WRONGVIF (S,G)=%s interface %s is not downstream for channel", + __func__, ch->sg_str, ifp->name); + } + return -5; + } + + if (assert_action_a1(ch)) { + if (PIM_DEBUG_MROUTE) { + zlog_debug( + "%s: WRONGVIF (S,G)=%s assert_action_a1 failure on interface %s", + __func__, ch->sg_str, ifp->name); + } + return -6; + } + + return 0; +} + +int pim_mroute_msg_wrvifwhole(int fd, struct interface *ifp, const char *buf, + size_t len) +{ + const ipv_hdr *ip_hdr = (const ipv_hdr *)buf; + struct pim_interface *pim_ifp; + struct pim_instance *pim; + struct pim_ifchannel *ch; + struct pim_upstream *up; + pim_sgaddr star_g; + pim_sgaddr sg; + + pim_ifp = ifp->info; + + memset(&sg, 0, sizeof(sg)); + sg.src = IPV_SRC(ip_hdr); + sg.grp = IPV_DST(ip_hdr); + + ch = pim_ifchannel_find(ifp, &sg); + if (ch) { + if (PIM_DEBUG_MROUTE) + zlog_debug( + "WRVIFWHOLE (S,G)=%s found ifchannel on interface %s", + ch->sg_str, ifp->name); + return -1; + } + + star_g = sg; + star_g.src = PIMADDR_ANY; + + pim = pim_ifp->pim; + /* + * If the incoming interface is the pimreg, then + * we know the callback is associated with a pim register + * packet and there is nothing to do here as that + * normal pim processing will see the packet and allow + * us to do the right thing. + */ + if (ifp == pim->regiface) { + return 0; + } + + up = pim_upstream_find(pim_ifp->pim, &sg); + if (up) { + struct pim_upstream *parent; + struct pim_nexthop source; + struct pim_rpf *rpf = RP(pim_ifp->pim, sg.grp); + + /* No RPF or No RPF interface or No mcast on RPF interface */ + if (!rpf || !rpf->source_nexthop.interface || + !rpf->source_nexthop.interface->info) + return 0; + + /* + * If we have received a WRVIFWHOLE and are at this + * point, we could be receiving the packet on the *,G + * tree, let's check and if so we can safely drop + * it. + */ + parent = pim_upstream_find(pim_ifp->pim, &star_g); + if (parent && parent->rpf.source_nexthop.interface == ifp) + return 0; + + pim_ifp = rpf->source_nexthop.interface->info; + + memset(&source, 0, sizeof(source)); + /* + * If we are the fhr that means we are getting a callback during + * the pimreg period, so I believe we can ignore this packet + */ + if (!PIM_UPSTREAM_FLAG_TEST_FHR(up->flags)) { + /* + * No if channel, but upstream we are at the RP. + * + * This could be a anycast RP too and we may + * not have received a register packet from + * the source here at all. So gracefully + * bow out of doing a nexthop lookup and + * setting the SPTBIT to true + */ + if (!(pim_addr_is_any(up->upstream_register)) && + pim_nexthop_lookup(pim_ifp->pim, &source, + up->upstream_register, 0)) { + pim_register_stop_send(source.interface, &sg, + pim_ifp->primary_address, + up->upstream_register); + up->sptbit = PIM_UPSTREAM_SPTBIT_TRUE; + } + + pim_upstream_inherited_olist(pim_ifp->pim, up); + if (!up->channel_oil->installed) + pim_upstream_mroute_add(up->channel_oil, + __func__); + } else { + if (I_am_RP(pim_ifp->pim, up->sg.grp)) { + if (pim_nexthop_lookup(pim_ifp->pim, &source, + up->upstream_register, + 0)) + pim_register_stop_send( + source.interface, &sg, + pim_ifp->primary_address, + up->upstream_register); + up->sptbit = PIM_UPSTREAM_SPTBIT_TRUE; + } else { + /* + * At this point pimd is connected to + * the source, it has a parent, we are not + * the RP and the SPTBIT should be set + * since we know *the* S,G is on the SPT. + * The first time this happens, let's cause + * an immediate join to go out so that + * the RP can trim this guy immediately + * if necessary, instead of waiting + * one join/prune send cycle + */ + if (up->sptbit != PIM_UPSTREAM_SPTBIT_TRUE && + up->parent && + up->rpf.source_nexthop.interface != + up->parent->rpf.source_nexthop + .interface) { + up->sptbit = PIM_UPSTREAM_SPTBIT_TRUE; + pim_jp_agg_single_upstream_send( + &up->parent->rpf, up->parent, + true); + } + } + pim_upstream_keep_alive_timer_start( + up, pim_ifp->pim->keep_alive_time); + pim_upstream_inherited_olist(pim_ifp->pim, up); + pim_mroute_msg_wholepkt(fd, ifp, buf, len); + } + return 0; + } + + pim_ifp = ifp->info; + if (pim_if_connected_to_source(ifp, sg.src)) { + up = pim_upstream_add(pim_ifp->pim, &sg, ifp, + PIM_UPSTREAM_FLAG_MASK_FHR, __func__, + NULL); + if (!up) { + if (PIM_DEBUG_MROUTE) + zlog_debug( + "%pSG: WRONGVIF%s unable to create upstream on interface", + &sg, ifp->name); + return -2; + } + PIM_UPSTREAM_FLAG_SET_SRC_STREAM(up->flags); + pim_upstream_keep_alive_timer_start( + up, pim_ifp->pim->keep_alive_time); + up->channel_oil->cc.pktcnt++; + pim_register_join(up); + pim_upstream_inherited_olist(pim_ifp->pim, up); + if (!up->channel_oil->installed) + pim_upstream_mroute_add(up->channel_oil, __func__); + + // Send the packet to the RP + pim_mroute_msg_wholepkt(fd, ifp, buf, len); + } else { + up = pim_upstream_add(pim_ifp->pim, &sg, ifp, + PIM_UPSTREAM_FLAG_MASK_SRC_NOCACHE, + __func__, NULL); + if (!up->channel_oil->installed) + pim_upstream_mroute_add(up->channel_oil, __func__); + } + + return 0; +} + +#if PIM_IPV == 4 +static int process_igmp_packet(struct pim_instance *pim, const char *buf, + size_t buf_size, ifindex_t ifindex) +{ + struct interface *ifp; + struct pim_interface *pim_ifp; + struct in_addr ifaddr; + struct gm_sock *igmp; + const struct prefix *connected_src; + const struct ip *ip_hdr = (const struct ip *)buf; + + /* We have the IP packet but we do not know which interface this + * packet was + * received on. Find the interface that is on the same subnet as + * the source + * of the IP packet. + */ + ifp = if_lookup_by_index(ifindex, pim->vrf->vrf_id); + + if (!ifp || !ifp->info) + return 0; + + connected_src = pim_if_connected_to_source(ifp, ip_hdr->ip_src); + + if (!connected_src) { + if (PIM_DEBUG_GM_PACKETS) { + zlog_debug( + "Recv IGMP packet on interface: %s from a non-connected source: %pI4", + ifp->name, &ip_hdr->ip_src); + } + return 0; + } + + pim_ifp = ifp->info; + ifaddr = connected_src->u.prefix4; + igmp = pim_igmp_sock_lookup_ifaddr(pim_ifp->gm_socket_list, ifaddr); + + if (PIM_DEBUG_GM_PACKETS) { + zlog_debug( + "%s(%s): igmp kernel upcall on %s(%p) for %pI4 -> %pI4", + __func__, pim->vrf->name, ifp->name, igmp, + &ip_hdr->ip_src, &ip_hdr->ip_dst); + } + if (igmp) + pim_igmp_packet(igmp, (char *)buf, buf_size); + else if (PIM_DEBUG_GM_PACKETS) { + zlog_debug( + "No IGMP socket on interface: %s with connected source: %pFX", + ifp->name, connected_src); + } + return 0; +} +#endif + +int pim_mroute_msg(struct pim_instance *pim, const char *buf, size_t buf_size, + ifindex_t ifindex) +{ + struct interface *ifp; + const ipv_hdr *ip_hdr; + const kernmsg *msg; + + if (buf_size < (int)sizeof(ipv_hdr)) + return 0; + + ip_hdr = (const ipv_hdr *)buf; + +#if PIM_IPV == 4 + if (ip_hdr->ip_p == IPPROTO_IGMP) { + process_igmp_packet(pim, buf, buf_size, ifindex); + } else if (ip_hdr->ip_p) { + if (PIM_DEBUG_MROUTE_DETAIL) { + zlog_debug( + "%s: no kernel upcall proto=%d src: %pI4 dst: %pI4 msg_size=%ld", + __func__, ip_hdr->ip_p, &ip_hdr->ip_src, + &ip_hdr->ip_dst, (long int)buf_size); + } + + } else { +#else + + if ((ip_hdr->ip6_vfc & 0xf) == 0) { +#endif + msg = (const kernmsg *)buf; + + ifp = pim_if_find_by_vif_index(pim, msg->msg_im_vif); + + if (!ifp) + return 0; + if (PIM_DEBUG_MROUTE) { +#if PIM_IPV == 4 + zlog_debug( + "%s: pim kernel upcall %s type=%d ip_p=%d from fd=%d for (S,G)=(%pI4,%pI4) on %s vifi=%d size=%ld", + __func__, gmmsgtype2str[msg->msg_im_msgtype], + msg->msg_im_msgtype, ip_hdr->ip_p, + pim->mroute_socket, &msg->msg_im_src, + &msg->msg_im_dst, ifp->name, msg->msg_im_vif, + (long int)buf_size); +#else + zlog_debug( + "%s: pim kernel upcall %s type=%d ip_p=%d from fd=%d for (S,G)=(%pI6,%pI6) on %s vifi=%d size=%ld", + __func__, gmmsgtype2str[msg->msg_im_msgtype], + msg->msg_im_msgtype, ip_hdr->ip6_nxt, + pim->mroute_socket, &msg->msg_im_src, + &msg->msg_im_dst, ifp->name, msg->msg_im_vif, + (long int)buf_size); +#endif + } + + switch (msg->msg_im_msgtype) { + case GMMSG_WRONGVIF: + return pim_mroute_msg_wrongvif(pim->mroute_socket, ifp, + msg); + case GMMSG_NOCACHE: + return pim_mroute_msg_nocache(pim->mroute_socket, ifp, + msg); + case GMMSG_WHOLEPKT: + return pim_mroute_msg_wholepkt(pim->mroute_socket, ifp, + (const char *)msg, + buf_size); + case GMMSG_WRVIFWHOLE: + return pim_mroute_msg_wrvifwhole(pim->mroute_socket, + ifp, (const char *)msg, + buf_size); + default: + break; + } + } + + return 0; +} + +static void mroute_read(struct thread *t) +{ + struct pim_instance *pim; + static long long count; + char buf[10000]; + int cont = 1; + int rd; + ifindex_t ifindex; + pim = THREAD_ARG(t); + + while (cont) { + rd = pim_socket_recvfromto(pim->mroute_socket, (uint8_t *)buf, + sizeof(buf), NULL, NULL, NULL, NULL, + &ifindex); + if (rd <= 0) { + if (errno == EINTR) + continue; + if (errno == EWOULDBLOCK || errno == EAGAIN) + break; + + zlog_warn( + "%s: failure reading rd=%d: fd=%d: errno=%d: %s", + __func__, rd, pim->mroute_socket, errno, + safe_strerror(errno)); + goto done; + } + + pim_mroute_msg(pim, buf, rd, ifindex); + + count++; + if (count % router->packet_process == 0) + cont = 0; + } +/* Keep reading */ +done: + mroute_read_on(pim); + + return; +} + +static void mroute_read_on(struct pim_instance *pim) +{ + thread_add_read(router->master, mroute_read, pim, pim->mroute_socket, + &pim->thread); +} + +static void mroute_read_off(struct pim_instance *pim) +{ + THREAD_OFF(pim->thread); +} + +int pim_mroute_socket_enable(struct pim_instance *pim) +{ + int fd; + + frr_with_privs(&pimd_privs) { + +#if PIM_IPV == 4 + fd = socket(AF_INET, SOCK_RAW, IPPROTO_IGMP); +#else + fd = socket(AF_INET6, SOCK_RAW, IPPROTO_ICMPV6); +#endif + if (fd < 0) { + zlog_warn("Could not create mroute socket: errno=%d: %s", + errno, + safe_strerror(errno)); + return -2; + } + +#if PIM_IPV == 6 + struct icmp6_filter filter[1]; + int ret; + + /* Unlike IPv4, this socket is not used for MLD, so just drop + * everything with an empty ICMP6 filter. Otherwise we get + * all kinds of garbage here, possibly even non-multicast + * related ICMPv6 traffic (e.g. ping) + * + * (mroute kernel upcall "packets" are injected directly on the + * socket, this sockopt -or any other- has no effect on them) + */ + ICMP6_FILTER_SETBLOCKALL(filter); + ret = setsockopt(fd, SOL_ICMPV6, ICMP6_FILTER, filter, + sizeof(filter)); + if (ret) + zlog_err( + "(VRF %s) failed to set mroute control filter: %m", + pim->vrf->name); +#endif + +#ifdef SO_BINDTODEVICE + if (pim->vrf->vrf_id != VRF_DEFAULT + && setsockopt(fd, SOL_SOCKET, SO_BINDTODEVICE, + pim->vrf->name, strlen(pim->vrf->name))) { + zlog_warn("Could not setsockopt SO_BINDTODEVICE: %s", + safe_strerror(errno)); + close(fd); + return -3; + } +#endif + + } + + pim->mroute_socket = fd; + if (pim_mroute_set(pim, 1)) { + zlog_warn( + "Could not enable mroute on socket fd=%d: errno=%d: %s", + fd, errno, safe_strerror(errno)); + close(fd); + pim->mroute_socket = -1; + return -3; + } + + pim->mroute_socket_creation = pim_time_monotonic_sec(); + + mroute_read_on(pim); + + return 0; +} + +int pim_mroute_socket_disable(struct pim_instance *pim) +{ + if (pim_mroute_set(pim, 0)) { + zlog_warn( + "Could not disable mroute on socket fd=%d: errno=%d: %s", + pim->mroute_socket, errno, safe_strerror(errno)); + return -2; + } + + if (close(pim->mroute_socket)) { + zlog_warn("Failure closing mroute socket: fd=%d errno=%d: %s", + pim->mroute_socket, errno, safe_strerror(errno)); + return -3; + } + + mroute_read_off(pim); + pim->mroute_socket = -1; + + return 0; +} + +/* + For each network interface (e.g., physical or a virtual tunnel) that + would be used for multicast forwarding, a corresponding multicast + interface must be added to the kernel. + */ +int pim_mroute_add_vif(struct interface *ifp, pim_addr ifaddr, + unsigned char flags) +{ + struct pim_interface *pim_ifp = ifp->info; + pim_vifctl vc; + int err; + + if (PIM_DEBUG_MROUTE) + zlog_debug("%s: Add Vif %d (%s[%s])", __func__, + pim_ifp->mroute_vif_index, ifp->name, + pim_ifp->pim->vrf->name); + + memset(&vc, 0, sizeof(vc)); + vc.vc_vifi = pim_ifp->mroute_vif_index; +#if PIM_IPV == 4 +#ifdef VIFF_USE_IFINDEX + vc.vc_lcl_ifindex = ifp->ifindex; +#else + if (ifaddr.s_addr == INADDR_ANY) { + zlog_warn( + "%s: unnumbered interfaces are not supported on this platform", + __func__); + return -1; + } + memcpy(&vc.vc_lcl_addr, &ifaddr, sizeof(vc.vc_lcl_addr)); +#endif +#else + vc.vc_pifi = ifp->ifindex; +#endif + vc.vc_flags = flags; + vc.vc_threshold = PIM_MROUTE_MIN_TTL; + vc.vc_rate_limit = 0; + +#if PIM_IPV == 4 +#ifdef PIM_DVMRP_TUNNEL + if (vc.vc_flags & VIFF_TUNNEL) { + memcpy(&vc.vc_rmt_addr, &vif_remote_addr, + sizeof(vc.vc_rmt_addr)); + } +#endif +#endif + + err = setsockopt(pim_ifp->pim->mroute_socket, PIM_IPPROTO, MRT_ADD_VIF, + (void *)&vc, sizeof(vc)); + if (err) { + zlog_warn( + "%s: failure: setsockopt(fd=%d,PIM_IPPROTO,MRT_ADD_VIF,vif_index=%d,ifaddr=%pPAs,flag=%d): errno=%d: %s", + __func__, pim_ifp->pim->mroute_socket, ifp->ifindex, + &ifaddr, flags, errno, safe_strerror(errno)); + return -2; + } + + return 0; +} + +int pim_mroute_del_vif(struct interface *ifp) +{ + struct pim_interface *pim_ifp = ifp->info; + pim_vifctl vc; + int err; + + if (PIM_DEBUG_MROUTE) + zlog_debug("%s: Del Vif %d (%s[%s])", __func__, + pim_ifp->mroute_vif_index, ifp->name, + pim_ifp->pim->vrf->name); + + memset(&vc, 0, sizeof(vc)); + vc.vc_vifi = pim_ifp->mroute_vif_index; + + err = setsockopt(pim_ifp->pim->mroute_socket, PIM_IPPROTO, MRT_DEL_VIF, + (void *)&vc, sizeof(vc)); + if (err) { + zlog_warn( + "%s %s: failure: setsockopt(fd=%d,PIM_IPPROTO,MRT_DEL_VIF,vif_index=%d): errno=%d: %s", + __FILE__, __func__, pim_ifp->pim->mroute_socket, + pim_ifp->mroute_vif_index, errno, safe_strerror(errno)); + return -2; + } + + return 0; +} + +/* + * Prevent creating MFC entry with OIF=IIF. + * + * This is a protection against implementation mistakes. + * + * PIM protocol implicitely ensures loopfree multicast topology. + * + * IGMP must be protected against adding looped MFC entries created + * by both source and receiver attached to the same interface. See + * TODO T22. + * We shall allow igmp to create upstream when it is DR for the intf. + * Assume RP reachable via non DR. + */ +bool pim_mroute_allow_iif_in_oil(struct channel_oil *c_oil, + int oif_index) +{ +#ifdef PIM_ENFORCE_LOOPFREE_MFC + struct interface *ifp_out; + struct pim_interface *pim_ifp; + + if (c_oil->up && + PIM_UPSTREAM_FLAG_TEST_ALLOW_IIF_IN_OIL(c_oil->up->flags)) + return true; + + ifp_out = pim_if_find_by_vif_index(c_oil->pim, oif_index); + if (!ifp_out) + return false; + pim_ifp = ifp_out->info; + if (!pim_ifp) + return false; + if ((c_oil->oif_flags[oif_index] & PIM_OIF_FLAG_PROTO_GM) && + PIM_I_am_DR(pim_ifp)) + return true; + + return false; +#else + return true; +#endif +} + +static inline void pim_mroute_copy(struct channel_oil *out, + struct channel_oil *in) +{ + int i; + + *oil_origin(out) = *oil_origin(in); + *oil_mcastgrp(out) = *oil_mcastgrp(in); + *oil_parent(out) = *oil_parent(in); + + for (i = 0; i < MAXVIFS; ++i) { + if (*oil_parent(out) == i && + !pim_mroute_allow_iif_in_oil(in, i)) { + oil_if_set(out, i, 0); + continue; + } + + if (in->oif_flags[i] & PIM_OIF_FLAG_MUTE) + oil_if_set(out, i, 0); + else + oil_if_set(out, i, oil_if_has(in, i)); + } +} + +/* This function must not be called directly 0 + * use pim_upstream_mroute_add or pim_static_mroute_add instead + */ +static int pim_mroute_add(struct channel_oil *c_oil, const char *name) +{ + struct pim_instance *pim = c_oil->pim; + struct channel_oil tmp_oil[1] = { }; + int err; + + pim->mroute_add_last = pim_time_monotonic_sec(); + ++pim->mroute_add_events; + + /* Copy the oil to a temporary structure to fixup (without need to + * later restore) before sending the mroute add to the dataplane + */ + pim_mroute_copy(tmp_oil, c_oil); + + /* The linux kernel *expects* the incoming + * vif to be part of the outgoing list + * in the case of a (*,G). + */ + if (pim_addr_is_any(*oil_origin(c_oil))) { + oil_if_set(tmp_oil, *oil_parent(c_oil), 1); + } + + /* + * If we have an unresolved cache entry for the S,G + * it is owned by the pimreg for the incoming IIF + * So set pimreg as the IIF temporarily to cause + * the packets to be forwarded. Then set it + * to the correct IIF afterwords. + */ + if (!c_oil->installed && !pim_addr_is_any(*oil_origin(c_oil)) + && *oil_parent(c_oil) != 0) { + *oil_parent(tmp_oil) = 0; + } + /* For IPv6 MRT_ADD_MFC is defined to MRT6_ADD_MFC */ + err = setsockopt(pim->mroute_socket, PIM_IPPROTO, MRT_ADD_MFC, + &tmp_oil->oil, sizeof(tmp_oil->oil)); + + if (!err && !c_oil->installed + && !pim_addr_is_any(*oil_origin(c_oil)) + && *oil_parent(c_oil) != 0) { + *oil_parent(tmp_oil) = *oil_parent(c_oil); + err = setsockopt(pim->mroute_socket, PIM_IPPROTO, MRT_ADD_MFC, + &tmp_oil->oil, sizeof(tmp_oil->oil)); + } + + if (err) { + zlog_warn( + "%s %s: failure: setsockopt(fd=%d,PIM_IPPROTO,MRT_ADD_MFC): errno=%d: %s", + __FILE__, __func__, pim->mroute_socket, errno, + safe_strerror(errno)); + return -2; + } + + if (PIM_DEBUG_MROUTE) { + char buf[1000]; + zlog_debug("%s(%s), vrf %s Added Route: %s", __func__, name, + pim->vrf->name, + pim_channel_oil_dump(c_oil, buf, sizeof(buf))); + } + + if (!c_oil->installed) { + c_oil->installed = 1; + c_oil->mroute_creation = pim_time_monotonic_sec(); + } + + return 0; +} + +static int pim_upstream_get_mroute_iif(struct channel_oil *c_oil, + const char *name) +{ + vifi_t iif = MAXVIFS; + struct interface *ifp = NULL; + struct pim_interface *pim_ifp; + struct pim_upstream *up = c_oil->up; + + if (up) { + if (PIM_UPSTREAM_FLAG_TEST_USE_RPT(up->flags)) { + if (up->parent) + ifp = up->parent->rpf.source_nexthop.interface; + } else { + ifp = up->rpf.source_nexthop.interface; + } + if (ifp) { + pim_ifp = (struct pim_interface *)ifp->info; + if (pim_ifp) + iif = pim_ifp->mroute_vif_index; + } + } + return iif; +} + +static int pim_upstream_mroute_update(struct channel_oil *c_oil, + const char *name) +{ + char buf[1000]; + + if (*oil_parent(c_oil) >= MAXVIFS) { + /* the c_oil cannot be installed as a mroute yet */ + if (PIM_DEBUG_MROUTE) + zlog_debug( + "%s(%s) %s mroute not ready to be installed; %s", + __func__, name, + pim_channel_oil_dump(c_oil, buf, + sizeof(buf)), + c_oil->installed ? + "uninstall" : "skip"); + /* if already installed flush it out as we are going to stop + * updates to it leaving it in a stale state + */ + if (c_oil->installed) + pim_mroute_del(c_oil, name); + /* return success (skipped) */ + return 0; + } + + return pim_mroute_add(c_oil, name); +} + +/* IIF associated with SGrpt entries are re-evaluated when the parent + * (*,G) entries IIF changes + */ +static void pim_upstream_all_sources_iif_update(struct pim_upstream *up) +{ + struct listnode *listnode; + struct pim_upstream *child; + + for (ALL_LIST_ELEMENTS_RO(up->sources, listnode, + child)) { + if (PIM_UPSTREAM_FLAG_TEST_USE_RPT(child->flags)) + pim_upstream_mroute_iif_update(child->channel_oil, + __func__); + } +} + +/* In the case of "PIM state machine" added mroutes an upstream entry + * must be present to decide on the SPT-forwarding vs. RPT-forwarding. + */ +int pim_upstream_mroute_add(struct channel_oil *c_oil, const char *name) +{ + vifi_t iif; + + iif = pim_upstream_get_mroute_iif(c_oil, name); + + if (*oil_parent(c_oil) != iif) { + *oil_parent(c_oil) = iif; + if (pim_addr_is_any(*oil_origin(c_oil)) && + c_oil->up) + pim_upstream_all_sources_iif_update(c_oil->up); + } else { + *oil_parent(c_oil) = iif; + } + + return pim_upstream_mroute_update(c_oil, name); +} + +/* Look for IIF changes and update the dateplane entry only if the IIF + * has changed. + */ +int pim_upstream_mroute_iif_update(struct channel_oil *c_oil, const char *name) +{ + vifi_t iif; + char buf[1000]; + + iif = pim_upstream_get_mroute_iif(c_oil, name); + if (*oil_parent(c_oil) == iif) { + /* no change */ + return 0; + } + *oil_parent(c_oil) = iif; + + if (pim_addr_is_any(*oil_origin(c_oil)) && + c_oil->up) + pim_upstream_all_sources_iif_update(c_oil->up); + + if (PIM_DEBUG_MROUTE_DETAIL) + zlog_debug("%s(%s) %s mroute iif update %d", + __func__, name, + pim_channel_oil_dump(c_oil, buf, + sizeof(buf)), iif); + /* XXX: is this hack needed? */ + c_oil->oil_inherited_rescan = 1; + return pim_upstream_mroute_update(c_oil, name); +} + +int pim_static_mroute_add(struct channel_oil *c_oil, const char *name) +{ + return pim_mroute_add(c_oil, name); +} + +void pim_static_mroute_iif_update(struct channel_oil *c_oil, + int input_vif_index, + const char *name) +{ + if (*oil_parent(c_oil) == input_vif_index) + return; + + *oil_parent(c_oil) = input_vif_index; + if (input_vif_index == MAXVIFS) + pim_mroute_del(c_oil, name); + else + pim_static_mroute_add(c_oil, name); +} + +int pim_mroute_del(struct channel_oil *c_oil, const char *name) +{ + struct pim_instance *pim = c_oil->pim; + int err; + + pim->mroute_del_last = pim_time_monotonic_sec(); + ++pim->mroute_del_events; + + if (!c_oil->installed) { + if (PIM_DEBUG_MROUTE) { + char buf[1000]; + zlog_debug( + "%s %s: vifi %d for route is %s not installed, do not need to send del req. ", + __FILE__, __func__, *oil_parent(c_oil), + pim_channel_oil_dump(c_oil, buf, sizeof(buf))); + } + return -2; + } + + err = setsockopt(pim->mroute_socket, PIM_IPPROTO, MRT_DEL_MFC, + &c_oil->oil, sizeof(c_oil->oil)); + if (err) { + if (PIM_DEBUG_MROUTE) + zlog_warn( + "%s %s: failure: setsockopt(fd=%d,PIM_IPPROTO,MRT_DEL_MFC): errno=%d: %s", + __FILE__, __func__, pim->mroute_socket, errno, + safe_strerror(errno)); + return -2; + } + + if (PIM_DEBUG_MROUTE) { + char buf[1000]; + zlog_debug("%s(%s), vrf %s Deleted Route: %s", __func__, name, + pim->vrf->name, + pim_channel_oil_dump(c_oil, buf, sizeof(buf))); + } + + // Reset kernel installed flag + c_oil->installed = 0; + + return 0; +} + +void pim_mroute_update_counters(struct channel_oil *c_oil) +{ + struct pim_instance *pim = c_oil->pim; + pim_sioc_sg_req sgreq; + + c_oil->cc.oldpktcnt = c_oil->cc.pktcnt; + c_oil->cc.oldbytecnt = c_oil->cc.bytecnt; + c_oil->cc.oldwrong_if = c_oil->cc.wrong_if; + + if (!c_oil->installed) { + c_oil->cc.lastused = 100 * pim->keep_alive_time; + if (PIM_DEBUG_MROUTE) { + pim_sgaddr sg; + + sg.src = *oil_origin(c_oil); + sg.grp = *oil_mcastgrp(c_oil); + zlog_debug("Channel%pSG is not installed no need to collect data from kernel", + &sg); + } + return; + } + + + memset(&sgreq, 0, sizeof(sgreq)); + + pim_zlookup_sg_statistics(c_oil); + +#if PIM_IPV == 4 + sgreq.src = *oil_origin(c_oil); + sgreq.grp = *oil_mcastgrp(c_oil); +#else + sgreq.src = c_oil->oil.mf6cc_origin; + sgreq.grp = c_oil->oil.mf6cc_mcastgrp; +#endif + if (ioctl(pim->mroute_socket, PIM_SIOCGETSGCNT, &sgreq)) { + pim_sgaddr sg; + + sg.src = *oil_origin(c_oil); + sg.grp = *oil_mcastgrp(c_oil); + + zlog_warn( + "ioctl(PIM_SIOCGETSGCNT=%lu) failure for (S,G)=%pSG: errno=%d: %s", + (unsigned long)PIM_SIOCGETSGCNT, &sg, errno, + safe_strerror(errno)); + return; + } + + c_oil->cc.pktcnt = sgreq.pktcnt; + c_oil->cc.bytecnt = sgreq.bytecnt; + c_oil->cc.wrong_if = sgreq.wrong_if; + return; +} |