2 * Linux IPv6 multicast routing support for BSD pim6sd
3 * Based on net/ipv4/ipmr.c.
6 * LSIIT Laboratory, Strasbourg, France
9 * Copyright (C)2007,2008 USAGI/WIDE Project
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 static struct timer_list ipmr_expire_timer;
91 struct ipmr_mfc_iter {
92 struct seq_net_private p;
93 struct mfc6_cache **cache;
98 static struct mfc6_cache *ipmr_mfc_seq_idx(struct net *net,
99 struct ipmr_mfc_iter *it, loff_t pos)
101 struct mfc6_cache *mfc;
103 it->cache = net->ipv6.mfc6_cache_array;
104 read_lock(&mrt_lock);
105 for (it->ct = 0; it->ct < MFC6_LINES; it->ct++)
106 for (mfc = net->ipv6.mfc6_cache_array[it->ct];
107 mfc; mfc = mfc->next)
110 read_unlock(&mrt_lock);
112 it->cache = &mfc_unres_queue;
113 spin_lock_bh(&mfc_unres_lock);
114 for (mfc = mfc_unres_queue; mfc; mfc = mfc->next)
115 if (net_eq(mfc6_net(mfc), net) &&
118 spin_unlock_bh(&mfc_unres_lock);
128 * The /proc interfaces to multicast routing /proc/ip6_mr_cache /proc/ip6_mr_vif
131 struct ipmr_vif_iter {
132 struct seq_net_private p;
136 static struct mif_device *ip6mr_vif_seq_idx(struct net *net,
137 struct ipmr_vif_iter *iter,
140 for (iter->ct = 0; iter->ct < net->ipv6.maxvif; ++iter->ct) {
141 if (!MIF_EXISTS(net, iter->ct))
144 return &net->ipv6.vif6_table[iter->ct];
149 static void *ip6mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
152 struct net *net = seq_file_net(seq);
154 read_lock(&mrt_lock);
155 return *pos ? ip6mr_vif_seq_idx(net, seq->private, *pos - 1)
159 static void *ip6mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
161 struct ipmr_vif_iter *iter = seq->private;
162 struct net *net = seq_file_net(seq);
165 if (v == SEQ_START_TOKEN)
166 return ip6mr_vif_seq_idx(net, iter, 0);
168 while (++iter->ct < net->ipv6.maxvif) {
169 if (!MIF_EXISTS(net, iter->ct))
171 return &net->ipv6.vif6_table[iter->ct];
176 static void ip6mr_vif_seq_stop(struct seq_file *seq, void *v)
179 read_unlock(&mrt_lock);
182 static int ip6mr_vif_seq_show(struct seq_file *seq, void *v)
184 struct net *net = seq_file_net(seq);
186 if (v == SEQ_START_TOKEN) {
188 "Interface BytesIn PktsIn BytesOut PktsOut Flags\n");
190 const struct mif_device *vif = v;
191 const char *name = vif->dev ? vif->dev->name : "none";
194 "%2td %-10s %8ld %7ld %8ld %7ld %05X\n",
195 vif - net->ipv6.vif6_table,
196 name, vif->bytes_in, vif->pkt_in,
197 vif->bytes_out, vif->pkt_out,
203 static const struct seq_operations ip6mr_vif_seq_ops = {
204 .start = ip6mr_vif_seq_start,
205 .next = ip6mr_vif_seq_next,
206 .stop = ip6mr_vif_seq_stop,
207 .show = ip6mr_vif_seq_show,
210 static int ip6mr_vif_open(struct inode *inode, struct file *file)
212 return seq_open_net(inode, file, &ip6mr_vif_seq_ops,
213 sizeof(struct ipmr_vif_iter));
216 static const struct file_operations ip6mr_vif_fops = {
217 .owner = THIS_MODULE,
218 .open = ip6mr_vif_open,
221 .release = seq_release_net,
224 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
226 struct net *net = seq_file_net(seq);
228 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
232 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
234 struct mfc6_cache *mfc = v;
235 struct ipmr_mfc_iter *it = seq->private;
236 struct net *net = seq_file_net(seq);
240 if (v == SEQ_START_TOKEN)
241 return ipmr_mfc_seq_idx(net, seq->private, 0);
246 if (it->cache == &mfc_unres_queue)
249 BUG_ON(it->cache != net->ipv6.mfc6_cache_array);
251 while (++it->ct < MFC6_LINES) {
252 mfc = net->ipv6.mfc6_cache_array[it->ct];
257 /* exhausted cache_array, show unresolved */
258 read_unlock(&mrt_lock);
259 it->cache = &mfc_unres_queue;
262 spin_lock_bh(&mfc_unres_lock);
263 mfc = mfc_unres_queue;
268 spin_unlock_bh(&mfc_unres_lock);
274 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
276 struct ipmr_mfc_iter *it = seq->private;
277 struct net *net = seq_file_net(seq);
279 if (it->cache == &mfc_unres_queue)
280 spin_unlock_bh(&mfc_unres_lock);
281 else if (it->cache == net->ipv6.mfc6_cache_array)
282 read_unlock(&mrt_lock);
285 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
288 struct net *net = seq_file_net(seq);
290 if (v == SEQ_START_TOKEN) {
294 "Iif Pkts Bytes Wrong Oifs\n");
296 const struct mfc6_cache *mfc = v;
297 const struct ipmr_mfc_iter *it = seq->private;
299 seq_printf(seq, "%pI6 %pI6 %-3hd",
300 &mfc->mf6c_mcastgrp, &mfc->mf6c_origin,
303 if (it->cache != &mfc_unres_queue) {
304 seq_printf(seq, " %8lu %8lu %8lu",
306 mfc->mfc_un.res.bytes,
307 mfc->mfc_un.res.wrong_if);
308 for (n = mfc->mfc_un.res.minvif;
309 n < mfc->mfc_un.res.maxvif; n++) {
310 if (MIF_EXISTS(net, n) &&
311 mfc->mfc_un.res.ttls[n] < 255)
314 n, mfc->mfc_un.res.ttls[n]);
317 /* unresolved mfc_caches don't contain
318 * pkt, bytes and wrong_if values
320 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
327 static const struct seq_operations ipmr_mfc_seq_ops = {
328 .start = ipmr_mfc_seq_start,
329 .next = ipmr_mfc_seq_next,
330 .stop = ipmr_mfc_seq_stop,
331 .show = ipmr_mfc_seq_show,
334 static int ipmr_mfc_open(struct inode *inode, struct file *file)
336 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
337 sizeof(struct ipmr_mfc_iter));
340 static const struct file_operations ip6mr_mfc_fops = {
341 .owner = THIS_MODULE,
342 .open = ipmr_mfc_open,
345 .release = seq_release_net,
349 #ifdef CONFIG_IPV6_PIMSM_V2
351 static int pim6_rcv(struct sk_buff *skb)
353 struct pimreghdr *pim;
354 struct ipv6hdr *encap;
355 struct net_device *reg_dev = NULL;
356 struct net *net = dev_net(skb->dev);
357 int reg_vif_num = net->ipv6.mroute_reg_vif_num;
359 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
362 pim = (struct pimreghdr *)skb_transport_header(skb);
363 if (pim->type != ((PIM_VERSION << 4) | PIM_REGISTER) ||
364 (pim->flags & PIM_NULL_REGISTER) ||
365 (csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
366 sizeof(*pim), IPPROTO_PIM,
367 csum_partial((void *)pim, sizeof(*pim), 0)) &&
368 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
371 /* check if the inner packet is destined to mcast group */
372 encap = (struct ipv6hdr *)(skb_transport_header(skb) +
375 if (!ipv6_addr_is_multicast(&encap->daddr) ||
376 encap->payload_len == 0 ||
377 ntohs(encap->payload_len) + sizeof(*pim) > skb->len)
380 read_lock(&mrt_lock);
381 if (reg_vif_num >= 0)
382 reg_dev = net->ipv6.vif6_table[reg_vif_num].dev;
385 read_unlock(&mrt_lock);
390 skb->mac_header = skb->network_header;
391 skb_pull(skb, (u8 *)encap - skb->data);
392 skb_reset_network_header(skb);
394 skb->protocol = htons(ETH_P_IPV6);
396 skb->pkt_type = PACKET_HOST;
398 reg_dev->stats.rx_bytes += skb->len;
399 reg_dev->stats.rx_packets++;
409 static const struct inet6_protocol pim6_protocol = {
413 /* Service routines creating virtual interfaces: PIMREG */
415 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb,
416 struct net_device *dev)
418 struct net *net = dev_net(dev);
420 read_lock(&mrt_lock);
421 dev->stats.tx_bytes += skb->len;
422 dev->stats.tx_packets++;
423 ip6mr_cache_report(net, skb, net->ipv6.mroute_reg_vif_num,
425 read_unlock(&mrt_lock);
430 static const struct net_device_ops reg_vif_netdev_ops = {
431 .ndo_start_xmit = reg_vif_xmit,
434 static void reg_vif_setup(struct net_device *dev)
436 dev->type = ARPHRD_PIMREG;
437 dev->mtu = 1500 - sizeof(struct ipv6hdr) - 8;
438 dev->flags = IFF_NOARP;
439 dev->netdev_ops = ®_vif_netdev_ops;
440 dev->destructor = free_netdev;
441 dev->features |= NETIF_F_NETNS_LOCAL;
444 static struct net_device *ip6mr_reg_vif(struct net *net)
446 struct net_device *dev;
448 dev = alloc_netdev(0, "pim6reg", reg_vif_setup);
452 dev_net_set(dev, net);
454 if (register_netdevice(dev)) {
467 /* allow the register to be completed before unregistering. */
471 unregister_netdevice(dev);
480 static int mif6_delete(struct net *net, int vifi, struct list_head *head)
482 struct mif_device *v;
483 struct net_device *dev;
484 struct inet6_dev *in6_dev;
485 if (vifi < 0 || vifi >= net->ipv6.maxvif)
486 return -EADDRNOTAVAIL;
488 v = &net->ipv6.vif6_table[vifi];
490 write_lock_bh(&mrt_lock);
495 write_unlock_bh(&mrt_lock);
496 return -EADDRNOTAVAIL;
499 #ifdef CONFIG_IPV6_PIMSM_V2
500 if (vifi == net->ipv6.mroute_reg_vif_num)
501 net->ipv6.mroute_reg_vif_num = -1;
504 if (vifi + 1 == net->ipv6.maxvif) {
506 for (tmp = vifi - 1; tmp >= 0; tmp--) {
507 if (MIF_EXISTS(net, tmp))
510 net->ipv6.maxvif = tmp + 1;
513 write_unlock_bh(&mrt_lock);
515 dev_set_allmulti(dev, -1);
517 in6_dev = __in6_dev_get(dev);
519 in6_dev->cnf.mc_forwarding--;
521 if (v->flags & MIFF_REGISTER)
522 unregister_netdevice_queue(dev, head);
528 static inline void ip6mr_cache_free(struct mfc6_cache *c)
530 release_net(mfc6_net(c));
531 kmem_cache_free(mrt_cachep, c);
534 /* Destroy an unresolved cache entry, killing queued skbs
535 and reporting error to netlink readers.
538 static void ip6mr_destroy_unres(struct mfc6_cache *c)
541 struct net *net = mfc6_net(c);
543 atomic_dec(&net->ipv6.cache_resolve_queue_len);
545 while((skb = skb_dequeue(&c->mfc_un.unres.unresolved)) != NULL) {
546 if (ipv6_hdr(skb)->version == 0) {
547 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
548 nlh->nlmsg_type = NLMSG_ERROR;
549 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
550 skb_trim(skb, nlh->nlmsg_len);
551 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -ETIMEDOUT;
552 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
561 /* Single timer process for all the unresolved queue. */
563 static void ipmr_do_expire_process(unsigned long dummy)
565 unsigned long now = jiffies;
566 unsigned long expires = 10 * HZ;
567 struct mfc6_cache *c, **cp;
569 cp = &mfc_unres_queue;
571 while ((c = *cp) != NULL) {
572 if (time_after(c->mfc_un.unres.expires, now)) {
574 unsigned long interval = c->mfc_un.unres.expires - now;
575 if (interval < expires)
582 ip6mr_destroy_unres(c);
585 if (mfc_unres_queue != NULL)
586 mod_timer(&ipmr_expire_timer, jiffies + expires);
589 static void ipmr_expire_process(unsigned long dummy)
591 if (!spin_trylock(&mfc_unres_lock)) {
592 mod_timer(&ipmr_expire_timer, jiffies + 1);
596 if (mfc_unres_queue != NULL)
597 ipmr_do_expire_process(dummy);
599 spin_unlock(&mfc_unres_lock);
602 /* Fill oifs list. It is called under write locked mrt_lock. */
604 static void ip6mr_update_thresholds(struct mfc6_cache *cache, unsigned char *ttls)
607 struct net *net = mfc6_net(cache);
609 cache->mfc_un.res.minvif = MAXMIFS;
610 cache->mfc_un.res.maxvif = 0;
611 memset(cache->mfc_un.res.ttls, 255, MAXMIFS);
613 for (vifi = 0; vifi < net->ipv6.maxvif; vifi++) {
614 if (MIF_EXISTS(net, vifi) &&
615 ttls[vifi] && ttls[vifi] < 255) {
616 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
617 if (cache->mfc_un.res.minvif > vifi)
618 cache->mfc_un.res.minvif = vifi;
619 if (cache->mfc_un.res.maxvif <= vifi)
620 cache->mfc_un.res.maxvif = vifi + 1;
625 static int mif6_add(struct net *net, struct mif6ctl *vifc, int mrtsock)
627 int vifi = vifc->mif6c_mifi;
628 struct mif_device *v = &net->ipv6.vif6_table[vifi];
629 struct net_device *dev;
630 struct inet6_dev *in6_dev;
634 if (MIF_EXISTS(net, vifi))
637 switch (vifc->mif6c_flags) {
638 #ifdef CONFIG_IPV6_PIMSM_V2
641 * Special Purpose VIF in PIM
642 * All the packets will be sent to the daemon
644 if (net->ipv6.mroute_reg_vif_num >= 0)
646 dev = ip6mr_reg_vif(net);
649 err = dev_set_allmulti(dev, 1);
651 unregister_netdevice(dev);
658 dev = dev_get_by_index(net, vifc->mif6c_pifi);
660 return -EADDRNOTAVAIL;
661 err = dev_set_allmulti(dev, 1);
671 in6_dev = __in6_dev_get(dev);
673 in6_dev->cnf.mc_forwarding++;
676 * Fill in the VIF structures
678 v->rate_limit = vifc->vifc_rate_limit;
679 v->flags = vifc->mif6c_flags;
681 v->flags |= VIFF_STATIC;
682 v->threshold = vifc->vifc_threshold;
687 v->link = dev->ifindex;
688 if (v->flags & MIFF_REGISTER)
689 v->link = dev->iflink;
691 /* And finish update writing critical data */
692 write_lock_bh(&mrt_lock);
694 #ifdef CONFIG_IPV6_PIMSM_V2
695 if (v->flags & MIFF_REGISTER)
696 net->ipv6.mroute_reg_vif_num = vifi;
698 if (vifi + 1 > net->ipv6.maxvif)
699 net->ipv6.maxvif = vifi + 1;
700 write_unlock_bh(&mrt_lock);
704 static struct mfc6_cache *ip6mr_cache_find(struct net *net,
705 struct in6_addr *origin,
706 struct in6_addr *mcastgrp)
708 int line = MFC6_HASH(mcastgrp, origin);
709 struct mfc6_cache *c;
711 for (c = net->ipv6.mfc6_cache_array[line]; c; c = c->next) {
712 if (ipv6_addr_equal(&c->mf6c_origin, origin) &&
713 ipv6_addr_equal(&c->mf6c_mcastgrp, mcastgrp))
720 * Allocate a multicast cache entry
722 static struct mfc6_cache *ip6mr_cache_alloc(struct net *net)
724 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
727 c->mfc_un.res.minvif = MAXMIFS;
728 mfc6_net_set(c, net);
732 static struct mfc6_cache *ip6mr_cache_alloc_unres(struct net *net)
734 struct mfc6_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
737 skb_queue_head_init(&c->mfc_un.unres.unresolved);
738 c->mfc_un.unres.expires = jiffies + 10 * HZ;
739 mfc6_net_set(c, net);
744 * A cache entry has gone into a resolved state from queued
747 static void ip6mr_cache_resolve(struct mfc6_cache *uc, struct mfc6_cache *c)
752 * Play the pending entries through our router
755 while((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
756 if (ipv6_hdr(skb)->version == 0) {
758 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct ipv6hdr));
760 if (ip6mr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) {
761 nlh->nlmsg_len = skb_tail_pointer(skb) - (u8 *)nlh;
763 nlh->nlmsg_type = NLMSG_ERROR;
764 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
765 skb_trim(skb, nlh->nlmsg_len);
766 ((struct nlmsgerr *)NLMSG_DATA(nlh))->error = -EMSGSIZE;
768 err = rtnl_unicast(skb, mfc6_net(uc), NETLINK_CB(skb).pid);
770 ip6_mr_forward(skb, c);
775 * Bounce a cache query up to pim6sd. We could use netlink for this but pim6sd
776 * expects the following bizarre scheme.
778 * Called under mrt_lock.
781 static int ip6mr_cache_report(struct net *net, struct sk_buff *pkt, mifi_t mifi,
788 #ifdef CONFIG_IPV6_PIMSM_V2
789 if (assert == MRT6MSG_WHOLEPKT)
790 skb = skb_realloc_headroom(pkt, -skb_network_offset(pkt)
794 skb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(*msg), GFP_ATOMIC);
799 /* I suppose that internal messages
800 * do not require checksums */
802 skb->ip_summed = CHECKSUM_UNNECESSARY;
804 #ifdef CONFIG_IPV6_PIMSM_V2
805 if (assert == MRT6MSG_WHOLEPKT) {
806 /* Ugly, but we have no choice with this interface.
807 Duplicate old header, fix length etc.
808 And all this only to mangle msg->im6_msgtype and
809 to set msg->im6_mbz to "mbz" :-)
811 skb_push(skb, -skb_network_offset(pkt));
813 skb_push(skb, sizeof(*msg));
814 skb_reset_transport_header(skb);
815 msg = (struct mrt6msg *)skb_transport_header(skb);
817 msg->im6_msgtype = MRT6MSG_WHOLEPKT;
818 msg->im6_mif = net->ipv6.mroute_reg_vif_num;
820 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
821 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
823 skb->ip_summed = CHECKSUM_UNNECESSARY;
831 skb_put(skb, sizeof(struct ipv6hdr));
832 skb_reset_network_header(skb);
833 skb_copy_to_linear_data(skb, ipv6_hdr(pkt), sizeof(struct ipv6hdr));
838 skb_put(skb, sizeof(*msg));
839 skb_reset_transport_header(skb);
840 msg = (struct mrt6msg *)skb_transport_header(skb);
843 msg->im6_msgtype = assert;
846 ipv6_addr_copy(&msg->im6_src, &ipv6_hdr(pkt)->saddr);
847 ipv6_addr_copy(&msg->im6_dst, &ipv6_hdr(pkt)->daddr);
849 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
850 skb->ip_summed = CHECKSUM_UNNECESSARY;
853 if (net->ipv6.mroute6_sk == NULL) {
859 * Deliver to user space multicast routing algorithms
861 ret = sock_queue_rcv_skb(net->ipv6.mroute6_sk, skb);
864 printk(KERN_WARNING "mroute6: pending queue full, dropping entries.\n");
872 * Queue a packet for resolution. It gets locked cache entry!
876 ip6mr_cache_unresolved(struct net *net, mifi_t mifi, struct sk_buff *skb)
879 struct mfc6_cache *c;
881 spin_lock_bh(&mfc_unres_lock);
882 for (c = mfc_unres_queue; c; c = c->next) {
883 if (net_eq(mfc6_net(c), net) &&
884 ipv6_addr_equal(&c->mf6c_mcastgrp, &ipv6_hdr(skb)->daddr) &&
885 ipv6_addr_equal(&c->mf6c_origin, &ipv6_hdr(skb)->saddr))
891 * Create a new entry if allowable
894 if (atomic_read(&net->ipv6.cache_resolve_queue_len) >= 10 ||
895 (c = ip6mr_cache_alloc_unres(net)) == NULL) {
896 spin_unlock_bh(&mfc_unres_lock);
903 * Fill in the new cache entry
906 c->mf6c_origin = ipv6_hdr(skb)->saddr;
907 c->mf6c_mcastgrp = ipv6_hdr(skb)->daddr;
910 * Reflect first query at pim6sd
912 err = ip6mr_cache_report(net, skb, mifi, MRT6MSG_NOCACHE);
914 /* If the report failed throw the cache entry
917 spin_unlock_bh(&mfc_unres_lock);
924 atomic_inc(&net->ipv6.cache_resolve_queue_len);
925 c->next = mfc_unres_queue;
928 ipmr_do_expire_process(1);
932 * See if we can append the packet
934 if (c->mfc_un.unres.unresolved.qlen > 3) {
938 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
942 spin_unlock_bh(&mfc_unres_lock);
947 * MFC6 cache manipulation by user space
950 static int ip6mr_mfc_delete(struct net *net, struct mf6cctl *mfc)
953 struct mfc6_cache *c, **cp;
955 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
957 for (cp = &net->ipv6.mfc6_cache_array[line];
958 (c = *cp) != NULL; cp = &c->next) {
959 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
960 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr)) {
961 write_lock_bh(&mrt_lock);
963 write_unlock_bh(&mrt_lock);
972 static int ip6mr_device_event(struct notifier_block *this,
973 unsigned long event, void *ptr)
975 struct net_device *dev = ptr;
976 struct net *net = dev_net(dev);
977 struct mif_device *v;
981 if (event != NETDEV_UNREGISTER)
984 v = &net->ipv6.vif6_table[0];
985 for (ct = 0; ct < net->ipv6.maxvif; ct++, v++) {
987 mif6_delete(net, ct, &list);
989 unregister_netdevice_many(&list);
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;
1081 #ifdef CONFIG_IPV6_PIMSM_V2
1082 if (inet6_add_protocol(&pim6_protocol, IPPROTO_PIM) < 0) {
1083 printk(KERN_ERR "ip6_mr_init: can't add PIM protocol\n");
1085 goto add_proto_fail;
1089 #ifdef CONFIG_IPV6_PIMSM_V2
1091 unregister_netdevice_notifier(&ip6_mr_notifier);
1094 del_timer(&ipmr_expire_timer);
1095 unregister_pernet_subsys(&ip6mr_net_ops);
1097 kmem_cache_destroy(mrt_cachep);
1101 void ip6_mr_cleanup(void)
1103 unregister_netdevice_notifier(&ip6_mr_notifier);
1104 del_timer(&ipmr_expire_timer);
1105 unregister_pernet_subsys(&ip6mr_net_ops);
1106 kmem_cache_destroy(mrt_cachep);
1109 static int ip6mr_mfc_add(struct net *net, struct mf6cctl *mfc, int mrtsock)
1112 struct mfc6_cache *uc, *c, **cp;
1113 unsigned char ttls[MAXMIFS];
1116 if (mfc->mf6cc_parent >= MAXMIFS)
1119 memset(ttls, 255, MAXMIFS);
1120 for (i = 0; i < MAXMIFS; i++) {
1121 if (IF_ISSET(i, &mfc->mf6cc_ifset))
1126 line = MFC6_HASH(&mfc->mf6cc_mcastgrp.sin6_addr, &mfc->mf6cc_origin.sin6_addr);
1128 for (cp = &net->ipv6.mfc6_cache_array[line];
1129 (c = *cp) != NULL; cp = &c->next) {
1130 if (ipv6_addr_equal(&c->mf6c_origin, &mfc->mf6cc_origin.sin6_addr) &&
1131 ipv6_addr_equal(&c->mf6c_mcastgrp, &mfc->mf6cc_mcastgrp.sin6_addr))
1136 write_lock_bh(&mrt_lock);
1137 c->mf6c_parent = mfc->mf6cc_parent;
1138 ip6mr_update_thresholds(c, ttls);
1140 c->mfc_flags |= MFC_STATIC;
1141 write_unlock_bh(&mrt_lock);
1145 if (!ipv6_addr_is_multicast(&mfc->mf6cc_mcastgrp.sin6_addr))
1148 c = ip6mr_cache_alloc(net);
1152 c->mf6c_origin = mfc->mf6cc_origin.sin6_addr;
1153 c->mf6c_mcastgrp = mfc->mf6cc_mcastgrp.sin6_addr;
1154 c->mf6c_parent = mfc->mf6cc_parent;
1155 ip6mr_update_thresholds(c, ttls);
1157 c->mfc_flags |= MFC_STATIC;
1159 write_lock_bh(&mrt_lock);
1160 c->next = net->ipv6.mfc6_cache_array[line];
1161 net->ipv6.mfc6_cache_array[line] = c;
1162 write_unlock_bh(&mrt_lock);
1165 * Check to see if we resolved a queued list. If so we
1166 * need to send on the frames and tidy up.
1168 spin_lock_bh(&mfc_unres_lock);
1169 for (cp = &mfc_unres_queue; (uc = *cp) != NULL;
1171 if (net_eq(mfc6_net(uc), net) &&
1172 ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
1173 ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
1175 atomic_dec(&net->ipv6.cache_resolve_queue_len);
1179 if (mfc_unres_queue == NULL)
1180 del_timer(&ipmr_expire_timer);
1181 spin_unlock_bh(&mfc_unres_lock);
1184 ip6mr_cache_resolve(uc, c);
1185 ip6mr_cache_free(uc);
1191 * Close the multicast socket, and clear the vif tables etc
1194 static void mroute_clean_tables(struct net *net)
1200 * Shut down all active vif entries
1202 for (i = 0; i < net->ipv6.maxvif; i++) {
1203 if (!(net->ipv6.vif6_table[i].flags & VIFF_STATIC))
1204 mif6_delete(net, i, &list);
1206 unregister_netdevice_many(&list);
1211 for (i = 0; i < MFC6_LINES; i++) {
1212 struct mfc6_cache *c, **cp;
1214 cp = &net->ipv6.mfc6_cache_array[i];
1215 while ((c = *cp) != NULL) {
1216 if (c->mfc_flags & MFC_STATIC) {
1220 write_lock_bh(&mrt_lock);
1222 write_unlock_bh(&mrt_lock);
1224 ip6mr_cache_free(c);
1228 if (atomic_read(&net->ipv6.cache_resolve_queue_len) != 0) {
1229 struct mfc6_cache *c, **cp;
1231 spin_lock_bh(&mfc_unres_lock);
1232 cp = &mfc_unres_queue;
1233 while ((c = *cp) != NULL) {
1234 if (!net_eq(mfc6_net(c), net)) {
1239 ip6mr_destroy_unres(c);
1241 spin_unlock_bh(&mfc_unres_lock);
1245 static int ip6mr_sk_init(struct sock *sk)
1248 struct net *net = sock_net(sk);
1251 write_lock_bh(&mrt_lock);
1252 if (likely(net->ipv6.mroute6_sk == NULL)) {
1253 net->ipv6.mroute6_sk = sk;
1254 net->ipv6.devconf_all->mc_forwarding++;
1258 write_unlock_bh(&mrt_lock);
1265 int ip6mr_sk_done(struct sock *sk)
1268 struct net *net = sock_net(sk);
1271 if (sk == net->ipv6.mroute6_sk) {
1272 write_lock_bh(&mrt_lock);
1273 net->ipv6.mroute6_sk = NULL;
1274 net->ipv6.devconf_all->mc_forwarding--;
1275 write_unlock_bh(&mrt_lock);
1277 mroute_clean_tables(net);
1286 * Socket options and virtual interface manipulation. The whole
1287 * virtual interface system is a complete heap, but unfortunately
1288 * that's how BSD mrouted happens to think. Maybe one day with a proper
1289 * MOSPF/PIM router set up we can clean this up.
1292 int ip6_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1298 struct net *net = sock_net(sk);
1300 if (optname != MRT6_INIT) {
1301 if (sk != net->ipv6.mroute6_sk && !capable(CAP_NET_ADMIN))
1307 if (sk->sk_type != SOCK_RAW ||
1308 inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1310 if (optlen < sizeof(int))
1313 return ip6mr_sk_init(sk);
1316 return ip6mr_sk_done(sk);
1319 if (optlen < sizeof(vif))
1321 if (copy_from_user(&vif, optval, sizeof(vif)))
1323 if (vif.mif6c_mifi >= MAXMIFS)
1326 ret = mif6_add(net, &vif, sk == net->ipv6.mroute6_sk);
1331 if (optlen < sizeof(mifi_t))
1333 if (copy_from_user(&mifi, optval, sizeof(mifi_t)))
1336 ret = mif6_delete(net, mifi, NULL);
1341 * Manipulate the forwarding caches. These live
1342 * in a sort of kernel/user symbiosis.
1346 if (optlen < sizeof(mfc))
1348 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1351 if (optname == MRT6_DEL_MFC)
1352 ret = ip6mr_mfc_delete(net, &mfc);
1354 ret = ip6mr_mfc_add(net, &mfc,
1355 sk == net->ipv6.mroute6_sk);
1360 * Control PIM assert (to activate pim will activate assert)
1365 if (get_user(v, (int __user *)optval))
1367 net->ipv6.mroute_do_assert = !!v;
1371 #ifdef CONFIG_IPV6_PIMSM_V2
1375 if (get_user(v, (int __user *)optval))
1380 if (v != net->ipv6.mroute_do_pim) {
1381 net->ipv6.mroute_do_pim = v;
1382 net->ipv6.mroute_do_assert = v;
1390 * Spurious command, or MRT6_VERSION which you cannot
1394 return -ENOPROTOOPT;
1399 * Getsock opt support for the multicast routing system.
1402 int ip6_mroute_getsockopt(struct sock *sk, int optname, char __user *optval,
1407 struct net *net = sock_net(sk);
1413 #ifdef CONFIG_IPV6_PIMSM_V2
1415 val = net->ipv6.mroute_do_pim;
1419 val = net->ipv6.mroute_do_assert;
1422 return -ENOPROTOOPT;
1425 if (get_user(olr, optlen))
1428 olr = min_t(int, olr, sizeof(int));
1432 if (put_user(olr, optlen))
1434 if (copy_to_user(optval, &val, olr))
1440 * The IP multicast ioctl support routines.
1443 int ip6mr_ioctl(struct sock *sk, int cmd, void __user *arg)
1445 struct sioc_sg_req6 sr;
1446 struct sioc_mif_req6 vr;
1447 struct mif_device *vif;
1448 struct mfc6_cache *c;
1449 struct net *net = sock_net(sk);
1452 case SIOCGETMIFCNT_IN6:
1453 if (copy_from_user(&vr, arg, sizeof(vr)))
1455 if (vr.mifi >= net->ipv6.maxvif)
1457 read_lock(&mrt_lock);
1458 vif = &net->ipv6.vif6_table[vr.mifi];
1459 if (MIF_EXISTS(net, vr.mifi)) {
1460 vr.icount = vif->pkt_in;
1461 vr.ocount = vif->pkt_out;
1462 vr.ibytes = vif->bytes_in;
1463 vr.obytes = vif->bytes_out;
1464 read_unlock(&mrt_lock);
1466 if (copy_to_user(arg, &vr, sizeof(vr)))
1470 read_unlock(&mrt_lock);
1471 return -EADDRNOTAVAIL;
1472 case SIOCGETSGCNT_IN6:
1473 if (copy_from_user(&sr, arg, sizeof(sr)))
1476 read_lock(&mrt_lock);
1477 c = ip6mr_cache_find(net, &sr.src.sin6_addr, &sr.grp.sin6_addr);
1479 sr.pktcnt = c->mfc_un.res.pkt;
1480 sr.bytecnt = c->mfc_un.res.bytes;
1481 sr.wrong_if = c->mfc_un.res.wrong_if;
1482 read_unlock(&mrt_lock);
1484 if (copy_to_user(arg, &sr, sizeof(sr)))
1488 read_unlock(&mrt_lock);
1489 return -EADDRNOTAVAIL;
1491 return -ENOIOCTLCMD;
1496 static inline int ip6mr_forward2_finish(struct sk_buff *skb)
1498 IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
1499 IPSTATS_MIB_OUTFORWDATAGRAMS);
1500 return dst_output(skb);
1504 * Processing handlers for ip6mr_forward
1507 static int ip6mr_forward2(struct sk_buff *skb, struct mfc6_cache *c, int vifi)
1509 struct ipv6hdr *ipv6h;
1510 struct net *net = mfc6_net(c);
1511 struct mif_device *vif = &net->ipv6.vif6_table[vifi];
1512 struct net_device *dev;
1513 struct dst_entry *dst;
1516 if (vif->dev == NULL)
1519 #ifdef CONFIG_IPV6_PIMSM_V2
1520 if (vif->flags & MIFF_REGISTER) {
1522 vif->bytes_out += skb->len;
1523 vif->dev->stats.tx_bytes += skb->len;
1524 vif->dev->stats.tx_packets++;
1525 ip6mr_cache_report(net, skb, vifi, MRT6MSG_WHOLEPKT);
1530 ipv6h = ipv6_hdr(skb);
1532 fl = (struct flowi) {
1535 { .daddr = ipv6h->daddr, }
1539 dst = ip6_route_output(net, NULL, &fl);
1544 skb_dst_set(skb, dst);
1547 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1548 * not only before forwarding, but after forwarding on all output
1549 * interfaces. It is clear, if mrouter runs a multicasting
1550 * program, it should receive packets not depending to what interface
1551 * program is joined.
1552 * If we will not make it, the program will have to join on all
1553 * interfaces. On the other hand, multihoming host (or router, but
1554 * not mrouter) cannot join to more than one interface - it will
1555 * result in receiving multiple packets.
1560 vif->bytes_out += skb->len;
1562 /* We are about to write */
1563 /* XXX: extension headers? */
1564 if (skb_cow(skb, sizeof(*ipv6h) + LL_RESERVED_SPACE(dev)))
1567 ipv6h = ipv6_hdr(skb);
1570 IP6CB(skb)->flags |= IP6SKB_FORWARDED;
1572 return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dev,
1573 ip6mr_forward2_finish);
1580 static int ip6mr_find_vif(struct net_device *dev)
1582 struct net *net = dev_net(dev);
1584 for (ct = net->ipv6.maxvif - 1; ct >= 0; ct--) {
1585 if (net->ipv6.vif6_table[ct].dev == dev)
1591 static int ip6_mr_forward(struct sk_buff *skb, struct mfc6_cache *cache)
1595 struct net *net = mfc6_net(cache);
1597 vif = cache->mf6c_parent;
1598 cache->mfc_un.res.pkt++;
1599 cache->mfc_un.res.bytes += skb->len;
1602 * Wrong interface: drop packet and (maybe) send PIM assert.
1604 if (net->ipv6.vif6_table[vif].dev != skb->dev) {
1607 cache->mfc_un.res.wrong_if++;
1608 true_vifi = ip6mr_find_vif(skb->dev);
1610 if (true_vifi >= 0 && net->ipv6.mroute_do_assert &&
1611 /* pimsm uses asserts, when switching from RPT to SPT,
1612 so that we cannot check that packet arrived on an oif.
1613 It is bad, but otherwise we would need to move pretty
1614 large chunk of pimd to kernel. Ough... --ANK
1616 (net->ipv6.mroute_do_pim ||
1617 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1619 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1620 cache->mfc_un.res.last_assert = jiffies;
1621 ip6mr_cache_report(net, skb, true_vifi, MRT6MSG_WRONGMIF);
1626 net->ipv6.vif6_table[vif].pkt_in++;
1627 net->ipv6.vif6_table[vif].bytes_in += skb->len;
1632 for (ct = cache->mfc_un.res.maxvif - 1; ct >= cache->mfc_un.res.minvif; ct--) {
1633 if (ipv6_hdr(skb)->hop_limit > cache->mfc_un.res.ttls[ct]) {
1635 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1637 ip6mr_forward2(skb2, cache, psend);
1643 ip6mr_forward2(skb, cache, psend);
1654 * Multicast packets for forwarding arrive here
1657 int ip6_mr_input(struct sk_buff *skb)
1659 struct mfc6_cache *cache;
1660 struct net *net = dev_net(skb->dev);
1662 read_lock(&mrt_lock);
1663 cache = ip6mr_cache_find(net,
1664 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr);
1667 * No usable cache entry
1669 if (cache == NULL) {
1672 vif = ip6mr_find_vif(skb->dev);
1674 int err = ip6mr_cache_unresolved(net, vif, skb);
1675 read_unlock(&mrt_lock);
1679 read_unlock(&mrt_lock);
1684 ip6_mr_forward(skb, cache);
1686 read_unlock(&mrt_lock);
1693 ip6mr_fill_mroute(struct sk_buff *skb, struct mfc6_cache *c, struct rtmsg *rtm)
1696 struct rtnexthop *nhp;
1697 struct net *net = mfc6_net(c);
1698 u8 *b = skb_tail_pointer(skb);
1699 struct rtattr *mp_head;
1701 /* If cache is unresolved, don't try to parse IIF and OIF */
1702 if (c->mf6c_parent > MAXMIFS)
1705 if (MIF_EXISTS(net, c->mf6c_parent))
1706 RTA_PUT(skb, RTA_IIF, 4, &net->ipv6.vif6_table[c->mf6c_parent].dev->ifindex);
1708 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
1710 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
1711 if (MIF_EXISTS(net, ct) && c->mfc_un.res.ttls[ct] < 255) {
1712 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
1713 goto rtattr_failure;
1714 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
1715 nhp->rtnh_flags = 0;
1716 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
1717 nhp->rtnh_ifindex = net->ipv6.vif6_table[ct].dev->ifindex;
1718 nhp->rtnh_len = sizeof(*nhp);
1721 mp_head->rta_type = RTA_MULTIPATH;
1722 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
1723 rtm->rtm_type = RTN_MULTICAST;
1731 int ip6mr_get_route(struct net *net,
1732 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
1735 struct mfc6_cache *cache;
1736 struct rt6_info *rt = (struct rt6_info *)skb_dst(skb);
1738 read_lock(&mrt_lock);
1739 cache = ip6mr_cache_find(net, &rt->rt6i_src.addr, &rt->rt6i_dst.addr);
1742 struct sk_buff *skb2;
1743 struct ipv6hdr *iph;
1744 struct net_device *dev;
1748 read_unlock(&mrt_lock);
1753 if (dev == NULL || (vif = ip6mr_find_vif(dev)) < 0) {
1754 read_unlock(&mrt_lock);
1758 /* really correct? */
1759 skb2 = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
1761 read_unlock(&mrt_lock);
1765 skb_reset_transport_header(skb2);
1767 skb_put(skb2, sizeof(struct ipv6hdr));
1768 skb_reset_network_header(skb2);
1770 iph = ipv6_hdr(skb2);
1773 iph->flow_lbl[0] = 0;
1774 iph->flow_lbl[1] = 0;
1775 iph->flow_lbl[2] = 0;
1776 iph->payload_len = 0;
1777 iph->nexthdr = IPPROTO_NONE;
1779 ipv6_addr_copy(&iph->saddr, &rt->rt6i_src.addr);
1780 ipv6_addr_copy(&iph->daddr, &rt->rt6i_dst.addr);
1782 err = ip6mr_cache_unresolved(net, vif, skb2);
1783 read_unlock(&mrt_lock);
1788 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
1789 cache->mfc_flags |= MFC_NOTIFY;
1791 err = ip6mr_fill_mroute(skb, cache, rtm);
1792 read_unlock(&mrt_lock);