2 * IP multicast routing support for mrouted 3.6/3.8
4 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
5 * Linux Consultancy and Custom Driver Development
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
13 * Michael Chastain : Incorrect size of copying.
14 * Alan Cox : Added the cache manager code
15 * Alan Cox : Fixed the clone/copy bug and device race.
16 * Mike McLagan : Routing by source
17 * Malcolm Beattie : Buffer handling fixes.
18 * Alexey Kuznetsov : Double buffer free and other fixes.
19 * SVR Anand : Fixed several multicast bugs and problems.
20 * Alexey Kuznetsov : Status, optimisations and more.
21 * Brad Parker : Better behaviour on mrouted upcall
23 * Carlos Picoto : PIMv1 Support
24 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header
25 * Relax this requrement to work with older peers.
29 #include <asm/system.h>
30 #include <asm/uaccess.h>
31 #include <linux/types.h>
32 #include <linux/capability.h>
33 #include <linux/errno.h>
34 #include <linux/timer.h>
36 #include <linux/kernel.h>
37 #include <linux/fcntl.h>
38 #include <linux/stat.h>
39 #include <linux/socket.h>
41 #include <linux/inet.h>
42 #include <linux/netdevice.h>
43 #include <linux/inetdevice.h>
44 #include <linux/igmp.h>
45 #include <linux/proc_fs.h>
46 #include <linux/seq_file.h>
47 #include <linux/mroute.h>
48 #include <linux/init.h>
49 #include <linux/if_ether.h>
50 #include <net/net_namespace.h>
52 #include <net/protocol.h>
53 #include <linux/skbuff.h>
54 #include <net/route.h>
59 #include <linux/notifier.h>
60 #include <linux/if_arp.h>
61 #include <linux/netfilter_ipv4.h>
63 #include <net/checksum.h>
64 #include <net/netlink.h>
66 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
67 #define CONFIG_IP_PIMSM 1
70 /* Big lock, protecting vif table, mrt cache and mroute socket state.
71 Note that the changes are semaphored via rtnl_lock.
74 static DEFINE_RWLOCK(mrt_lock);
77 * Multicast router control variables
80 #define VIF_EXISTS(_net, _idx) ((_net)->ipv4.vif_table[_idx].dev != NULL)
82 static struct mfc_cache *mfc_unres_queue; /* Queue of unresolved entries */
84 /* Special spinlock for queue of unresolved entries */
85 static DEFINE_SPINLOCK(mfc_unres_lock);
87 /* We return to original Alan's scheme. Hash table of resolved
88 entries is changed only in process context and protected
89 with weak lock mrt_lock. Queue of unresolved entries is protected
90 with strong spinlock mfc_unres_lock.
92 In this case data path is free of exclusive locks at all.
95 static struct kmem_cache *mrt_cachep __read_mostly;
97 static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local);
98 static int ipmr_cache_report(struct net *net,
99 struct sk_buff *pkt, vifi_t vifi, int assert);
100 static int ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm);
102 static struct timer_list ipmr_expire_timer;
104 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
106 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
108 struct net *net = dev_net(dev);
112 dev = __dev_get_by_name(net, "tunl0");
114 const struct net_device_ops *ops = dev->netdev_ops;
116 struct ip_tunnel_parm p;
118 memset(&p, 0, sizeof(p));
119 p.iph.daddr = v->vifc_rmt_addr.s_addr;
120 p.iph.saddr = v->vifc_lcl_addr.s_addr;
123 p.iph.protocol = IPPROTO_IPIP;
124 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
125 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
127 if (ops->ndo_do_ioctl) {
128 mm_segment_t oldfs = get_fs();
131 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
138 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
140 struct net_device *dev;
142 dev = __dev_get_by_name(net, "tunl0");
145 const struct net_device_ops *ops = dev->netdev_ops;
148 struct ip_tunnel_parm p;
149 struct in_device *in_dev;
151 memset(&p, 0, sizeof(p));
152 p.iph.daddr = v->vifc_rmt_addr.s_addr;
153 p.iph.saddr = v->vifc_lcl_addr.s_addr;
156 p.iph.protocol = IPPROTO_IPIP;
157 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
158 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
160 if (ops->ndo_do_ioctl) {
161 mm_segment_t oldfs = get_fs();
164 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
172 (dev = __dev_get_by_name(net, p.name)) != NULL) {
173 dev->flags |= IFF_MULTICAST;
175 in_dev = __in_dev_get_rtnl(dev);
179 ipv4_devconf_setall(in_dev);
180 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
190 /* allow the register to be completed before unregistering. */
194 unregister_netdevice(dev);
198 #ifdef CONFIG_IP_PIMSM
200 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
202 struct net *net = dev_net(dev);
204 read_lock(&mrt_lock);
205 dev->stats.tx_bytes += skb->len;
206 dev->stats.tx_packets++;
207 ipmr_cache_report(net, skb, net->ipv4.mroute_reg_vif_num,
209 read_unlock(&mrt_lock);
214 static const struct net_device_ops reg_vif_netdev_ops = {
215 .ndo_start_xmit = reg_vif_xmit,
218 static void reg_vif_setup(struct net_device *dev)
220 dev->type = ARPHRD_PIMREG;
221 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
222 dev->flags = IFF_NOARP;
223 dev->netdev_ops = ®_vif_netdev_ops,
224 dev->destructor = free_netdev;
225 dev->features |= NETIF_F_NETNS_LOCAL;
228 static struct net_device *ipmr_reg_vif(struct net *net)
230 struct net_device *dev;
231 struct in_device *in_dev;
233 dev = alloc_netdev(0, "pimreg", reg_vif_setup);
238 dev_net_set(dev, net);
240 if (register_netdevice(dev)) {
247 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
252 ipv4_devconf_setall(in_dev);
253 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
264 /* allow the register to be completed before unregistering. */
268 unregister_netdevice(dev);
275 * @notify: Set to 1, if the caller is a notifier_call
278 static int vif_delete(struct net *net, int vifi, int notify,
279 struct list_head *head)
281 struct vif_device *v;
282 struct net_device *dev;
283 struct in_device *in_dev;
285 if (vifi < 0 || vifi >= net->ipv4.maxvif)
286 return -EADDRNOTAVAIL;
288 v = &net->ipv4.vif_table[vifi];
290 write_lock_bh(&mrt_lock);
295 write_unlock_bh(&mrt_lock);
296 return -EADDRNOTAVAIL;
299 #ifdef CONFIG_IP_PIMSM
300 if (vifi == net->ipv4.mroute_reg_vif_num)
301 net->ipv4.mroute_reg_vif_num = -1;
304 if (vifi+1 == net->ipv4.maxvif) {
306 for (tmp=vifi-1; tmp>=0; tmp--) {
307 if (VIF_EXISTS(net, tmp))
310 net->ipv4.maxvif = tmp+1;
313 write_unlock_bh(&mrt_lock);
315 dev_set_allmulti(dev, -1);
317 if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
318 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
319 ip_rt_multicast_event(in_dev);
322 if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER) && !notify)
323 unregister_netdevice_queue(dev, head);
329 static inline void ipmr_cache_free(struct mfc_cache *c)
331 release_net(mfc_net(c));
332 kmem_cache_free(mrt_cachep, c);
335 /* Destroy an unresolved cache entry, killing queued skbs
336 and reporting error to netlink readers.
339 static void ipmr_destroy_unres(struct mfc_cache *c)
343 struct net *net = mfc_net(c);
345 atomic_dec(&net->ipv4.cache_resolve_queue_len);
347 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
348 if (ip_hdr(skb)->version == 0) {
349 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
350 nlh->nlmsg_type = NLMSG_ERROR;
351 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
352 skb_trim(skb, nlh->nlmsg_len);
354 e->error = -ETIMEDOUT;
355 memset(&e->msg, 0, sizeof(e->msg));
357 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
366 /* Single timer process for all the unresolved queue. */
368 static void ipmr_expire_process(unsigned long dummy)
371 unsigned long expires;
372 struct mfc_cache *c, **cp;
374 if (!spin_trylock(&mfc_unres_lock)) {
375 mod_timer(&ipmr_expire_timer, jiffies+HZ/10);
379 if (mfc_unres_queue == NULL)
384 cp = &mfc_unres_queue;
386 while ((c=*cp) != NULL) {
387 if (time_after(c->mfc_un.unres.expires, now)) {
388 unsigned long interval = c->mfc_un.unres.expires - now;
389 if (interval < expires)
397 ipmr_destroy_unres(c);
400 if (mfc_unres_queue != NULL)
401 mod_timer(&ipmr_expire_timer, jiffies + expires);
404 spin_unlock(&mfc_unres_lock);
407 /* Fill oifs list. It is called under write locked mrt_lock. */
409 static void ipmr_update_thresholds(struct mfc_cache *cache, unsigned char *ttls)
412 struct net *net = mfc_net(cache);
414 cache->mfc_un.res.minvif = MAXVIFS;
415 cache->mfc_un.res.maxvif = 0;
416 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
418 for (vifi = 0; vifi < net->ipv4.maxvif; vifi++) {
419 if (VIF_EXISTS(net, vifi) &&
420 ttls[vifi] && ttls[vifi] < 255) {
421 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
422 if (cache->mfc_un.res.minvif > vifi)
423 cache->mfc_un.res.minvif = vifi;
424 if (cache->mfc_un.res.maxvif <= vifi)
425 cache->mfc_un.res.maxvif = vifi + 1;
430 static int vif_add(struct net *net, struct vifctl *vifc, int mrtsock)
432 int vifi = vifc->vifc_vifi;
433 struct vif_device *v = &net->ipv4.vif_table[vifi];
434 struct net_device *dev;
435 struct in_device *in_dev;
439 if (VIF_EXISTS(net, vifi))
442 switch (vifc->vifc_flags) {
443 #ifdef CONFIG_IP_PIMSM
446 * Special Purpose VIF in PIM
447 * All the packets will be sent to the daemon
449 if (net->ipv4.mroute_reg_vif_num >= 0)
451 dev = ipmr_reg_vif(net);
454 err = dev_set_allmulti(dev, 1);
456 unregister_netdevice(dev);
463 dev = ipmr_new_tunnel(net, vifc);
466 err = dev_set_allmulti(dev, 1);
468 ipmr_del_tunnel(dev, vifc);
474 case VIFF_USE_IFINDEX:
476 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
477 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
478 if (dev && dev->ip_ptr == NULL) {
480 return -EADDRNOTAVAIL;
483 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
486 return -EADDRNOTAVAIL;
487 err = dev_set_allmulti(dev, 1);
497 if ((in_dev = __in_dev_get_rtnl(dev)) == NULL)
498 return -EADDRNOTAVAIL;
499 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
500 ip_rt_multicast_event(in_dev);
503 * Fill in the VIF structures
505 v->rate_limit = vifc->vifc_rate_limit;
506 v->local = vifc->vifc_lcl_addr.s_addr;
507 v->remote = vifc->vifc_rmt_addr.s_addr;
508 v->flags = vifc->vifc_flags;
510 v->flags |= VIFF_STATIC;
511 v->threshold = vifc->vifc_threshold;
516 v->link = dev->ifindex;
517 if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER))
518 v->link = dev->iflink;
520 /* And finish update writing critical data */
521 write_lock_bh(&mrt_lock);
523 #ifdef CONFIG_IP_PIMSM
524 if (v->flags&VIFF_REGISTER)
525 net->ipv4.mroute_reg_vif_num = vifi;
527 if (vifi+1 > net->ipv4.maxvif)
528 net->ipv4.maxvif = vifi+1;
529 write_unlock_bh(&mrt_lock);
533 static struct mfc_cache *ipmr_cache_find(struct net *net,
537 int line = MFC_HASH(mcastgrp, origin);
540 for (c = net->ipv4.mfc_cache_array[line]; c; c = c->next) {
541 if (c->mfc_origin==origin && c->mfc_mcastgrp==mcastgrp)
548 * Allocate a multicast cache entry
550 static struct mfc_cache *ipmr_cache_alloc(struct net *net)
552 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
555 c->mfc_un.res.minvif = MAXVIFS;
560 static struct mfc_cache *ipmr_cache_alloc_unres(struct net *net)
562 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
565 skb_queue_head_init(&c->mfc_un.unres.unresolved);
566 c->mfc_un.unres.expires = jiffies + 10*HZ;
572 * A cache entry has gone into a resolved state from queued
575 static void ipmr_cache_resolve(struct mfc_cache *uc, struct mfc_cache *c)
581 * Play the pending entries through our router
584 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
585 if (ip_hdr(skb)->version == 0) {
586 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
588 if (ipmr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) {
589 nlh->nlmsg_len = (skb_tail_pointer(skb) -
592 nlh->nlmsg_type = NLMSG_ERROR;
593 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
594 skb_trim(skb, nlh->nlmsg_len);
596 e->error = -EMSGSIZE;
597 memset(&e->msg, 0, sizeof(e->msg));
600 rtnl_unicast(skb, mfc_net(c), NETLINK_CB(skb).pid);
602 ip_mr_forward(skb, c, 0);
607 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
608 * expects the following bizarre scheme.
610 * Called under mrt_lock.
613 static int ipmr_cache_report(struct net *net,
614 struct sk_buff *pkt, vifi_t vifi, int assert)
617 const int ihl = ip_hdrlen(pkt);
618 struct igmphdr *igmp;
622 #ifdef CONFIG_IP_PIMSM
623 if (assert == IGMPMSG_WHOLEPKT)
624 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
627 skb = alloc_skb(128, GFP_ATOMIC);
632 #ifdef CONFIG_IP_PIMSM
633 if (assert == IGMPMSG_WHOLEPKT) {
634 /* Ugly, but we have no choice with this interface.
635 Duplicate old header, fix ihl, length etc.
636 And all this only to mangle msg->im_msgtype and
637 to set msg->im_mbz to "mbz" :-)
639 skb_push(skb, sizeof(struct iphdr));
640 skb_reset_network_header(skb);
641 skb_reset_transport_header(skb);
642 msg = (struct igmpmsg *)skb_network_header(skb);
643 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
644 msg->im_msgtype = IGMPMSG_WHOLEPKT;
646 msg->im_vif = net->ipv4.mroute_reg_vif_num;
647 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
648 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
649 sizeof(struct iphdr));
658 skb->network_header = skb->tail;
660 skb_copy_to_linear_data(skb, pkt->data, ihl);
661 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */
662 msg = (struct igmpmsg *)skb_network_header(skb);
664 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
670 igmp=(struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
672 msg->im_msgtype = assert;
674 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
675 skb->transport_header = skb->network_header;
678 if (net->ipv4.mroute_sk == NULL) {
686 ret = sock_queue_rcv_skb(net->ipv4.mroute_sk, skb);
689 printk(KERN_WARNING "mroute: pending queue full, dropping entries.\n");
697 * Queue a packet for resolution. It gets locked cache entry!
701 ipmr_cache_unresolved(struct net *net, vifi_t vifi, struct sk_buff *skb)
705 const struct iphdr *iph = ip_hdr(skb);
707 spin_lock_bh(&mfc_unres_lock);
708 for (c=mfc_unres_queue; c; c=c->next) {
709 if (net_eq(mfc_net(c), net) &&
710 c->mfc_mcastgrp == iph->daddr &&
711 c->mfc_origin == iph->saddr)
717 * Create a new entry if allowable
720 if (atomic_read(&net->ipv4.cache_resolve_queue_len) >= 10 ||
721 (c = ipmr_cache_alloc_unres(net)) == NULL) {
722 spin_unlock_bh(&mfc_unres_lock);
729 * Fill in the new cache entry
732 c->mfc_origin = iph->saddr;
733 c->mfc_mcastgrp = iph->daddr;
736 * Reflect first query at mrouted.
738 err = ipmr_cache_report(net, skb, vifi, IGMPMSG_NOCACHE);
740 /* If the report failed throw the cache entry
743 spin_unlock_bh(&mfc_unres_lock);
750 atomic_inc(&net->ipv4.cache_resolve_queue_len);
751 c->next = mfc_unres_queue;
754 mod_timer(&ipmr_expire_timer, c->mfc_un.unres.expires);
758 * See if we can append the packet
760 if (c->mfc_un.unres.unresolved.qlen>3) {
764 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
768 spin_unlock_bh(&mfc_unres_lock);
773 * MFC cache manipulation by user space mroute daemon
776 static int ipmr_mfc_delete(struct net *net, struct mfcctl *mfc)
779 struct mfc_cache *c, **cp;
781 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
783 for (cp = &net->ipv4.mfc_cache_array[line];
784 (c = *cp) != NULL; cp = &c->next) {
785 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
786 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) {
787 write_lock_bh(&mrt_lock);
789 write_unlock_bh(&mrt_lock);
798 static int ipmr_mfc_add(struct net *net, struct mfcctl *mfc, int mrtsock)
801 struct mfc_cache *uc, *c, **cp;
803 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
805 for (cp = &net->ipv4.mfc_cache_array[line];
806 (c = *cp) != NULL; cp = &c->next) {
807 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
808 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr)
813 write_lock_bh(&mrt_lock);
814 c->mfc_parent = mfc->mfcc_parent;
815 ipmr_update_thresholds(c, mfc->mfcc_ttls);
817 c->mfc_flags |= MFC_STATIC;
818 write_unlock_bh(&mrt_lock);
822 if (!ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
825 c = ipmr_cache_alloc(net);
829 c->mfc_origin = mfc->mfcc_origin.s_addr;
830 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
831 c->mfc_parent = mfc->mfcc_parent;
832 ipmr_update_thresholds(c, mfc->mfcc_ttls);
834 c->mfc_flags |= MFC_STATIC;
836 write_lock_bh(&mrt_lock);
837 c->next = net->ipv4.mfc_cache_array[line];
838 net->ipv4.mfc_cache_array[line] = c;
839 write_unlock_bh(&mrt_lock);
842 * Check to see if we resolved a queued list. If so we
843 * need to send on the frames and tidy up.
845 spin_lock_bh(&mfc_unres_lock);
846 for (cp = &mfc_unres_queue; (uc=*cp) != NULL;
848 if (net_eq(mfc_net(uc), net) &&
849 uc->mfc_origin == c->mfc_origin &&
850 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
852 atomic_dec(&net->ipv4.cache_resolve_queue_len);
856 if (mfc_unres_queue == NULL)
857 del_timer(&ipmr_expire_timer);
858 spin_unlock_bh(&mfc_unres_lock);
861 ipmr_cache_resolve(uc, c);
868 * Close the multicast socket, and clear the vif tables etc
871 static void mroute_clean_tables(struct net *net)
877 * Shut down all active vif entries
879 for (i = 0; i < net->ipv4.maxvif; i++) {
880 if (!(net->ipv4.vif_table[i].flags&VIFF_STATIC))
881 vif_delete(net, i, 0, &list);
883 unregister_netdevice_many(&list);
888 for (i=0; i<MFC_LINES; i++) {
889 struct mfc_cache *c, **cp;
891 cp = &net->ipv4.mfc_cache_array[i];
892 while ((c = *cp) != NULL) {
893 if (c->mfc_flags&MFC_STATIC) {
897 write_lock_bh(&mrt_lock);
899 write_unlock_bh(&mrt_lock);
905 if (atomic_read(&net->ipv4.cache_resolve_queue_len) != 0) {
906 struct mfc_cache *c, **cp;
908 spin_lock_bh(&mfc_unres_lock);
909 cp = &mfc_unres_queue;
910 while ((c = *cp) != NULL) {
911 if (!net_eq(mfc_net(c), net)) {
917 ipmr_destroy_unres(c);
919 spin_unlock_bh(&mfc_unres_lock);
923 static void mrtsock_destruct(struct sock *sk)
925 struct net *net = sock_net(sk);
928 if (sk == net->ipv4.mroute_sk) {
929 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
931 write_lock_bh(&mrt_lock);
932 net->ipv4.mroute_sk = NULL;
933 write_unlock_bh(&mrt_lock);
935 mroute_clean_tables(net);
941 * Socket options and virtual interface manipulation. The whole
942 * virtual interface system is a complete heap, but unfortunately
943 * that's how BSD mrouted happens to think. Maybe one day with a proper
944 * MOSPF/PIM router set up we can clean this up.
947 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
952 struct net *net = sock_net(sk);
954 if (optname != MRT_INIT) {
955 if (sk != net->ipv4.mroute_sk && !capable(CAP_NET_ADMIN))
961 if (sk->sk_type != SOCK_RAW ||
962 inet_sk(sk)->inet_num != IPPROTO_IGMP)
964 if (optlen != sizeof(int))
968 if (net->ipv4.mroute_sk) {
973 ret = ip_ra_control(sk, 1, mrtsock_destruct);
975 write_lock_bh(&mrt_lock);
976 net->ipv4.mroute_sk = sk;
977 write_unlock_bh(&mrt_lock);
979 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
984 if (sk != net->ipv4.mroute_sk)
986 return ip_ra_control(sk, 0, NULL);
989 if (optlen != sizeof(vif))
991 if (copy_from_user(&vif, optval, sizeof(vif)))
993 if (vif.vifc_vifi >= MAXVIFS)
996 if (optname == MRT_ADD_VIF) {
997 ret = vif_add(net, &vif, sk == net->ipv4.mroute_sk);
999 ret = vif_delete(net, vif.vifc_vifi, 0, NULL);
1005 * Manipulate the forwarding caches. These live
1006 * in a sort of kernel/user symbiosis.
1010 if (optlen != sizeof(mfc))
1012 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1015 if (optname == MRT_DEL_MFC)
1016 ret = ipmr_mfc_delete(net, &mfc);
1018 ret = ipmr_mfc_add(net, &mfc, sk == net->ipv4.mroute_sk);
1022 * Control PIM assert.
1027 if (get_user(v,(int __user *)optval))
1029 net->ipv4.mroute_do_assert = (v) ? 1 : 0;
1032 #ifdef CONFIG_IP_PIMSM
1037 if (get_user(v,(int __user *)optval))
1043 if (v != net->ipv4.mroute_do_pim) {
1044 net->ipv4.mroute_do_pim = v;
1045 net->ipv4.mroute_do_assert = v;
1052 * Spurious command, or MRT_VERSION which you cannot
1056 return -ENOPROTOOPT;
1061 * Getsock opt support for the multicast routing system.
1064 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1068 struct net *net = sock_net(sk);
1070 if (optname != MRT_VERSION &&
1071 #ifdef CONFIG_IP_PIMSM
1074 optname!=MRT_ASSERT)
1075 return -ENOPROTOOPT;
1077 if (get_user(olr, optlen))
1080 olr = min_t(unsigned int, olr, sizeof(int));
1084 if (put_user(olr, optlen))
1086 if (optname == MRT_VERSION)
1088 #ifdef CONFIG_IP_PIMSM
1089 else if (optname == MRT_PIM)
1090 val = net->ipv4.mroute_do_pim;
1093 val = net->ipv4.mroute_do_assert;
1094 if (copy_to_user(optval, &val, olr))
1100 * The IP multicast ioctl support routines.
1103 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1105 struct sioc_sg_req sr;
1106 struct sioc_vif_req vr;
1107 struct vif_device *vif;
1108 struct mfc_cache *c;
1109 struct net *net = sock_net(sk);
1113 if (copy_from_user(&vr, arg, sizeof(vr)))
1115 if (vr.vifi >= net->ipv4.maxvif)
1117 read_lock(&mrt_lock);
1118 vif = &net->ipv4.vif_table[vr.vifi];
1119 if (VIF_EXISTS(net, vr.vifi)) {
1120 vr.icount = vif->pkt_in;
1121 vr.ocount = vif->pkt_out;
1122 vr.ibytes = vif->bytes_in;
1123 vr.obytes = vif->bytes_out;
1124 read_unlock(&mrt_lock);
1126 if (copy_to_user(arg, &vr, sizeof(vr)))
1130 read_unlock(&mrt_lock);
1131 return -EADDRNOTAVAIL;
1133 if (copy_from_user(&sr, arg, sizeof(sr)))
1136 read_lock(&mrt_lock);
1137 c = ipmr_cache_find(net, sr.src.s_addr, sr.grp.s_addr);
1139 sr.pktcnt = c->mfc_un.res.pkt;
1140 sr.bytecnt = c->mfc_un.res.bytes;
1141 sr.wrong_if = c->mfc_un.res.wrong_if;
1142 read_unlock(&mrt_lock);
1144 if (copy_to_user(arg, &sr, sizeof(sr)))
1148 read_unlock(&mrt_lock);
1149 return -EADDRNOTAVAIL;
1151 return -ENOIOCTLCMD;
1156 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1158 struct net_device *dev = ptr;
1159 struct net *net = dev_net(dev);
1160 struct vif_device *v;
1164 if (!net_eq(dev_net(dev), net))
1167 if (event != NETDEV_UNREGISTER)
1169 v = &net->ipv4.vif_table[0];
1170 for (ct = 0; ct < net->ipv4.maxvif; ct++, v++) {
1172 vif_delete(net, ct, 1, &list);
1174 unregister_netdevice_many(&list);
1179 static struct notifier_block ip_mr_notifier = {
1180 .notifier_call = ipmr_device_event,
1184 * Encapsulate a packet by attaching a valid IPIP header to it.
1185 * This avoids tunnel drivers and other mess and gives us the speed so
1186 * important for multicast video.
1189 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1192 struct iphdr *old_iph = ip_hdr(skb);
1194 skb_push(skb, sizeof(struct iphdr));
1195 skb->transport_header = skb->network_header;
1196 skb_reset_network_header(skb);
1200 iph->tos = old_iph->tos;
1201 iph->ttl = old_iph->ttl;
1205 iph->protocol = IPPROTO_IPIP;
1207 iph->tot_len = htons(skb->len);
1208 ip_select_ident(iph, skb_dst(skb), NULL);
1211 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1215 static inline int ipmr_forward_finish(struct sk_buff *skb)
1217 struct ip_options * opt = &(IPCB(skb)->opt);
1219 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1221 if (unlikely(opt->optlen))
1222 ip_forward_options(skb);
1224 return dst_output(skb);
1228 * Processing handlers for ipmr_forward
1231 static void ipmr_queue_xmit(struct sk_buff *skb, struct mfc_cache *c, int vifi)
1233 struct net *net = mfc_net(c);
1234 const struct iphdr *iph = ip_hdr(skb);
1235 struct vif_device *vif = &net->ipv4.vif_table[vifi];
1236 struct net_device *dev;
1240 if (vif->dev == NULL)
1243 #ifdef CONFIG_IP_PIMSM
1244 if (vif->flags & VIFF_REGISTER) {
1246 vif->bytes_out += skb->len;
1247 vif->dev->stats.tx_bytes += skb->len;
1248 vif->dev->stats.tx_packets++;
1249 ipmr_cache_report(net, skb, vifi, IGMPMSG_WHOLEPKT);
1254 if (vif->flags&VIFF_TUNNEL) {
1255 struct flowi fl = { .oif = vif->link,
1257 { .daddr = vif->remote,
1258 .saddr = vif->local,
1259 .tos = RT_TOS(iph->tos) } },
1260 .proto = IPPROTO_IPIP };
1261 if (ip_route_output_key(net, &rt, &fl))
1263 encap = sizeof(struct iphdr);
1265 struct flowi fl = { .oif = vif->link,
1267 { .daddr = iph->daddr,
1268 .tos = RT_TOS(iph->tos) } },
1269 .proto = IPPROTO_IPIP };
1270 if (ip_route_output_key(net, &rt, &fl))
1274 dev = rt->u.dst.dev;
1276 if (skb->len+encap > dst_mtu(&rt->u.dst) && (ntohs(iph->frag_off) & IP_DF)) {
1277 /* Do not fragment multicasts. Alas, IPv4 does not
1278 allow to send ICMP, so that packets will disappear
1282 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1287 encap += LL_RESERVED_SPACE(dev) + rt->u.dst.header_len;
1289 if (skb_cow(skb, encap)) {
1295 vif->bytes_out += skb->len;
1298 skb_dst_set(skb, &rt->u.dst);
1299 ip_decrease_ttl(ip_hdr(skb));
1301 /* FIXME: forward and output firewalls used to be called here.
1302 * What do we do with netfilter? -- RR */
1303 if (vif->flags & VIFF_TUNNEL) {
1304 ip_encap(skb, vif->local, vif->remote);
1305 /* FIXME: extra output firewall step used to be here. --RR */
1306 vif->dev->stats.tx_packets++;
1307 vif->dev->stats.tx_bytes += skb->len;
1310 IPCB(skb)->flags |= IPSKB_FORWARDED;
1313 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1314 * not only before forwarding, but after forwarding on all output
1315 * interfaces. It is clear, if mrouter runs a multicasting
1316 * program, it should receive packets not depending to what interface
1317 * program is joined.
1318 * If we will not make it, the program will have to join on all
1319 * interfaces. On the other hand, multihoming host (or router, but
1320 * not mrouter) cannot join to more than one interface - it will
1321 * result in receiving multiple packets.
1323 NF_HOOK(PF_INET, NF_INET_FORWARD, skb, skb->dev, dev,
1324 ipmr_forward_finish);
1332 static int ipmr_find_vif(struct net_device *dev)
1334 struct net *net = dev_net(dev);
1336 for (ct = net->ipv4.maxvif-1; ct >= 0; ct--) {
1337 if (net->ipv4.vif_table[ct].dev == dev)
1343 /* "local" means that we should preserve one skb (for local delivery) */
1345 static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local)
1349 struct net *net = mfc_net(cache);
1351 vif = cache->mfc_parent;
1352 cache->mfc_un.res.pkt++;
1353 cache->mfc_un.res.bytes += skb->len;
1356 * Wrong interface: drop packet and (maybe) send PIM assert.
1358 if (net->ipv4.vif_table[vif].dev != skb->dev) {
1361 if (skb_rtable(skb)->fl.iif == 0) {
1362 /* It is our own packet, looped back.
1363 Very complicated situation...
1365 The best workaround until routing daemons will be
1366 fixed is not to redistribute packet, if it was
1367 send through wrong interface. It means, that
1368 multicast applications WILL NOT work for
1369 (S,G), which have default multicast route pointing
1370 to wrong oif. In any case, it is not a good
1371 idea to use multicasting applications on router.
1376 cache->mfc_un.res.wrong_if++;
1377 true_vifi = ipmr_find_vif(skb->dev);
1379 if (true_vifi >= 0 && net->ipv4.mroute_do_assert &&
1380 /* pimsm uses asserts, when switching from RPT to SPT,
1381 so that we cannot check that packet arrived on an oif.
1382 It is bad, but otherwise we would need to move pretty
1383 large chunk of pimd to kernel. Ough... --ANK
1385 (net->ipv4.mroute_do_pim ||
1386 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1388 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1389 cache->mfc_un.res.last_assert = jiffies;
1390 ipmr_cache_report(net, skb, true_vifi, IGMPMSG_WRONGVIF);
1395 net->ipv4.vif_table[vif].pkt_in++;
1396 net->ipv4.vif_table[vif].bytes_in += skb->len;
1401 for (ct = cache->mfc_un.res.maxvif-1; ct >= cache->mfc_un.res.minvif; ct--) {
1402 if (ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1404 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1406 ipmr_queue_xmit(skb2, cache, psend);
1413 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1415 ipmr_queue_xmit(skb2, cache, psend);
1417 ipmr_queue_xmit(skb, cache, psend);
1430 * Multicast packets for forwarding arrive here
1433 int ip_mr_input(struct sk_buff *skb)
1435 struct mfc_cache *cache;
1436 struct net *net = dev_net(skb->dev);
1437 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1439 /* Packet is looped back after forward, it should not be
1440 forwarded second time, but still can be delivered locally.
1442 if (IPCB(skb)->flags&IPSKB_FORWARDED)
1446 if (IPCB(skb)->opt.router_alert) {
1447 if (ip_call_ra_chain(skb))
1449 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP){
1450 /* IGMPv1 (and broken IGMPv2 implementations sort of
1451 Cisco IOS <= 11.2(8)) do not put router alert
1452 option to IGMP packets destined to routable
1453 groups. It is very bad, because it means
1454 that we can forward NO IGMP messages.
1456 read_lock(&mrt_lock);
1457 if (net->ipv4.mroute_sk) {
1459 raw_rcv(net->ipv4.mroute_sk, skb);
1460 read_unlock(&mrt_lock);
1463 read_unlock(&mrt_lock);
1467 read_lock(&mrt_lock);
1468 cache = ipmr_cache_find(net, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1471 * No usable cache entry
1473 if (cache == NULL) {
1477 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1478 ip_local_deliver(skb);
1480 read_unlock(&mrt_lock);
1486 vif = ipmr_find_vif(skb->dev);
1488 int err = ipmr_cache_unresolved(net, vif, skb);
1489 read_unlock(&mrt_lock);
1493 read_unlock(&mrt_lock);
1498 ip_mr_forward(skb, cache, local);
1500 read_unlock(&mrt_lock);
1503 return ip_local_deliver(skb);
1509 return ip_local_deliver(skb);
1514 #ifdef CONFIG_IP_PIMSM
1515 static int __pim_rcv(struct sk_buff *skb, unsigned int pimlen)
1517 struct net_device *reg_dev = NULL;
1518 struct iphdr *encap;
1519 struct net *net = dev_net(skb->dev);
1521 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
1524 a. packet is really destinted to a multicast group
1525 b. packet is not a NULL-REGISTER
1526 c. packet is not truncated
1528 if (!ipv4_is_multicast(encap->daddr) ||
1529 encap->tot_len == 0 ||
1530 ntohs(encap->tot_len) + pimlen > skb->len)
1533 read_lock(&mrt_lock);
1534 if (net->ipv4.mroute_reg_vif_num >= 0)
1535 reg_dev = net->ipv4.vif_table[net->ipv4.mroute_reg_vif_num].dev;
1538 read_unlock(&mrt_lock);
1540 if (reg_dev == NULL)
1543 skb->mac_header = skb->network_header;
1544 skb_pull(skb, (u8*)encap - skb->data);
1545 skb_reset_network_header(skb);
1547 skb->protocol = htons(ETH_P_IP);
1549 skb->pkt_type = PACKET_HOST;
1551 reg_dev->stats.rx_bytes += skb->len;
1552 reg_dev->stats.rx_packets++;
1561 #ifdef CONFIG_IP_PIMSM_V1
1563 * Handle IGMP messages of PIMv1
1566 int pim_rcv_v1(struct sk_buff * skb)
1568 struct igmphdr *pim;
1569 struct net *net = dev_net(skb->dev);
1571 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1574 pim = igmp_hdr(skb);
1576 if (!net->ipv4.mroute_do_pim ||
1577 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
1580 if (__pim_rcv(skb, sizeof(*pim))) {
1588 #ifdef CONFIG_IP_PIMSM_V2
1589 static int pim_rcv(struct sk_buff * skb)
1591 struct pimreghdr *pim;
1593 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1596 pim = (struct pimreghdr *)skb_transport_header(skb);
1597 if (pim->type != ((PIM_VERSION<<4)|(PIM_REGISTER)) ||
1598 (pim->flags&PIM_NULL_REGISTER) ||
1599 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
1600 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
1603 if (__pim_rcv(skb, sizeof(*pim))) {
1612 ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm)
1615 struct rtnexthop *nhp;
1616 struct net *net = mfc_net(c);
1617 struct net_device *dev = net->ipv4.vif_table[c->mfc_parent].dev;
1618 u8 *b = skb_tail_pointer(skb);
1619 struct rtattr *mp_head;
1622 RTA_PUT(skb, RTA_IIF, 4, &dev->ifindex);
1624 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
1626 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
1627 if (c->mfc_un.res.ttls[ct] < 255) {
1628 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
1629 goto rtattr_failure;
1630 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
1631 nhp->rtnh_flags = 0;
1632 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
1633 nhp->rtnh_ifindex = net->ipv4.vif_table[ct].dev->ifindex;
1634 nhp->rtnh_len = sizeof(*nhp);
1637 mp_head->rta_type = RTA_MULTIPATH;
1638 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
1639 rtm->rtm_type = RTN_MULTICAST;
1647 int ipmr_get_route(struct net *net,
1648 struct sk_buff *skb, struct rtmsg *rtm, int nowait)
1651 struct mfc_cache *cache;
1652 struct rtable *rt = skb_rtable(skb);
1654 read_lock(&mrt_lock);
1655 cache = ipmr_cache_find(net, rt->rt_src, rt->rt_dst);
1657 if (cache == NULL) {
1658 struct sk_buff *skb2;
1660 struct net_device *dev;
1664 read_unlock(&mrt_lock);
1669 if (dev == NULL || (vif = ipmr_find_vif(dev)) < 0) {
1670 read_unlock(&mrt_lock);
1673 skb2 = skb_clone(skb, GFP_ATOMIC);
1675 read_unlock(&mrt_lock);
1679 skb_push(skb2, sizeof(struct iphdr));
1680 skb_reset_network_header(skb2);
1682 iph->ihl = sizeof(struct iphdr) >> 2;
1683 iph->saddr = rt->rt_src;
1684 iph->daddr = rt->rt_dst;
1686 err = ipmr_cache_unresolved(net, vif, skb2);
1687 read_unlock(&mrt_lock);
1691 if (!nowait && (rtm->rtm_flags&RTM_F_NOTIFY))
1692 cache->mfc_flags |= MFC_NOTIFY;
1693 err = ipmr_fill_mroute(skb, cache, rtm);
1694 read_unlock(&mrt_lock);
1698 #ifdef CONFIG_PROC_FS
1700 * The /proc interfaces to multicast routing /proc/ip_mr_cache /proc/ip_mr_vif
1702 struct ipmr_vif_iter {
1703 struct seq_net_private p;
1707 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
1708 struct ipmr_vif_iter *iter,
1711 for (iter->ct = 0; iter->ct < net->ipv4.maxvif; ++iter->ct) {
1712 if (!VIF_EXISTS(net, iter->ct))
1715 return &net->ipv4.vif_table[iter->ct];
1720 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
1721 __acquires(mrt_lock)
1723 struct net *net = seq_file_net(seq);
1725 read_lock(&mrt_lock);
1726 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
1730 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1732 struct ipmr_vif_iter *iter = seq->private;
1733 struct net *net = seq_file_net(seq);
1736 if (v == SEQ_START_TOKEN)
1737 return ipmr_vif_seq_idx(net, iter, 0);
1739 while (++iter->ct < net->ipv4.maxvif) {
1740 if (!VIF_EXISTS(net, iter->ct))
1742 return &net->ipv4.vif_table[iter->ct];
1747 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
1748 __releases(mrt_lock)
1750 read_unlock(&mrt_lock);
1753 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
1755 struct net *net = seq_file_net(seq);
1757 if (v == SEQ_START_TOKEN) {
1759 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
1761 const struct vif_device *vif = v;
1762 const char *name = vif->dev ? vif->dev->name : "none";
1765 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
1766 vif - net->ipv4.vif_table,
1767 name, vif->bytes_in, vif->pkt_in,
1768 vif->bytes_out, vif->pkt_out,
1769 vif->flags, vif->local, vif->remote);
1774 static const struct seq_operations ipmr_vif_seq_ops = {
1775 .start = ipmr_vif_seq_start,
1776 .next = ipmr_vif_seq_next,
1777 .stop = ipmr_vif_seq_stop,
1778 .show = ipmr_vif_seq_show,
1781 static int ipmr_vif_open(struct inode *inode, struct file *file)
1783 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
1784 sizeof(struct ipmr_vif_iter));
1787 static const struct file_operations ipmr_vif_fops = {
1788 .owner = THIS_MODULE,
1789 .open = ipmr_vif_open,
1791 .llseek = seq_lseek,
1792 .release = seq_release_net,
1795 struct ipmr_mfc_iter {
1796 struct seq_net_private p;
1797 struct mfc_cache **cache;
1802 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
1803 struct ipmr_mfc_iter *it, loff_t pos)
1805 struct mfc_cache *mfc;
1807 it->cache = net->ipv4.mfc_cache_array;
1808 read_lock(&mrt_lock);
1809 for (it->ct = 0; it->ct < MFC_LINES; it->ct++)
1810 for (mfc = net->ipv4.mfc_cache_array[it->ct];
1811 mfc; mfc = mfc->next)
1814 read_unlock(&mrt_lock);
1816 it->cache = &mfc_unres_queue;
1817 spin_lock_bh(&mfc_unres_lock);
1818 for (mfc = mfc_unres_queue; mfc; mfc = mfc->next)
1819 if (net_eq(mfc_net(mfc), net) &&
1822 spin_unlock_bh(&mfc_unres_lock);
1829 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
1831 struct ipmr_mfc_iter *it = seq->private;
1832 struct net *net = seq_file_net(seq);
1836 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
1840 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1842 struct mfc_cache *mfc = v;
1843 struct ipmr_mfc_iter *it = seq->private;
1844 struct net *net = seq_file_net(seq);
1848 if (v == SEQ_START_TOKEN)
1849 return ipmr_mfc_seq_idx(net, seq->private, 0);
1854 if (it->cache == &mfc_unres_queue)
1857 BUG_ON(it->cache != net->ipv4.mfc_cache_array);
1859 while (++it->ct < MFC_LINES) {
1860 mfc = net->ipv4.mfc_cache_array[it->ct];
1865 /* exhausted cache_array, show unresolved */
1866 read_unlock(&mrt_lock);
1867 it->cache = &mfc_unres_queue;
1870 spin_lock_bh(&mfc_unres_lock);
1871 mfc = mfc_unres_queue;
1872 while (mfc && !net_eq(mfc_net(mfc), net))
1878 spin_unlock_bh(&mfc_unres_lock);
1884 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
1886 struct ipmr_mfc_iter *it = seq->private;
1887 struct net *net = seq_file_net(seq);
1889 if (it->cache == &mfc_unres_queue)
1890 spin_unlock_bh(&mfc_unres_lock);
1891 else if (it->cache == net->ipv4.mfc_cache_array)
1892 read_unlock(&mrt_lock);
1895 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
1898 struct net *net = seq_file_net(seq);
1900 if (v == SEQ_START_TOKEN) {
1902 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
1904 const struct mfc_cache *mfc = v;
1905 const struct ipmr_mfc_iter *it = seq->private;
1907 seq_printf(seq, "%08lX %08lX %-3hd",
1908 (unsigned long) mfc->mfc_mcastgrp,
1909 (unsigned long) mfc->mfc_origin,
1912 if (it->cache != &mfc_unres_queue) {
1913 seq_printf(seq, " %8lu %8lu %8lu",
1914 mfc->mfc_un.res.pkt,
1915 mfc->mfc_un.res.bytes,
1916 mfc->mfc_un.res.wrong_if);
1917 for (n = mfc->mfc_un.res.minvif;
1918 n < mfc->mfc_un.res.maxvif; n++ ) {
1919 if (VIF_EXISTS(net, n) &&
1920 mfc->mfc_un.res.ttls[n] < 255)
1923 n, mfc->mfc_un.res.ttls[n]);
1926 /* unresolved mfc_caches don't contain
1927 * pkt, bytes and wrong_if values
1929 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
1931 seq_putc(seq, '\n');
1936 static const struct seq_operations ipmr_mfc_seq_ops = {
1937 .start = ipmr_mfc_seq_start,
1938 .next = ipmr_mfc_seq_next,
1939 .stop = ipmr_mfc_seq_stop,
1940 .show = ipmr_mfc_seq_show,
1943 static int ipmr_mfc_open(struct inode *inode, struct file *file)
1945 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
1946 sizeof(struct ipmr_mfc_iter));
1949 static const struct file_operations ipmr_mfc_fops = {
1950 .owner = THIS_MODULE,
1951 .open = ipmr_mfc_open,
1953 .llseek = seq_lseek,
1954 .release = seq_release_net,
1958 #ifdef CONFIG_IP_PIMSM_V2
1959 static const struct net_protocol pim_protocol = {
1967 * Setup for IP multicast routing
1969 static int __net_init ipmr_net_init(struct net *net)
1973 net->ipv4.vif_table = kcalloc(MAXVIFS, sizeof(struct vif_device),
1975 if (!net->ipv4.vif_table) {
1980 /* Forwarding cache */
1981 net->ipv4.mfc_cache_array = kcalloc(MFC_LINES,
1982 sizeof(struct mfc_cache *),
1984 if (!net->ipv4.mfc_cache_array) {
1986 goto fail_mfc_cache;
1989 #ifdef CONFIG_IP_PIMSM
1990 net->ipv4.mroute_reg_vif_num = -1;
1993 #ifdef CONFIG_PROC_FS
1995 if (!proc_net_fops_create(net, "ip_mr_vif", 0, &ipmr_vif_fops))
1997 if (!proc_net_fops_create(net, "ip_mr_cache", 0, &ipmr_mfc_fops))
1998 goto proc_cache_fail;
2002 #ifdef CONFIG_PROC_FS
2004 proc_net_remove(net, "ip_mr_vif");
2006 kfree(net->ipv4.mfc_cache_array);
2009 kfree(net->ipv4.vif_table);
2014 static void __net_exit ipmr_net_exit(struct net *net)
2016 #ifdef CONFIG_PROC_FS
2017 proc_net_remove(net, "ip_mr_cache");
2018 proc_net_remove(net, "ip_mr_vif");
2020 kfree(net->ipv4.mfc_cache_array);
2021 kfree(net->ipv4.vif_table);
2024 static struct pernet_operations ipmr_net_ops = {
2025 .init = ipmr_net_init,
2026 .exit = ipmr_net_exit,
2029 int __init ip_mr_init(void)
2033 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2034 sizeof(struct mfc_cache),
2035 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2040 err = register_pernet_subsys(&ipmr_net_ops);
2042 goto reg_pernet_fail;
2044 setup_timer(&ipmr_expire_timer, ipmr_expire_process, 0);
2045 err = register_netdevice_notifier(&ip_mr_notifier);
2047 goto reg_notif_fail;
2048 #ifdef CONFIG_IP_PIMSM_V2
2049 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2050 printk(KERN_ERR "ip_mr_init: can't add PIM protocol\n");
2052 goto add_proto_fail;
2057 #ifdef CONFIG_IP_PIMSM_V2
2059 unregister_netdevice_notifier(&ip_mr_notifier);
2062 del_timer(&ipmr_expire_timer);
2063 unregister_pernet_subsys(&ipmr_net_ops);
2065 kmem_cache_destroy(mrt_cachep);