2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
5 * (c) 2004 Mickael Hoerdt, <hoerdt@clarinet.u-strasbg.fr>
6 * LSIIT Laboratory, Strasbourg, France
7 * (c) 2004 Jean-Philippe Andriot, <jean-philippe.andriot@6WIND.com>
9 * Copyright (C)2007,2008 USAGI/WIDE Project
10 * YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
19 #include <asm/system.h>
20 #include <asm/uaccess.h>
21 #include <linux/types.h>
22 #include <linux/sched.h>
23 #include <linux/errno.h>
24 #include <linux/timer.h>
26 #include <linux/kernel.h>
27 #include <linux/fcntl.h>
28 #include <linux/stat.h>
29 #include <linux/socket.h>
30 #include <linux/inet.h>
31 #include <linux/netdevice.h>
32 #include <linux/inetdevice.h>
33 #include <linux/proc_fs.h>
34 #include <linux/seq_file.h>
35 #include <linux/init.h>
36 #include <net/protocol.h>
37 #include <linux/skbuff.h>
40 #include <linux/notifier.h>
41 #include <linux/if_arp.h>
42 #include <net/checksum.h>
43 #include <net/netlink.h>
46 #include <net/ip6_route.h>
47 #include <linux/mroute6.h>
48 #include <linux/pim.h>
49 #include <net/addrconf.h>
50 #include <linux/netfilter_ipv6.h>
51 #include <net/ip6_checksum.h>
53 /* Big lock, protecting vif table, mrt cache and mroute socket state.
54 Note that the changes are semaphored via rtnl_lock.
57 static DEFINE_RWLOCK(mrt_lock);
60 * Multicast router control variables
63 #define MIF_EXISTS(_net, _idx) ((_net)->ipv6.vif6_table[_idx].dev != NULL)
65 static struct mfc6_cache *mfc_unres_queue; /* Queue of unresolved entries */
67 /* Special spinlock for queue of unresolved entries */
68 static DEFINE_SPINLOCK(mfc_unres_lock);
70 /* We return to original Alan's scheme. Hash table of resolved
71 entries is changed only in process context and protected
72 with weak lock mrt_lock. Queue of unresolved entries is protected
73 with strong spinlock mfc_unres_lock.
75 In this case data path is free of exclusive locks at all.
78 static struct kmem_cache *mrt_cachep __read_mostly;
80 static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache);
81 static int ip6mr_cache_report(struct net *net, struct sk_buff *pkt,
82 mifi_t mifi, int assert);
83 static int ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm);
84 static void mroute_clean_tables(struct net *net);
86 #ifdef CONFIG_IPV6_PIMSM_V2
87 static struct inet6_protocol pim6_protocol;
90 static struct timer_list ipmr_expire_timer;
95 struct ipmr_mfc_iter {
96 struct seq_net_private p;
97 struct mfc6_cache **cache;
102 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
103 struct ipmr_mfc_iter *it, loff_t pos)
105 struct mfc6_cache *mfc;
107 it->cache = net->ipv6.mfc6_cache_array;
108 read_lock(&mrt_lock);
109 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++)
110 for (mfc = net->ipv6.mfc6_cache_array[it->ct];
111 mfc; mfc = mfc->next)
114 read_unlock(&mrt_lock);
116 it->cache = &mfc_unres_queue;
117 spin_lock_bh(&mfc_unres_lock);
118 for (mfc = mfc_unres_queue; mfc; mfc = mfc->next)
119 if (net_eq(mfc6_net(mfc), net) &&
122 spin_unlock_bh(&mfc_unres_lock);
132 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
135 struct ipmr_vif_iter {
136 struct seq_net_private p;
140 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
141 struct ipmr_vif_iter *iter,
144 for (iter->ct = 0; iter->ct < net->ipv6.maxvif; ++iter->ct) {
145 if (!MIF_EXISTS(net, iter->ct))
148 return &net->ipv6.vif6_table[iter->ct];
153 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
156 struct net *net = seq_file_net(seq);
158 read_lock(&mrt_lock);
159 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
163 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
165 struct ipmr_vif_iter *iter = seq->private;
166 struct net *net = seq_file_net(seq);
169 if (v == SEQ_START_TOKEN)
170 return ip6mr_vif_seq_idx(net, iter, 0);
172 while (++iter->ct < net->ipv6.maxvif) {
173 if (!MIF_EXISTS(net, iter->ct))
175 return &net->ipv6.vif6_table[iter->ct];
180 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
183 read_unlock(&mrt_lock);
186 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
188 struct net *net = seq_file_net(seq);
190 if (v == SEQ_START_TOKEN) {
192 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
194 const struct mif_device *vif = v;
195 const char *name = vif->dev ? vif->dev->name : "none";
198 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
199 vif - net->ipv6.vif6_table,
200 name, vif->bytes_in, vif->pkt_in,
201 vif->bytes_out, vif->pkt_out,
207 static struct seq_operations ip6mr_vif_seq_ops = {
208 .start = ip6mr_vif_seq_start,
209 .next = ip6mr_vif_seq_next,
210 .stop = ip6mr_vif_seq_stop,
211 .show = ip6mr_vif_seq_show,
214 static int ip6mr_vif_open(struct inode *inode, struct file *file)
216 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
217 sizeof(struct ipmr_vif_iter));
220 static struct file_operations ip6mr_vif_fops = {
221 .owner = THIS_MODULE,
222 .open = ip6mr_vif_open,
225 .release = seq_release_net,
228 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
230 struct net *net = seq_file_net(seq);
232 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
236 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
238 struct mfc6_cache *mfc = v;
239 struct ipmr_mfc_iter *it = seq->private;
240 struct net *net = seq_file_net(seq);
244 if (v == SEQ_START_TOKEN)
245 return ipmr_mfc_seq_idx(net, seq->private, 0);
250 if (it->cache == &mfc_unres_queue)
253 BUG_ON(it->cache != net->ipv6.mfc6_cache_array);
255 while (++it->ct < MFC6_LINES) {
256 mfc = net->ipv6.mfc6_cache_array[it->ct];
261 /* exhausted cache_array, show unresolved */
262 read_unlock(&mrt_lock);
263 it->cache = &mfc_unres_queue;
266 spin_lock_bh(&mfc_unres_lock);
267 mfc = mfc_unres_queue;
272 spin_unlock_bh(&mfc_unres_lock);
278 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
280 struct ipmr_mfc_iter *it = seq->private;
281 struct net *net = seq_file_net(seq);
283 if (it->cache == &mfc_unres_queue)
284 spin_unlock_bh(&mfc_unres_lock);
285 else if (it->cache == net->ipv6.mfc6_cache_array)
286 read_unlock(&mrt_lock);
289 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
292 struct net *net = seq_file_net(seq);
294 if (v == SEQ_START_TOKEN) {
298 "Iif Pkts Bytes Wrong Oifs\n");
300 const struct mfc6_cache *mfc = v;
301 const struct ipmr_mfc_iter *it = seq->private;
303 seq_printf(seq, "%pI6 %pI6 %-3hd",
304 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
307 if (it->cache != &mfc_unres_queue) {
308 seq_printf(seq, " %8lu %8lu %8lu",
310 mfc->mfc_un.res.bytes,
311 mfc->mfc_un.res.wrong_if);
312 for (n = mfc->mfc_un.res.minvif;
313 n < mfc->mfc_un.res.maxvif; n++) {
314 if (MIF_EXISTS(net, n) &&
315 mfc->mfc_un.res.ttls[n] < 255)
318 n, mfc->mfc_un.res.ttls[n]);
321 /* unresolved mfc_caches don't contain
322 * pkt, bytes and wrong_if values
324 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
331 static struct seq_operations ipmr_mfc_seq_ops = {
332 .start = ipmr_mfc_seq_start,
333 .next = ipmr_mfc_seq_next,
334 .stop = ipmr_mfc_seq_stop,
335 .show = ipmr_mfc_seq_show,
338 static int ipmr_mfc_open(struct inode *inode, struct file *file)
340 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
341 sizeof(struct ipmr_mfc_iter));
344 static struct file_operations ip6mr_mfc_fops = {
345 .owner = THIS_MODULE,
346 .open = ipmr_mfc_open,
349 .release = seq_release_net,
353 #ifdef CONFIG_IPV6_PIMSM_V2
355 static int pim6_rcv(struct sk_buff *skb)
357 struct pimreghdr *pim;
358 struct ipv6hdr *encap;
359 struct net_device *reg_dev = NULL;
360 struct net *net = dev_net(skb->dev);
361 int reg_vif_num = net->ipv6.mroute_reg_vif_num;
363 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
366 pim = (struct pimreghdr *)skb_transport_header(skb);
367 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
368 (pim->flags & PIM_NULL_REGISTER) ||
369 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
370 sizeof(*pim), IPPROTO_PIM,
371 csum_partial((void *)pim, sizeof(*pim), 0)) &&
372 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
375 /* check if the inner packet is destined to mcast group */
376 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
379 if (!ipv6_addr_is_multicast(&encap->daddr) ||
380 encap->payload_len == 0 ||
381 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
384 read_lock(&mrt_lock);
385 if (reg_vif_num >= 0)
386 reg_dev = net->ipv6.vif6_table[reg_vif_num].dev;
389 read_unlock(&mrt_lock);
394 skb->mac_header = skb->network_header;
395 skb_pull(skb, (u8 *)encap - skb->data);
396 skb_reset_network_header(skb);
398 skb->protocol = htons(ETH_P_IPV6);
400 skb->pkt_type = PACKET_HOST;
401 dst_release(skb->dst);
402 reg_dev->stats.rx_bytes += skb->len;
403 reg_dev->stats.rx_packets++;
414 static struct inet6_protocol pim6_protocol = {
418 /* Service routines creating virtual interfaces: PIMREG */
420 static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
422 struct net *net = dev_net(dev);
424 read_lock(&mrt_lock);
425 dev->stats.tx_bytes += skb->len;
426 dev->stats.tx_packets++;
427 ip6mr_cache_report(net, skb, net->ipv6.mroute_reg_vif_num,
429 read_unlock(&mrt_lock);
434 static const struct net_device_ops reg_vif_netdev_ops = {
435 .ndo_start_xmit = reg_vif_xmit,
438 static void reg_vif_setup(struct net_device *dev)
440 dev->type = ARPHRD_PIMREG;
441 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
442 dev->flags = IFF_NOARP;
443 dev->netdev_ops = ®_vif_netdev_ops;
444 dev->destructor = free_netdev;
447 static struct net_device *ip6mr_reg_vif(struct net *net)
449 struct net_device *dev;
451 dev = alloc_netdev(0, "pim6reg", reg_vif_setup);
455 dev_net_set(dev, net);
457 if (register_netdevice(dev)) {
470 /* allow the register to be completed before unregistering. */
474 unregister_netdevice(dev);
483 static int mif6_delete(struct net *net, int vifi)
485 struct mif_device *v;
486 struct net_device *dev;
487 struct inet6_dev *in6_dev;
488 if (vifi < 0 || vifi >= net->ipv6.maxvif)
489 return -EADDRNOTAVAIL;
491 v = &net->ipv6.vif6_table[vifi];
493 write_lock_bh(&mrt_lock);
498 write_unlock_bh(&mrt_lock);
499 return -EADDRNOTAVAIL;
502 #ifdef CONFIG_IPV6_PIMSM_V2
503 if (vifi == net->ipv6.mroute_reg_vif_num)
504 net->ipv6.mroute_reg_vif_num = -1;
507 if (vifi + 1 == net->ipv6.maxvif) {
509 for (tmp = vifi - 1; tmp >= 0; tmp--) {
510 if (MIF_EXISTS(net, tmp))
513 net->ipv6.maxvif = tmp + 1;
516 write_unlock_bh(&mrt_lock);
518 dev_set_allmulti(dev, -1);
520 in6_dev = __in6_dev_get(dev);
522 in6_dev->cnf.mc_forwarding--;
524 if (v->flags & MIFF_REGISTER)
525 unregister_netdevice(dev);
531 static inline void ip6mr_cache_free(struct mfc6_cache *c)
533 release_net(mfc6_net(c));
534 kmem_cache_free(mrt_cachep, c);
537 /* Destroy an unresolved cache entry, killing queued skbs
538 and reporting error to netlink readers.
541 static void ip6mr_destroy_unres(struct mfc6_cache *c)
544 struct net *net = mfc6_net(c);
546 atomic_dec(&net->ipv6.cache_resolve_queue_len);
548 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
549 if (ipv6_hdr(skb)->version == 0) {
550 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
551 nlh->nlmsg_type = NLMSG_ERROR;
552 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
553 skb_trim(skb, nlh->nlmsg_len);
554 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
555 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
564 /* Single timer process for all the unresolved queue. */
566 static void ipmr_do_expire_process(unsigned long dummy)
568 unsigned long now = jiffies;
569 unsigned long expires = 10 * HZ;
570 struct mfc6_cache *c, **cp;
572 cp = &mfc_unres_queue;
574 while ((c = *cp) != NULL) {
575 if (time_after(c->mfc_un.unres.expires, now)) {
577 unsigned long interval = c->mfc_un.unres.expires - now;
578 if (interval < expires)
585 ip6mr_destroy_unres(c);
588 if (mfc_unres_queue != NULL)
589 mod_timer(&ipmr_expire_timer, jiffies + expires);
592 static void ipmr_expire_process(unsigned long dummy)
594 if (!spin_trylock(&mfc_unres_lock)) {
595 mod_timer(&ipmr_expire_timer, jiffies + 1);
599 if (mfc_unres_queue != NULL)
600 ipmr_do_expire_process(dummy);
602 spin_unlock(&mfc_unres_lock);
605 /* Fill oifs list. It is called under write locked mrt_lock. */
607 static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls)
610 struct net *net = mfc6_net(cache);
612 cache->mfc_un.res.minvif = MAXMIFS;
613 cache->mfc_un.res.maxvif = 0;
614 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
616 for (vifi = 0; vifi < net->ipv6.maxvif; vifi++) {
617 if (MIF_EXISTS(net, vifi) &&
618 ttls[vifi] && ttls[vifi] < 255) {
619 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
620 if (cache->mfc_un.res.minvif > vifi)
621 cache->mfc_un.res.minvif = vifi;
622 if (cache->mfc_un.res.maxvif <= vifi)
623 cache->mfc_un.res.maxvif = vifi + 1;
628 static int mif6_add(struct net *net, struct mif6ctl *vifc, int mrtsock)
630 int vifi = vifc->mif6c_mifi;
631 struct mif_device *v = &net->ipv6.vif6_table[vifi];
632 struct net_device *dev;
633 struct inet6_dev *in6_dev;
637 if (MIF_EXISTS(net, vifi))
640 switch (vifc->mif6c_flags) {
641 #ifdef CONFIG_IPV6_PIMSM_V2
644 * Special Purpose VIF in PIM
645 * All the packets will be sent to the daemon
647 if (net->ipv6.mroute_reg_vif_num >= 0)
649 dev = ip6mr_reg_vif(net);
652 err = dev_set_allmulti(dev, 1);
654 unregister_netdevice(dev);
661 dev = dev_get_by_index(net, vifc->mif6c_pifi);
663 return -EADDRNOTAVAIL;
664 err = dev_set_allmulti(dev, 1);
674 in6_dev = __in6_dev_get(dev);
676 in6_dev->cnf.mc_forwarding++;
679 * Fill in the VIF structures
681 v->rate_limit = vifc->vifc_rate_limit;
682 v->flags = vifc->mif6c_flags;
684 v->flags |= VIFF_STATIC;
685 v->threshold = vifc->vifc_threshold;
690 v->link = dev->ifindex;
691 if (v->flags & MIFF_REGISTER)
692 v->link = dev->iflink;
694 /* And finish update writing critical data */
695 write_lock_bh(&mrt_lock);
697 #ifdef CONFIG_IPV6_PIMSM_V2
698 if (v->flags & MIFF_REGISTER)
699 net->ipv6.mroute_reg_vif_num = vifi;
701 if (vifi + 1 > net->ipv6.maxvif)
702 net->ipv6.maxvif = vifi + 1;
703 write_unlock_bh(&mrt_lock);
707 static struct mfc6_cache *ip6mr_cache_find(struct net *net,
708 struct in6_addr *origin,
709 struct in6_addr *mcastgrp)
711 int line = MFC6_HASH(mcastgrp, origin);
712 struct mfc6_cache *c;
714 for (c = net->ipv6.mfc6_cache_array[line]; c; c = c->next) {
715 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
716 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
723 * Allocate a multicast cache entry
725 static struct mfc6_cache *ip6mr_cache_alloc(struct net *net)
727 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
730 c->mfc_un.res.minvif = MAXMIFS;
731 mfc6_net_set(c, net);
735 static struct mfc6_cache *ip6mr_cache_alloc_unres(struct net *net)
737 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
740 skb_queue_head_init(&c->mfc_un.unres.unresolved);
741 c->mfc_un.unres.expires = jiffies + 10 * HZ;
742 mfc6_net_set(c, net);
747 * A cache entry has gone into a resolved state from queued
750 static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c)
755 * Play the pending entries through our router
758 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
759 if (ipv6_hdr(skb)->version == 0) {
761 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
763 if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) {
764 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
766 nlh->nlmsg_type = NLMSG_ERROR;
767 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
768 skb_trim(skb, nlh->nlmsg_len);
769 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE;
771 err = rtnl_unicast(skb, mfc6_net(uc), NETLINK_CB(skb).pid);
773 ip6_mr_forward(skb, c);
778 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
779 * expects the following bizarre scheme.
781 * Called under mrt_lock.
784 static int ip6mr_cache_report(struct net *net, struct sk_buff *pkt, mifi_t mifi,
791 #ifdef CONFIG_IPV6_PIMSM_V2
792 if (assert == MRT6MSG_WHOLEPKT)
793 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
797 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
802 /* I suppose that internal messages
803 * do not require checksums */
805 skb->ip_summed = CHECKSUM_UNNECESSARY;
807 #ifdef CONFIG_IPV6_PIMSM_V2
808 if (assert == MRT6MSG_WHOLEPKT) {
809 /* Ugly, but we have no choice with this interface.
810 Duplicate old header, fix length etc.
811 And all this only to mangle msg->im6_msgtype and
812 to set msg->im6_mbz to "mbz" :-)
814 skb_push(skb, -skb_network_offset(pkt));
816 skb_push(skb, sizeof(*msg));
817 skb_reset_transport_header(skb);
818 msg = (struct mrt6msg *)skb_transport_header(skb);
820 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
821 msg->im6_mif = net->ipv6.mroute_reg_vif_num;
823 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
824 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
826 skb->ip_summed = CHECKSUM_UNNECESSARY;
834 skb_put(skb, sizeof(struct ipv6hdr));
835 skb_reset_network_header(skb);
836 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
841 skb_put(skb, sizeof(*msg));
842 skb_reset_transport_header(skb);
843 msg = (struct mrt6msg *)skb_transport_header(skb);
846 msg->im6_msgtype = assert;
849 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
850 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
852 skb->dst = dst_clone(pkt->dst);
853 skb->ip_summed = CHECKSUM_UNNECESSARY;
856 if (net->ipv6.mroute6_sk == NULL) {
862 * Deliver to user space multicast routing algorithms
864 ret = sock_queue_rcv_skb(net->ipv6.mroute6_sk, skb);
867 printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n");
875 * Queue a packet for resolution. It gets locked cache entry!
879 ip6mr_cache_unresolved(struct net *net, mifi_t mifi, struct sk_buff *skb)
882 struct mfc6_cache *c;
884 spin_lock_bh(&mfc_unres_lock);
885 for (c = mfc_unres_queue; c; c = c->next) {
886 if (net_eq(mfc6_net(c), net) &&
887 ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
888 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr))
894 * Create a new entry if allowable
897 if (atomic_read(&net->ipv6.cache_resolve_queue_len) >= 10 ||
898 (c = ip6mr_cache_alloc_unres(net)) == NULL) {
899 spin_unlock_bh(&mfc_unres_lock);
906 * Fill in the new cache entry
909 c->mf6c_origin = ipv6_hdr(skb)->saddr;
910 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
913 * Reflect first query at pim6sd
915 err = ip6mr_cache_report(net, skb, mifi, MRT6MSG_NOCACHE);
917 /* If the report failed throw the cache entry
920 spin_unlock_bh(&mfc_unres_lock);
927 atomic_inc(&net->ipv6.cache_resolve_queue_len);
928 c->next = mfc_unres_queue;
931 ipmr_do_expire_process(1);
935 * See if we can append the packet
937 if (c->mfc_un.unres.unresolved.qlen > 3) {
941 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
945 spin_unlock_bh(&mfc_unres_lock);
950 * MFC6 cache manipulation by user space
953 static int ip6mr_mfc_delete(struct net *net, struct mf6cctl *mfc)
956 struct mfc6_cache *c, **cp;
958 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
960 for (cp = &net->ipv6.mfc6_cache_array[line];
961 (c = *cp) != NULL; cp = &c->next) {
962 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
963 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
964 write_lock_bh(&mrt_lock);
966 write_unlock_bh(&mrt_lock);
975 static int ip6mr_device_event(struct notifier_block *this,
976 unsigned long event, void *ptr)
978 struct net_device *dev = ptr;
979 struct net *net = dev_net(dev);
980 struct mif_device *v;
983 if (event != NETDEV_UNREGISTER)
986 v = &net->ipv6.vif6_table[0];
987 for (ct = 0; ct < net->ipv6.maxvif; ct++, v++) {
989 mif6_delete(net, ct);
994 static struct notifier_block ip6_mr_notifier = {
995 .notifier_call = ip6mr_device_event
999 * Setup for IP multicast routing
1002 static int __net_init ip6mr_net_init(struct net *net)
1005 net->ipv6.vif6_table = kcalloc(MAXMIFS, sizeof(struct mif_device),
1007 if (!net->ipv6.vif6_table) {
1012 /* Forwarding cache */
1013 net->ipv6.mfc6_cache_array = kcalloc(MFC6_LINES,
1014 sizeof(struct mfc6_cache *),
1016 if (!net->ipv6.mfc6_cache_array) {
1018 goto fail_mfc6_cache;
1021 #ifdef CONFIG_IPV6_PIMSM_V2
1022 net->ipv6.mroute_reg_vif_num = -1;
1025 #ifdef CONFIG_PROC_FS
1027 if (!proc_net_fops_create(net, "ip6_mr_vif", 0, &ip6mr_vif_fops))
1029 if (!proc_net_fops_create(net, "ip6_mr_cache", 0, &ip6mr_mfc_fops))
1030 goto proc_cache_fail;
1034 #ifdef CONFIG_PROC_FS
1036 proc_net_remove(net, "ip6_mr_vif");
1038 kfree(net->ipv6.mfc6_cache_array);
1041 kfree(net->ipv6.vif6_table);
1046 static void __net_exit ip6mr_net_exit(struct net *net)
1048 #ifdef CONFIG_PROC_FS
1049 proc_net_remove(net, "ip6_mr_cache");
1050 proc_net_remove(net, "ip6_mr_vif");
1052 mroute_clean_tables(net);
1053 kfree(net->ipv6.mfc6_cache_array);
1054 kfree(net->ipv6.vif6_table);
1057 static struct pernet_operations ip6mr_net_ops = {
1058 .init = ip6mr_net_init,
1059 .exit = ip6mr_net_exit,
1062 int __init ip6_mr_init(void)
1066 mrt_cachep = kmem_cache_create("ip6_mrt_cache",
1067 sizeof(struct mfc6_cache),
1068 0, SLAB_HWCACHE_ALIGN,
1073 err = register_pernet_subsys(&ip6mr_net_ops);
1075 goto reg_pernet_fail;
1077 setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0);
1078 err = register_netdevice_notifier(&ip6_mr_notifier);
1080 goto reg_notif_fail;
1083 del_timer(&ipmr_expire_timer);
1084 unregister_pernet_subsys(&ip6mr_net_ops);
1086 kmem_cache_destroy(mrt_cachep);
1090 void ip6_mr_cleanup(void)
1092 unregister_netdevice_notifier(&ip6_mr_notifier);
1093 del_timer(&ipmr_expire_timer);
1094 unregister_pernet_subsys(&ip6mr_net_ops);
1095 kmem_cache_destroy(mrt_cachep);
1098 static int ip6mr_mfc_add(struct net *net, struct mf6cctl *mfc, int mrtsock)
1101 struct mfc6_cache *uc, *c, **cp;
1102 unsigned char ttls[MAXMIFS];
1105 memset(ttls, 255, MAXMIFS);
1106 for (i = 0; i < MAXMIFS; i++) {
1107 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1112 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1114 for (cp = &net->ipv6.mfc6_cache_array[line];
1115 (c = *cp) != NULL; cp = &c->next) {
1116 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1117 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr))
1122 write_lock_bh(&mrt_lock);
1123 c->mf6c_parent = mfc->mf6cc_parent;
1124 ip6mr_update_thresholds(c, ttls);
1126 c->mfc_flags |= MFC_STATIC;
1127 write_unlock_bh(&mrt_lock);
1131 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1134 c = ip6mr_cache_alloc(net);
1138 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1139 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1140 c->mf6c_parent = mfc->mf6cc_parent;
1141 ip6mr_update_thresholds(c, ttls);
1143 c->mfc_flags |= MFC_STATIC;
1145 write_lock_bh(&mrt_lock);
1146 c->next = net->ipv6.mfc6_cache_array[line];
1147 net->ipv6.mfc6_cache_array[line] = c;
1148 write_unlock_bh(&mrt_lock);
1151 * Check to see if we resolved a queued list. If so we
1152 * need to send on the frames and tidy up.
1154 spin_lock_bh(&mfc_unres_lock);
1155 for (cp = &mfc_unres_queue; (uc = *cp) != NULL;
1157 if (net_eq(mfc6_net(uc), net) &&
1158 ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1159 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1161 atomic_dec(&net->ipv6.cache_resolve_queue_len);
1165 if (mfc_unres_queue == NULL)
1166 del_timer(&ipmr_expire_timer);
1167 spin_unlock_bh(&mfc_unres_lock);
1170 ip6mr_cache_resolve(uc, c);
1171 ip6mr_cache_free(uc);
1177 * Close the multicast socket, and clear the vif tables etc
1180 static void mroute_clean_tables(struct net *net)
1185 * Shut down all active vif entries
1187 for (i = 0; i < net->ipv6.maxvif; i++) {
1188 if (!(net->ipv6.vif6_table[i].flags & VIFF_STATIC))
1189 mif6_delete(net, i);
1195 for (i = 0; i < MFC6_LINES; i++) {
1196 struct mfc6_cache *c, **cp;
1198 cp = &net->ipv6.mfc6_cache_array[i];
1199 while ((c = *cp) != NULL) {
1200 if (c->mfc_flags & MFC_STATIC) {
1204 write_lock_bh(&mrt_lock);
1206 write_unlock_bh(&mrt_lock);
1208 ip6mr_cache_free(c);
1212 if (atomic_read(&net->ipv6.cache_resolve_queue_len) != 0) {
1213 struct mfc6_cache *c, **cp;
1215 spin_lock_bh(&mfc_unres_lock);
1216 cp = &mfc_unres_queue;
1217 while ((c = *cp) != NULL) {
1218 if (!net_eq(mfc6_net(c), net)) {
1223 ip6mr_destroy_unres(c);
1225 spin_unlock_bh(&mfc_unres_lock);
1229 static int ip6mr_sk_init(struct sock *sk)
1232 struct net *net = sock_net(sk);
1235 write_lock_bh(&mrt_lock);
1236 if (likely(net->ipv6.mroute6_sk == NULL)) {
1237 net->ipv6.mroute6_sk = sk;
1238 net->ipv6.devconf_all->mc_forwarding++;
1242 write_unlock_bh(&mrt_lock);
1249 int ip6mr_sk_done(struct sock *sk)
1252 struct net *net = sock_net(sk);
1255 if (sk == net->ipv6.mroute6_sk) {
1256 write_lock_bh(&mrt_lock);
1257 net->ipv6.mroute6_sk = NULL;
1258 net->ipv6.devconf_all->mc_forwarding--;
1259 write_unlock_bh(&mrt_lock);
1261 mroute_clean_tables(net);
1270 * Socket options and virtual interface manipulation. The whole
1271 * virtual interface system is a complete heap, but unfortunately
1272 * that's how BSD mrouted happens to think. Maybe one day with a proper
1273 * MOSPF/PIM router set up we can clean this up.
1276 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, int optlen)
1282 struct net *net = sock_net(sk);
1284 if (optname != MRT6_INIT) {
1285 if (sk != net->ipv6.mroute6_sk && !capable(CAP_NET_ADMIN))
1291 if (sk->sk_type != SOCK_RAW ||
1292 inet_sk(sk)->num != IPPROTO_ICMPV6)
1294 if (optlen < sizeof(int))
1297 return ip6mr_sk_init(sk);
1300 return ip6mr_sk_done(sk);
1303 if (optlen < sizeof(vif))
1305 if (copy_from_user(&vif, optval, sizeof(vif)))
1307 if (vif.mif6c_mifi >= MAXMIFS)
1310 ret = mif6_add(net, &vif, sk == net->ipv6.mroute6_sk);
1315 if (optlen < sizeof(mifi_t))
1317 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1320 ret = mif6_delete(net, mifi);
1325 * Manipulate the forwarding caches. These live
1326 * in a sort of kernel/user symbiosis.
1330 if (optlen < sizeof(mfc))
1332 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1335 if (optname == MRT6_DEL_MFC)
1336 ret = ip6mr_mfc_delete(net, &mfc);
1338 ret = ip6mr_mfc_add(net, &mfc,
1339 sk == net->ipv6.mroute6_sk);
1344 * Control PIM assert (to activate pim will activate assert)
1349 if (get_user(v, (int __user *)optval))
1351 net->ipv6.mroute_do_assert = !!v;
1355 #ifdef CONFIG_IPV6_PIMSM_V2
1359 if (get_user(v, (int __user *)optval))
1364 if (v != net->ipv6.mroute_do_pim) {
1365 net->ipv6.mroute_do_pim = v;
1366 net->ipv6.mroute_do_assert = v;
1367 if (net->ipv6.mroute_do_pim)
1368 ret = inet6_add_protocol(&pim6_protocol,
1371 ret = inet6_del_protocol(&pim6_protocol,
1382 * Spurious command, or MRT6_VERSION which you cannot
1386 return -ENOPROTOOPT;
1391 * Getsock opt support for the multicast routing system.
1394 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1399 struct net *net = sock_net(sk);
1405 #ifdef CONFIG_IPV6_PIMSM_V2
1407 val = net->ipv6.mroute_do_pim;
1411 val = net->ipv6.mroute_do_assert;
1414 return -ENOPROTOOPT;
1417 if (get_user(olr, optlen))
1420 olr = min_t(int, olr, sizeof(int));
1424 if (put_user(olr, optlen))
1426 if (copy_to_user(optval, &val, olr))
1432 * The IP multicast ioctl support routines.
1435 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1437 struct sioc_sg_req6 sr;
1438 struct sioc_mif_req6 vr;
1439 struct mif_device *vif;
1440 struct mfc6_cache *c;
1441 struct net *net = sock_net(sk);
1444 case SIOCGETMIFCNT_IN6:
1445 if (copy_from_user(&vr, arg, sizeof(vr)))
1447 if (vr.mifi >= net->ipv6.maxvif)
1449 read_lock(&mrt_lock);
1450 vif = &net->ipv6.vif6_table[vr.mifi];
1451 if (MIF_EXISTS(net, vr.mifi)) {
1452 vr.icount = vif->pkt_in;
1453 vr.ocount = vif->pkt_out;
1454 vr.ibytes = vif->bytes_in;
1455 vr.obytes = vif->bytes_out;
1456 read_unlock(&mrt_lock);
1458 if (copy_to_user(arg, &vr, sizeof(vr)))
1462 read_unlock(&mrt_lock);
1463 return -EADDRNOTAVAIL;
1464 case SIOCGETSGCNT_IN6:
1465 if (copy_from_user(&sr, arg, sizeof(sr)))
1468 read_lock(&mrt_lock);
1469 c = ip6mr_cache_find(net, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1471 sr.pktcnt = c->mfc_un.res.pkt;
1472 sr.bytecnt = c->mfc_un.res.bytes;
1473 sr.wrong_if = c->mfc_un.res.wrong_if;
1474 read_unlock(&mrt_lock);
1476 if (copy_to_user(arg, &sr, sizeof(sr)))
1480 read_unlock(&mrt_lock);
1481 return -EADDRNOTAVAIL;
1483 return -ENOIOCTLCMD;
1488 static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1490 IP6_INC_STATS_BH(dev_net(skb->dst->dev), ip6_dst_idev(skb->dst),
1491 IPSTATS_MIB_OUTFORWDATAGRAMS);
1492 return dst_output(skb);
1496 * Processing handlers for ip6mr_forward
1499 static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1501 struct ipv6hdr *ipv6h;
1502 struct net *net = mfc6_net(c);
1503 struct mif_device *vif = &net->ipv6.vif6_table[vifi];
1504 struct net_device *dev;
1505 struct dst_entry *dst;
1508 if (vif->dev == NULL)
1511 #ifdef CONFIG_IPV6_PIMSM_V2
1512 if (vif->flags & MIFF_REGISTER) {
1514 vif->bytes_out += skb->len;
1515 vif->dev->stats.tx_bytes += skb->len;
1516 vif->dev->stats.tx_packets++;
1517 ip6mr_cache_report(net, skb, vifi, MRT6MSG_WHOLEPKT);
1522 ipv6h = ipv6_hdr(skb);
1524 fl = (struct flowi) {
1527 { .daddr = ipv6h->daddr, }
1531 dst = ip6_route_output(net, NULL, &fl);
1535 dst_release(skb->dst);
1539 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1540 * not only before forwarding, but after forwarding on all output
1541 * interfaces. It is clear, if mrouter runs a multicasting
1542 * program, it should receive packets not depending to what interface
1543 * program is joined.
1544 * If we will not make it, the program will have to join on all
1545 * interfaces. On the other hand, multihoming host (or router, but
1546 * not mrouter) cannot join to more than one interface - it will
1547 * result in receiving multiple packets.
1552 vif->bytes_out += skb->len;
1554 /* We are about to write */
1555 /* XXX: extension headers? */
1556 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1559 ipv6h = ipv6_hdr(skb);
1562 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1564 return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev,
1565 ip6mr_forward2_finish);
1572 static int ip6mr_find_vif(struct net_device *dev)
1574 struct net *net = dev_net(dev);
1576 for (ct = net->ipv6.maxvif - 1; ct >= 0; ct--) {
1577 if (net->ipv6.vif6_table[ct].dev == dev)
1583 static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache)
1587 struct net *net = mfc6_net(cache);
1589 vif = cache->mf6c_parent;
1590 cache->mfc_un.res.pkt++;
1591 cache->mfc_un.res.bytes += skb->len;
1594 * Wrong interface: drop packet and (maybe) send PIM assert.
1596 if (net->ipv6.vif6_table[vif].dev != skb->dev) {
1599 cache->mfc_un.res.wrong_if++;
1600 true_vifi = ip6mr_find_vif(skb->dev);
1602 if (true_vifi >= 0 && net->ipv6.mroute_do_assert &&
1603 /* pimsm uses asserts, when switching from RPT to SPT,
1604 so that we cannot check that packet arrived on an oif.
1605 It is bad, but otherwise we would need to move pretty
1606 large chunk of pimd to kernel. Ough... --ANK
1608 (net->ipv6.mroute_do_pim ||
1609 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1611 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1612 cache->mfc_un.res.last_assert = jiffies;
1613 ip6mr_cache_report(net, skb, true_vifi, MRT6MSG_WRONGMIF);
1618 net->ipv6.vif6_table[vif].pkt_in++;
1619 net->ipv6.vif6_table[vif].bytes_in += skb->len;
1624 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
1625 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
1627 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1629 ip6mr_forward2(skb2, cache, psend);
1635 ip6mr_forward2(skb, cache, psend);
1646 * Multicast packets for forwarding arrive here
1649 int ip6_mr_input(struct sk_buff *skb)
1651 struct mfc6_cache *cache;
1652 struct net *net = dev_net(skb->dev);
1654 read_lock(&mrt_lock);
1655 cache = ip6mr_cache_find(net,
1656 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
1659 * No usable cache entry
1661 if (cache == NULL) {
1664 vif = ip6mr_find_vif(skb->dev);
1666 int err = ip6mr_cache_unresolved(net, vif, skb);
1667 read_unlock(&mrt_lock);
1671 read_unlock(&mrt_lock);
1676 ip6_mr_forward(skb, cache);
1678 read_unlock(&mrt_lock);
1685 ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm)
1688 struct rtnexthop *nhp;
1689 struct net *net = mfc6_net(c);
1690 struct net_device *dev = net->ipv6.vif6_table[c->mf6c_parent].dev;
1691 u8 *b = skb_tail_pointer(skb);
1692 struct rtattr *mp_head;
1695 RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex);
1697 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
1699 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
1700 if (c->mfc_un.res.ttls[ct] < 255) {
1701 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
1702 goto rtattr_failure;
1703 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
1704 nhp->rtnh_flags = 0;
1705 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
1706 nhp->rtnh_ifindex = net->ipv6.vif6_table[ct].dev->ifindex;
1707 nhp->rtnh_len = sizeof(*nhp);
1710 mp_head->rta_type = RTA_MULTIPATH;
1711 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
1712 rtm->rtm_type = RTN_MULTICAST;
1720 int ip6mr_get_route(struct net *net,
1721 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
1724 struct mfc6_cache *cache;
1725 struct rt6_info *rt = (struct rt6_info *)skb->dst;
1727 read_lock(&mrt_lock);
1728 cache = ip6mr_cache_find(net, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
1731 struct sk_buff *skb2;
1732 struct ipv6hdr *iph;
1733 struct net_device *dev;
1737 read_unlock(&mrt_lock);
1742 if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) {
1743 read_unlock(&mrt_lock);
1747 /* really correct? */
1748 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
1750 read_unlock(&mrt_lock);
1754 skb_reset_transport_header(skb2);
1756 skb_put(skb2, sizeof(struct ipv6hdr));
1757 skb_reset_network_header(skb2);
1759 iph = ipv6_hdr(skb2);
1762 iph->flow_lbl[0] = 0;
1763 iph->flow_lbl[1] = 0;
1764 iph->flow_lbl[2] = 0;
1765 iph->payload_len = 0;
1766 iph->nexthdr = IPPROTO_NONE;
1768 ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr);
1769 ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr);
1771 err = ip6mr_cache_unresolved(net, vif, skb2);
1772 read_unlock(&mrt_lock);
1777 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
1778 cache->mfc_flags |= MFC_NOTIFY;
1780 err = ip6mr_fill_mroute(skb, cache, rtm);
1781 read_unlock(&mrt_lock);