ipmr.c

/*
 *	IP multicast routing support for mrouted 3.6/3.8
 *
 *		(c) 1995 Alan Cox, <alan@redhat.com>
 *	  Linux Consultancy and Custom Driver Development
 *
 *	This program is free software; you can redistribute it and/or
 *	modify it under the terms of the GNU General Public License
 *	as published by the Free Software Foundation; either version
 *	2 of the License, or (at your option) any later version.
 *
 *	Version: $Id: ipmr.c,v 1.65 2001/10/31 21:55:54 davem Exp $
 *
 *	Fixes:
 *	Michael Chastain	:	Incorrect size of copying.
 *	Alan Cox		:	Added the cache manager code
 *	Alan Cox		:	Fixed the clone/copy bug and device race.
 *	Mike McLagan		:	Routing by source
 *	Malcolm Beattie		:	Buffer handling fixes.
 *	Alexey Kuznetsov	:	Double buffer free and other fixes.
 *	SVR Anand		:	Fixed several multicast bugs and problems.
 *	Alexey Kuznetsov	:	Status, optimisations and more.
 *	Brad Parker		:	Better behaviour on mrouted upcall
 *					overflow.
 *      Carlos Picoto           :       PIMv1 Support
 *	Pavlin Ivanov Radoslavov:	PIMv2 Registers must checksum only PIM header
 *					Relax this requrement to work with older peers.
 *
 */

#include <asm/system.h>
#include <asm/uaccess.h>
#include <linux/types.h>
#include <linux/capability.h>
#include <linux/errno.h>
#include <linux/timer.h>
#include <linux/mm.h>
#include <linux/kernel.h>
#include <linux/fcntl.h>
#include <linux/stat.h>
#include <linux/socket.h>
#include <linux/in.h>
#include <linux/inet.h>
#include <linux/netdevice.h>
#include <linux/inetdevice.h>
#include <linux/igmp.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/mroute.h>
#include <linux/init.h>
#include <linux/if_ether.h>
#include <net/net_namespace.h>
#include <net/ip.h>
#include <net/protocol.h>
#include <linux/skbuff.h>
#include <net/route.h>
#include <net/sock.h>
#include <net/icmp.h>
#include <net/udp.h>
#include <net/raw.h>
#include <linux/notifier.h>
#include <linux/if_arp.h>
#include <linux/netfilter_ipv4.h>
#include <net/ipip.h>
#include <net/checksum.h>
#include <net/netlink.h>

#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
#define CONFIG_IP_PIMSM	1
#endif

static struct sock *mroute_socket;


/* Big lock, protecting vif table, mrt cache and mroute socket state.
   Note that the changes are semaphored via rtnl_lock.
 */

static DEFINE_RWLOCK(mrt_lock);

/*
 *	Multicast router control variables
 */

static struct vif_device vif_table[MAXVIFS];		/* Devices 		*/
static int maxvif;

#define VIF_EXISTS(idx) (vif_table[idx].dev != NULL)

static int mroute_do_assert;				/* Set in PIM assert	*/
static int mroute_do_pim;

static struct mfc_cache *mfc_cache_array[MFC_LINES];	/* Forwarding cache	*/

static struct mfc_cache *mfc_unres_queue;		/* Queue of unresolved entries */
static atomic_t cache_resolve_queue_len;		/* Size of unresolved	*/

/* Special spinlock for queue of unresolved entries */
static DEFINE_SPINLOCK(mfc_unres_lock);

/* We return to original Alan's scheme. Hash table of resolved
   entries is changed only in process context and protected
   with weak lock mrt_lock. Queue of unresolved entries is protected
   with strong spinlock mfc_unres_lock.

   In this case data path is free of exclusive locks at all.
 */

static struct kmem_cache *mrt_cachep __read_mostly;

static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local);
static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert);
static int ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm);

#ifdef CONFIG_IP_PIMSM_V2
static struct net_protocol pim_protocol;
#endif

static struct timer_list ipmr_expire_timer;

/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */

static
struct net_device *ipmr_new_tunnel(struct vifctl *v)
{
	struct net_device  *dev;

	dev = __dev_get_by_name(&init_net, "tunl0");

	if (dev) {
		int err;
		struct ifreq ifr;
		mm_segment_t	oldfs;
		struct ip_tunnel_parm p;
		struct in_device  *in_dev;

		memset(&p, 0, sizeof(p));
		p.iph.daddr = v->vifc_rmt_addr.s_addr;
		p.iph.saddr = v->vifc_lcl_addr.s_addr;
		p.iph.version = 4;
		p.iph.ihl = 5;
		p.iph.protocol = IPPROTO_IPIP;
		sprintf(p.name, "dvmrp%d", v->vifc_vifi);
		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;

		oldfs = get_fs(); set_fs(KERNEL_DS);
		err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
		set_fs(oldfs);

		dev = NULL;

		if (err == 0 && (dev = __dev_get_by_name(&init_net, p.name)) != NULL) {
			dev->flags |= IFF_MULTICAST;

			in_dev = __in_dev_get_rtnl(dev);
			if (in_dev == NULL)
				goto failure;

			ipv4_devconf_setall(in_dev);
			IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;

			if (dev_open(dev))
				goto failure;
		}
	}
	return dev;

failure:
	/* allow the register to be completed before unregistering. */
	rtnl_unlock();
	rtnl_lock();

	unregister_netdevice(dev);
	return NULL;
}

#ifdef CONFIG_IP_PIMSM

static int reg_vif_num = -1;

static int reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
{
	read_lock(&mrt_lock);
	((struct net_device_stats*)netdev_priv(dev))->tx_bytes += skb->len;
	((struct net_device_stats*)netdev_priv(dev))->tx_packets++;
	ipmr_cache_report(skb, reg_vif_num, IGMPMSG_WHOLEPKT);
	read_unlock(&mrt_lock);
	kfree_skb(skb);
	return 0;
}

static struct net_device_stats *reg_vif_get_stats(struct net_device *dev)
{
	return (struct net_device_stats*)netdev_priv(dev);
}

static void reg_vif_setup(struct net_device *dev)
{
	dev->type		= ARPHRD_PIMREG;
	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr) - 8;
	dev->flags		= IFF_NOARP;
	dev->hard_start_xmit	= reg_vif_xmit;
	dev->get_stats		= reg_vif_get_stats;
	dev->destructor		= free_netdev;
}

static struct net_device *ipmr_reg_vif(void)
{
	struct net_device *dev;
	struct in_device *in_dev;

	dev = alloc_netdev(sizeof(struct net_device_stats), "pimreg",
			   reg_vif_setup);

	if (dev == NULL)
		return NULL;

	if (register_netdevice(dev)) {
		free_netdev(dev);
		return NULL;
	}
	dev->iflink = 0;

	rcu_read_lock();
	if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
		rcu_read_unlock();
		goto failure;
	}

	ipv4_devconf_setall(in_dev);
	IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
	rcu_read_unlock();

	if (dev_open(dev))
		goto failure;

	return dev;

failure:
	/* allow the register to be completed before unregistering. */
	rtnl_unlock();
	rtnl_lock();

	unregister_netdevice(dev);
	return NULL;
}
#endif

/*
 *	Delete a VIF entry
 */

static int vif_delete(int vifi)
{
	struct vif_device *v;
	struct net_device *dev;
	struct in_device *in_dev;

	if (vifi < 0 || vifi >= maxvif)
		return -EADDRNOTAVAIL;

	v = &vif_table[vifi];

	write_lock_bh(&mrt_lock);
	dev = v->dev;
	v->dev = NULL;

	if (!dev) {
		write_unlock_bh(&mrt_lock);
		return -EADDRNOTAVAIL;
	}

#ifdef CONFIG_IP_PIMSM
	if (vifi == reg_vif_num)
		reg_vif_num = -1;
#endif

	if (vifi+1 == maxvif) {
		int tmp;
		for (tmp=vifi-1; tmp>=0; tmp--) {
			if (VIF_EXISTS(tmp))
				break;
		}
		maxvif = tmp+1;
	}

	write_unlock_bh(&mrt_lock);

	dev_set_allmulti(dev, -1);

	if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
		IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
		ip_rt_multicast_event(in_dev);
	}

	if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER))
		unregister_netdevice(dev);

	dev_put(dev);
	return 0;
}

/* Destroy an unresolved cache entry, killing queued skbs
   and reporting error to netlink readers.
 */

static void ipmr_destroy_unres(struct mfc_cache *c)
{
	struct sk_buff *skb;
	struct nlmsgerr *e;

	atomic_dec(&cache_resolve_queue_len);

	while ((skb=skb_dequeue(&c->mfc_un.unres.unresolved))) {
		if (ip_hdr(skb)->version == 0) {
			struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
			nlh->nlmsg_type = NLMSG_ERROR;
			nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
			skb_trim(skb, nlh->nlmsg_len);
			e = NLMSG_DATA(nlh);
			e->error = -ETIMEDOUT;
			memset(&e->msg, 0, sizeof(e->msg));

			rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid);
		} else
			kfree_skb(skb);
	}

	kmem_cache_free(mrt_cachep, c);
}


/* Single timer process for all the unresolved queue. */

static void ipmr_expire_process(unsigned long dummy)
{
	unsigned long now;
	unsigned long expires;
	struct mfc_cache *c, **cp;

	if (!spin_trylock(&mfc_unres_lock)) {
		mod_timer(&ipmr_expire_timer, jiffies+HZ/10);
		return;
	}

	if (atomic_read(&cache_resolve_queue_len) == 0)
		goto out;

	now = jiffies;
	expires = 10*HZ;
	cp = &mfc_unres_queue;

	while ((c=*cp) != NULL) {
		if (time_after(c->mfc_un.unres.expires, now)) {
			unsigned long interval = c->mfc_un.unres.expires - now;
			if (interval < expires)
				expires = interval;
			cp = &c->next;
			continue;
		}

		*cp = c->next;

		ipmr_destroy_unres(c);
	}

	if (atomic_read(&cache_resolve_queue_len))
		mod_timer(&ipmr_expire_timer, jiffies + expires);

out:
	spin_unlock(&mfc_unres_lock);
}

/* Fill oifs list. It is called under write locked mrt_lock. */

static void ipmr_update_thresholds(struct mfc_cache *cache, unsigned char *ttls)
{
	int vifi;

	cache->mfc_un.res.minvif = MAXVIFS;
	cache->mfc_un.res.maxvif = 0;
	memset(cache->mfc_un.res.ttls, 255, MAXVIFS);

	for (vifi=0; vifi<maxvif; vifi++) {
		if (VIF_EXISTS(vifi) && ttls[vifi] && ttls[vifi] < 255) {
			cache->mfc_un.res.ttls[vifi] = ttls[vifi];
			if (cache->mfc_un.res.minvif > vifi)
				cache->mfc_un.res.minvif = vifi;
			if (cache->mfc_un.res.maxvif <= vifi)
				cache->mfc_un.res.maxvif = vifi + 1;
		}
	}
}

static int vif_add(struct vifctl *vifc, int mrtsock)
{
	int vifi = vifc->vifc_vifi;
	struct vif_device *v = &vif_table[vifi];
	struct net_device *dev;
	struct in_device *in_dev;

	/* Is vif busy ? */
	if (VIF_EXISTS(vifi))
		return -EADDRINUSE;

	switch (vifc->vifc_flags) {
#ifdef CONFIG_IP_PIMSM
	case VIFF_REGISTER:
		/*
		 * Special Purpose VIF in PIM
		 * All the packets will be sent to the daemon
		 */
		if (reg_vif_num >= 0)
			return -EADDRINUSE;
		dev = ipmr_reg_vif();
		if (!dev)
			return -ENOBUFS;
		break;
#endif
	case VIFF_TUNNEL:
		dev = ipmr_new_tunnel(vifc);
		if (!dev)
			return -ENOBUFS;
		break;
	case 0:
		dev = ip_dev_find(&init_net, vifc->vifc_lcl_addr.s_addr);
		if (!dev)
			return -EADDRNOTAVAIL;
		dev_put(dev);
		break;
	default:
		return -EINVAL;
	}

	if ((in_dev = __in_dev_get_rtnl(dev)) == NULL)
		return -EADDRNOTAVAIL;
	IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
	dev_set_allmulti(dev, +1);
	ip_rt_multicast_event(in_dev);

	/*
	 *	Fill in the VIF structures
	 */
	v->rate_limit=vifc->vifc_rate_limit;
	v->local=vifc->vifc_lcl_addr.s_addr;
	v->remote=vifc->vifc_rmt_addr.s_addr;
	v->flags=vifc->vifc_flags;
	if (!mrtsock)
		v->flags |= VIFF_STATIC;
	v->threshold=vifc->vifc_threshold;
	v->bytes_in = 0;
	v->bytes_out = 0;
	v->pkt_in = 0;
	v->pkt_out = 0;
	v->link = dev->ifindex;
	if (v->flags&(VIFF_TUNNEL|VIFF_REGISTER))
		v->link = dev->iflink;

	/* And finish update writing critical data */
	write_lock_bh(&mrt_lock);
	dev_hold(dev);
	v->dev=dev;
#ifdef CONFIG_IP_PIMSM
	if (v->flags&VIFF_REGISTER)
		reg_vif_num = vifi;
#endif
	if (vifi+1 > maxvif)
		maxvif = vifi+1;
	write_unlock_bh(&mrt_lock);
	return 0;
}

static struct mfc_cache *ipmr_cache_find(__be32 origin, __be32 mcastgrp)
{
	int line=MFC_HASH(mcastgrp,origin);
	struct mfc_cache *c;

	for (c=mfc_cache_array[line]; c; c = c->next) {
		if (c->mfc_origin==origin && c->mfc_mcastgrp==mcastgrp)
			break;
	}
	return c;
}

/*
 *	Allocate a multicast cache entry
 */
static struct mfc_cache *ipmr_cache_alloc(void)
{
	struct mfc_cache *c=kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
	if (c==NULL)
		return NULL;
	c->mfc_un.res.minvif = MAXVIFS;
	return c;
}

static struct mfc_cache *ipmr_cache_alloc_unres(void)
{
	struct mfc_cache *c=kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
	if (c==NULL)
		return NULL;
	skb_queue_head_init(&c->mfc_un.unres.unresolved);
	c->mfc_un.unres.expires = jiffies + 10*HZ;
	return c;
}

/*
 *	A cache entry has gone into a resolved state from queued
 */

static void ipmr_cache_resolve(struct mfc_cache *uc, struct mfc_cache *c)
{
	struct sk_buff *skb;
	struct nlmsgerr *e;

	/*
	 *	Play the pending entries through our router
	 */

	while ((skb=__skb_dequeue(&uc->mfc_un.unres.unresolved))) {
		if (ip_hdr(skb)->version == 0) {
			struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));

			if (ipmr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) {
				nlh->nlmsg_len = (skb_tail_pointer(skb) -
						  (u8 *)nlh);
			} else {
				nlh->nlmsg_type = NLMSG_ERROR;
				nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
				skb_trim(skb, nlh->nlmsg_len);
				e = NLMSG_DATA(nlh);
				e->error = -EMSGSIZE;
				memset(&e->msg, 0, sizeof(e->msg));
			}

			rtnl_unicast(skb, &init_net, NETLINK_CB(skb).pid);
		} else
			ip_mr_forward(skb, c, 0);
	}
}

/*
 *	Bounce a cache query up to mrouted. We could use netlink for this but mrouted
 *	expects the following bizarre scheme.
 *
 *	Called under mrt_lock.
 */

static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert)
{
	struct sk_buff *skb;
	const int ihl = ip_hdrlen(pkt);
	struct igmphdr *igmp;
	struct igmpmsg *msg;
	int ret;

#ifdef CONFIG_IP_PIMSM
	if (assert == IGMPMSG_WHOLEPKT)
		skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
	else
#endif
		skb = alloc_skb(128, GFP_ATOMIC);

	if (!skb)
		return -ENOBUFS;

#ifdef CONFIG_IP_PIMSM
	if (assert == IGMPMSG_WHOLEPKT) {
		/* Ugly, but we have no choice with this interface.
		   Duplicate old header, fix ihl, length etc.
		   And all this only to mangle msg->im_msgtype and
		   to set msg->im_mbz to "mbz" :-)
		 */
		skb_push(skb, sizeof(struct iphdr));
		skb_reset_network_header(skb);
		skb_reset_transport_header(skb);
		msg = (struct igmpmsg *)skb_network_header(skb);
		memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
		msg->im_msgtype = IGMPMSG_WHOLEPKT;
		msg->im_mbz = 0;
		msg->im_vif = reg_vif_num;
		ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
		ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
					     sizeof(struct iphdr));
	} else
#endif
	{

	/*
	 *	Copy the IP header
	 */

	skb->network_header = skb->tail;
	skb_put(skb, ihl);
	skb_copy_to_linear_data(skb, pkt->data, ihl);
	ip_hdr(skb)->protocol = 0;			/* Flag to the kernel this is a route add */
	msg = (struct igmpmsg *)skb_network_header(skb);
	msg->im_vif = vifi;
	skb->dst = dst_clone(pkt->dst);

	/*
	 *	Add our header
	 */

	igmp=(struct igmphdr *)skb_put(skb,sizeof(struct igmphdr));
	igmp->type	=
	msg->im_msgtype = assert;
	igmp->code 	=	0;
	ip_hdr(skb)->tot_len = htons(skb->len);			/* Fix the length */
	skb->transport_header = skb->network_header;
	}

	if (mroute_socket == NULL) {
		kfree_skb(skb);
		return -EINVAL;
	}

	/*
	 *	Deliver to mrouted
	 */
	if ((ret=sock_queue_rcv_skb(mroute_socket,skb))<0) {
		if (net_ratelimit())
			printk(KERN_WARNING "mroute: pending queue full, dropping entries.\n");
		kfree_skb(skb);
	}

	return ret;
}

/*
 *	Queue a packet for resolution. It gets locked cache entry!
 */

static int
ipmr_cache_unresolved(vifi_t vifi, struct sk_buff *skb)
{
	int err;
	struct mfc_cache *c;
	const struct iphdr *iph = ip_hdr(skb);

	spin_lock_bh(&mfc_unres_lock);
	for (c=mfc_unres_queue; c; c=c->next) {
		if (c->mfc_mcastgrp == iph->daddr &&
		    c->mfc_origin == iph->saddr)
			break;
	}

	if (c == NULL) {
		/*
		 *	Create a new entry if allowable
		 */

		if (atomic_read(&cache_resolve_queue_len)>=10 ||
		    (c=ipmr_cache_alloc_unres())==NULL) {
			spin_unlock_bh(&mfc_unres_lock);

			kfree_skb(skb);
			return -ENOBUFS;
		}

		/*
		 *	Fill in the new cache entry
		 */
		c->mfc_parent	= -1;
		c->mfc_origin	= iph->saddr;
		c->mfc_mcastgrp	= iph->daddr;

		/*
		 *	Reflect first query at mrouted.
		 */
		if ((err = ipmr_cache_report(skb, vifi, IGMPMSG_NOCACHE))<0) {
			/* If the report failed throw the cache entry
			   out - Brad Parker
			 */
			spin_unlock_bh(&mfc_unres_lock);

			kmem_cache_free(mrt_cachep, c);
			kfree_skb(skb);
			return err;
		}

		atomic_inc(&cache_resolve_queue_len);
		c->next = mfc_unres_queue;
		mfc_unres_queue = c;

		mod_timer(&ipmr_expire_timer, c->mfc_un.unres.expires);
	}

	/*
	 *	See if we can append the packet
	 */
	if (c->mfc_un.unres.unresolved.qlen>3) {
		kfree_skb(skb);
		err = -ENOBUFS;
	} else {
		skb_queue_tail(&c->mfc_un.unres.unresolved,skb);
		err = 0;
	}

	spin_unlock_bh(&mfc_unres_lock);
	return err;
}

/*
 *	MFC cache manipulation by user space mroute daemon
 */

static int ipmr_mfc_delete(struct mfcctl *mfc)
{
	int line;
	struct mfc_cache *c, **cp;

	line=MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);

	for (cp=&mfc_cache_array[line]; (c=*cp) != NULL; cp = &c->next) {
		if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
		    c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) {
			write_lock_bh(&mrt_lock);
			*cp = c->next;
			write_unlock_bh(&mrt_lock);

			kmem_cache_free(mrt_cachep, c);
			return 0;
		}
	}
	return -ENOENT;
}

static int ipmr_mfc_add(struct mfcctl *mfc, int mrtsock)
{
	int line;
	struct mfc_cache *uc, *c, **cp;

	line=MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);

	for (cp=&mfc_cache_array[line]; (c=*cp) != NULL; cp = &c->next) {
		if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
		    c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr)
			break;
	}

	if (c != NULL) {
		write_lock_bh(&mrt_lock);
		c->mfc_parent = mfc->mfcc_parent;
		ipmr_update_thresholds(c, mfc->mfcc_ttls);
		if (!mrtsock)
			c->mfc_flags |= MFC_STATIC;
		write_unlock_bh(&mrt_lock);
		return 0;
	}

	if (!ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
		return -EINVAL;

	c=ipmr_cache_alloc();
	if (c==NULL)
		return -ENOMEM;

	c->mfc_origin=mfc->mfcc_origin.s_addr;
	c->mfc_mcastgrp=mfc->mfcc_mcastgrp.s_addr;
	c->mfc_parent=mfc->mfcc_parent;
	ipmr_update_thresholds(c, mfc->mfcc_ttls);
	if (!mrtsock)
		c->mfc_flags |= MFC_STATIC;

	write_lock_bh(&mrt_lock);
	c->next = mfc_cache_array[line];
	mfc_cache_array[line] = c;
	write_unlock_bh(&mrt_lock);

	/*
	 *	Check to see if we resolved a queued list. If so we
	 *	need to send on the frames and tidy up.
	 */
	spin_lock_bh(&mfc_unres_lock);
	for (cp = &mfc_unres_queue; (uc=*cp) != NULL;
	     cp = &uc->next) {
		if (uc->mfc_origin == c->mfc_origin &&
		    uc->mfc_mcastgrp == c->mfc_mcastgrp) {
			*cp = uc->next;
			if (atomic_dec_and_test(&cache_resolve_queue_len))
				del_timer(&ipmr_expire_timer);
			break;
		}
	}
	spin_unlock_bh(&mfc_unres_lock);

	if (uc) {
		ipmr_cache_resolve(uc, c);
		kmem_cache_free(mrt_cachep, uc);
	}
	return 0;
}

/*
 *	Close the multicast socket, and clear the vif tables etc
 */

static void mroute_clean_tables(struct sock *sk)
{
	int i;

	/*
	 *	Shut down all active vif entries
	 */
	for (i=0; i<maxvif; i++) {
		if (!(vif_table[i].flags&VIFF_STATIC))
			vif_delete(i);
	}

	/*
	 *	Wipe the cache
	 */
	for (i=0;i<MFC_LINES;i++) {
		struct mfc_cache *c, **cp;

		cp = &mfc_cache_array[i];
		while ((c = *cp) != NULL) {
			if (c->mfc_flags&MFC_STATIC) {
				cp = &c->next;
				continue;
			}
			write_lock_bh(&mrt_lock);
			*cp = c->next;
			write_unlock_bh(&mrt_lock);

			kmem_cache_free(mrt_cachep, c);
		}
	}

	if (atomic_read(&cache_resolve_queue_len) != 0) {
		struct mfc_cache *c;

		spin_lock_bh(&mfc_unres_lock);
		while (mfc_unres_queue != NULL) {
			c = mfc_unres_queue;
			mfc_unres_queue = c->next;
			spin_unlock_bh(&mfc_unres_lock);

			ipmr_destroy_unres(c);

			spin_lock_bh(&mfc_unres_lock);
		}
		spin_unlock_bh(&mfc_unres_lock);
	}
}

static void mrtsock_destruct(struct sock *sk)
{
	rtnl_lock();
	if (sk == mroute_socket) {
		IPV4_DEVCONF_ALL(sk->sk_net, MC_FORWARDING)--;

		write_lock_bh(&mrt_lock);
		mroute_socket=NULL;
		write_unlock_bh(&mrt_lock);

		mroute_clean_tables(sk);
	}
	rtnl_unlock();
}

/*
 *	Socket options and virtual interface manipulation. The whole
 *	virtual interface system is a complete heap, but unfortunately
 *	that's how BSD mrouted happens to think. Maybe one day with a proper
 *	MOSPF/PIM router set up we can clean this up.
 */

int ip_mroute_setsockopt(struct sock *sk,int optname,char __user *optval,int optlen)
{
	int ret;
	struct vifctl vif;
	struct mfcctl mfc;

	if (optname != MRT_INIT) {
		if (sk != mroute_socket && !capable(CAP_NET_ADMIN))
			return -EACCES;
	}

	switch (optname) {
	case MRT_INIT:
		if (sk->sk_type != SOCK_RAW ||
		    inet_sk(sk)->num != IPPROTO_IGMP)
			return -EOPNOTSUPP;
		if (optlen!=sizeof(int))
			return -ENOPROTOOPT;

		rtnl_lock();
		if (mroute_socket) {
			rtnl_unlock();
			return -EADDRINUSE;
		}

		ret = ip_ra_control(sk, 1, mrtsock_destruct);
		if (ret == 0) {
			write_lock_bh(&mrt_lock);
			mroute_socket=sk;
			write_unlock_bh(&mrt_lock);

			IPV4_DEVCONF_ALL(sk->sk_net, MC_FORWARDING)++;
		}
		rtnl_unlock();
		return ret;
	case MRT_DONE:
		if (sk!=mroute_socket)
			return -EACCES;
		return ip_ra_control(sk, 0, NULL);
	case MRT_ADD_VIF:
	case MRT_DEL_VIF:
		if (optlen!=sizeof(vif))
			return -EINVAL;
		if (copy_from_user(&vif,optval,sizeof(vif)))
			return -EFAULT;
		if (vif.vifc_vifi >= MAXVIFS)
			return -ENFILE;
		rtnl_lock();
		if (optname==MRT_ADD_VIF) {
			ret = vif_add(&vif, sk==mroute_socket);
		} else {
			ret = vif_delete(vif.vifc_vifi);
		}
		rtnl_unlock();
		return ret;

		/*
		 *	Manipulate the forwarding caches. These live
		 *	in a sort of kernel/user symbiosis.
		 */
	case MRT_ADD_MFC:
	case MRT_DEL_MFC:
		if (optlen!=sizeof(mfc))
			return -EINVAL;
		if (copy_from_user(&mfc,optval, sizeof(mfc)))
			return -EFAULT;
		rtnl_lock();
		if (optname==MRT_DEL_MFC)
			ret = ipmr_mfc_delete(&mfc);
		else
			ret = ipmr_mfc_add(&mfc, sk==mroute_socket);
		rtnl_unlock();
		return ret;
		/*
		 *	Control PIM assert.
		 */
	case MRT_ASSERT:
	{
		int v;
		if (get_user(v,(int __user *)optval))
			return -EFAULT;
		mroute_do_assert=(v)?1:0;
		return 0;
	}
#ifdef CONFIG_IP_PIMSM
	case MRT_PIM:
	{
		int v;

		if (get_user(v,(int __user *)optval))
			return -EFAULT;
		v = (v) ? 1 : 0;

		rtnl_lock();
		ret = 0;
		if (v != mroute_do_pim) {
			mroute_do_pim = v;
			mroute_do_assert = v;
#ifdef CONFIG_IP_PIMSM_V2
			if (mroute_do_pim)
				ret = inet_add_protocol(&pim_protocol,
							IPPROTO_PIM);
			else
				ret = inet_del_protocol(&pim_protocol,
							IPPROTO_PIM);
			if (ret < 0)
				ret = -EAGAIN;
#endif
		}
		rtnl_unlock();
		return ret;
	}
#endif
	/*
	 *	Spurious command, or MRT_VERSION which you cannot
	 *	set.
	 */
	default:
		return -ENOPROTOOPT;
	}
}

/*
 *	Getsock opt support for the multicast routing system.
 */

int ip_mroute_getsockopt(struct sock *sk,int optname,char __user *optval,int __user *optlen)
{
	int olr;
	int val;