dn_route.c

Linux内核源代码 为压缩文件 是<<Linux内核>>一书中的源代码

{
	struct dn_route *rt = NULL;
	struct net_device *dev = decnet_default_device;
	struct neighbour *neigh = NULL;
	struct dn_dev *dn_db;
	unsigned hash;
#ifdef CONFIG_DECNET_ROUTER
	struct dn_fib_key key;
	struct dn_fib_res res;
	int err;

	key.src = src;
	key.dst = dst;
	key.iif = 0;
	key.oif = 0;
	key.fwmark = 0;
	key.scope = RT_SCOPE_UNIVERSE;

	if ((err = dn_fib_lookup(&key, &res)) == 0) {
		switch(res.type) {
			case RTN_UNICAST:
				/*
				 * This method of handling multipath
				 * routes is a hack and will change.
				 * It works for now though.
				 */
				if (res.fi->fib_nhs)
					dn_fib_select_multipath(&key, &res);
				neigh = __neigh_lookup(&dn_neigh_table, &DN_FIB_RES_GW(res), DN_FIB_RES_DEV(res), 1);
				err = -ENOBUFS;
				if (!neigh)
					break;
				err = 0;
				break;
			case RTN_UNREACHABLE:
				err = -EHOSTUNREACH;
				break;
			default:
				err = -EINVAL;
		}
		dn_fib_res_put(&res);
		if (err < 0)
			return err;
		goto got_route;
	}

	if (err != -ESRCH)
		return err;
#endif 

	/* Look in On-Ethernet cache first */
	if (!(flags & MSG_TRYHARD)) {
		if ((neigh = dn_neigh_lookup(&dn_neigh_table, &dst)) != NULL)
			goto got_route;
	}

	if (dev == NULL)
		return -EINVAL;

	dn_db = dev->dn_ptr;

	if (dn_db == NULL)
		return -EINVAL;

	/* Try default router */
	if ((neigh = neigh_clone(dn_db->router)) != NULL)
		goto got_route;

	/* Send to default device (and hope for the best) if above fail */
	if ((neigh = __neigh_lookup(&dn_neigh_table, &dst, dev, 1)) != NULL)
		goto got_route;


	return -EINVAL;

got_route:

	if ((rt = dst_alloc(&dn_dst_ops)) == NULL) {
		neigh_release(neigh);
		return -EINVAL;
	}

	dn_db = (struct dn_dev *)neigh->dev->dn_ptr;
	
	rt->key.saddr  = src;
	rt->rt_saddr   = src;
	rt->key.daddr  = dst;
	rt->rt_daddr   = dst;
	rt->key.oif    = neigh ? neigh->dev->ifindex : -1;
	rt->key.iif    = 0;
	rt->key.fwmark = 0;

	rt->u.dst.neighbour = neigh;
	rt->u.dst.dev = neigh ? neigh->dev : NULL;
	rt->u.dst.lastuse = jiffies;
	rt->u.dst.output = dn_output;
	rt->u.dst.input  = dn_rt_bug;

	if (dn_dev_islocal(neigh->dev, rt->rt_daddr))
		rt->u.dst.input = dn_nsp_rx;

	hash = dn_hash(rt->key.saddr, rt->key.daddr);
	dn_insert_route(rt, hash);
	*pprt = &rt->u.dst;

	return 0;
}

int dn_route_output(struct dst_entry **pprt, dn_address dst, dn_address src, int flags)
{
	unsigned hash = dn_hash(src, dst);
	struct dn_route *rt = NULL;

	if (!(flags & MSG_TRYHARD)) {
		read_lock_bh(&dn_rt_hash_table[hash].lock);
		for(rt = dn_rt_hash_table[hash].chain; rt; rt = rt->u.rt_next) {
			if ((dst == rt->key.daddr) &&
					(src == rt->key.saddr) &&
					(rt->key.iif == 0) &&
					(rt->key.oif != 0)) {
				rt->u.dst.lastuse = jiffies;
				dst_hold(&rt->u.dst);
				rt->u.dst.__use++;
				read_unlock_bh(&dn_rt_hash_table[hash].lock);
				*pprt = &rt->u.dst;
				return 0;
			}
		}
		read_unlock_bh(&dn_rt_hash_table[hash].lock);
	}

	return dn_route_output_slow(pprt, dst, src, flags);
}

static int dn_route_input_slow(struct sk_buff *skb)
{
	struct dn_route *rt = NULL;
	struct dn_skb_cb *cb = (struct dn_skb_cb *)skb->cb;
	struct net_device *dev = skb->dev;
	struct dn_dev *dn_db;
	struct neighbour *neigh = NULL;
	int (*dnrt_input)(struct sk_buff *skb);
	int (*dnrt_output)(struct sk_buff *skb);
	u32 fwmark = 0;
	unsigned hash;
	dn_address saddr = cb->src;
	dn_address daddr = cb->dst;
#ifdef CONFIG_DECNET_ROUTER
	struct dn_fib_key key;
	struct dn_fib_res res;
	int err;
#endif

	if (dev == NULL)
		return -EINVAL;

	if ((dn_db = dev->dn_ptr) == NULL)
		return -EINVAL;

	/*
	 * In this case we've just received a packet from a source
	 * outside ourselves pretending to come from us. We don't
	 * allow it any further to prevent routing loops, spoofing and
	 * other nasties. Loopback packets already have the dst attached
	 * so this only affects packets which have originated elsewhere.
	 */
	if (dn_dev_islocal(dev, cb->src))
		return -ENOTUNIQ;

	/*
	 * Default is to create a drop everything entry
	 */
	dnrt_input  = dn_blackhole;
	dnrt_output = dn_rt_bug;

	/*
	 * Is the destination us ?
	 */
	if (!dn_dev_islocal(dev, cb->dst))
		goto non_local_input;

	/*
	 * Local input... find source of skb
	 */
	dnrt_input  = dn_nsp_rx;
	dnrt_output = dn_output;
	saddr = cb->dst;
	daddr = cb->src;

	if ((neigh = neigh_lookup(&dn_neigh_table, &cb->src, dev)) != NULL)
		goto add_entry;

	if (dn_db->router && ((neigh = neigh_clone(dn_db->router)) != NULL))
		goto add_entry;

	neigh = neigh_create(&dn_neigh_table, &cb->src, dev);
	if (!IS_ERR(neigh)) {
		if (dev->type == ARPHRD_ETHER)
			memcpy(neigh->ha, skb->mac.ethernet->h_source, ETH_ALEN);
		goto add_entry;
	}

	return PTR_ERR(neigh);

non_local_input:

#ifdef CONFIG_DECNET_ROUTER
	/*
	 * Destination is another node... find next hop in
	 * routing table here.
	 */

	key.src = cb->src;
	key.dst = cb->dst;
	key.iif = dev->ifindex;
	key.oif = 0;
	key.scope = RT_SCOPE_UNIVERSE;

#ifdef CONFIG_DECNET_ROUTE_FWMARK
	key.fwmark = skb->nfmark;
#else
	key.fwmark = 0;
#endif

	if ((err = dn_fib_lookup(&key, &res)) == 0) {
		switch(res.type) {
			case RTN_UNICAST:
				if (res.fi->fib_nhs)
					dn_fib_select_multipath(&key, &res);
				neigh = __neigh_lookup(&dn_neigh_table, &DN_FIB_RES_GW(res), DN_FIB_RES_DEV(res), 1);
				err = -ENOBUFS;
				if (!neigh)
					break;
				err = 0;
				dnrt_input = dn_forward;
				fwmark = key.fwmark;
				break;
			case RTN_UNREACHABLE:
				dnrt_input = dn_blackhole;
				fwmark = key.fwmark;
				break;
			default:
				err = -EINVAL;
		}
		dn_fib_res_put(&res);
		if (err < 0)
			return err;
		goto add_entry;
	}

	return err;

#endif /* CONFIG_DECNET_ROUTER */

add_entry:

	if ((rt = dst_alloc(&dn_dst_ops)) == NULL) {
                neigh_release(neigh);
                return -EINVAL;
        }

	rt->key.saddr  = cb->src;
	rt->rt_saddr   = saddr;
	rt->key.daddr  = cb->dst;
	rt->rt_daddr   = daddr;
	rt->key.oif    = 0;
	rt->key.iif    = dev->ifindex;
	rt->key.fwmark = fwmark;

	rt->u.dst.neighbour = neigh;
	rt->u.dst.dev = neigh ? neigh->dev : NULL;
	rt->u.dst.lastuse = jiffies;
	rt->u.dst.output = dnrt_output;
	rt->u.dst.input = dnrt_input;

	hash = dn_hash(rt->key.saddr, rt->key.daddr);
	dn_insert_route(rt, hash);
	skb->dst = (struct dst_entry *)rt;

	return 0;
}

int dn_route_input(struct sk_buff *skb)
{
	struct dn_route *rt;
	struct dn_skb_cb *cb = (struct dn_skb_cb *)skb->cb;
	unsigned hash = dn_hash(cb->src, cb->dst);

	if (skb->dst)
		return 0;

	read_lock(&dn_rt_hash_table[hash].lock);
	for(rt = dn_rt_hash_table[hash].chain; rt != NULL; rt = rt->u.rt_next) {
		if ((rt->key.saddr == cb->src) &&
				(rt->key.daddr == cb->dst) &&
				(rt->key.oif == 0) &&
#ifdef CONFIG_DECNET_ROUTE_FWMARK
				(rt->key.fwmark == skb->nfmark) &&
#endif
				(rt->key.iif == cb->iif)) {
			rt->u.dst.lastuse = jiffies;
			dst_hold(&rt->u.dst);
			rt->u.dst.__use++;
			read_unlock(&dn_rt_hash_table[hash].lock);
			skb->dst = (struct dst_entry *)rt;
			return 0;
		}
	}
	read_unlock(&dn_rt_hash_table[hash].lock);

	return dn_route_input_slow(skb);
}

#ifdef CONFIG_RTNETLINK
static int dn_rt_fill_info(struct sk_buff *skb, u32 pid, u32 seq, int event, int nowait)
{
	struct dn_route *rt = (struct dn_route *)skb->dst;
	struct rtmsg *r;
	struct nlmsghdr *nlh;
	unsigned char *b = skb->tail;

	nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(*r));
	r = NLMSG_DATA(nlh);
	nlh->nlmsg_flags = nowait ? NLM_F_MULTI : 0;
	r->rtm_family = AF_DECnet;
	r->rtm_dst_len = 16;
	r->rtm_src_len = 16;
	r->rtm_tos = 0;
	r->rtm_table = 0;
	r->rtm_type = 0;
	r->rtm_flags = 0;
	r->rtm_scope = RT_SCOPE_UNIVERSE;
	r->rtm_protocol = RTPROT_UNSPEC;
	RTA_PUT(skb, RTA_DST, 2, &rt->rt_daddr);
	RTA_PUT(skb, RTA_SRC, 2, &rt->rt_saddr);
	if (rt->u.dst.dev)
		RTA_PUT(skb, RTA_OIF, sizeof(int), &rt->u.dst.dev->ifindex);
	if (rt->u.dst.window)
		RTA_PUT(skb, RTAX_WINDOW, sizeof(unsigned), &rt->u.dst.window);
	if (rt->u.dst.rtt)
		RTA_PUT(skb, RTAX_RTT, sizeof(unsigned), &rt->u.dst.rtt);

	nlh->nlmsg_len = skb->tail - b;
	return skb->len;

nlmsg_failure:
rtattr_failure:
        skb_trim(skb, b - skb->data);
        return -1;
}

/*
 * This is called by both endnodes and routers now.
 */
int dn_cache_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, void *arg)
{
	struct rtattr **rta = arg;
	struct dn_route *rt = NULL;
	struct dn_skb_cb *cb;
	dn_address dst = 0;
	dn_address src = 0;
	int iif = 0;
	int err;
	struct sk_buff *skb;

	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
	if (skb == NULL)
		return -ENOBUFS;
	skb->mac.raw = skb->data;
	cb = (struct dn_skb_cb *)skb->cb;

	if (rta[RTA_SRC-1])
		memcpy(&src, RTA_DATA(rta[RTA_SRC-1]), 2);
	if (rta[RTA_DST-1])
		memcpy(&dst, RTA_DATA(rta[RTA_DST-1]), 2);
	if (rta[RTA_IIF-1])
		memcpy(&iif, RTA_DATA(rta[RTA_IIF-1]), sizeof(int));

	if (iif) {
		struct net_device *dev;
		if ((dev = dev_get_by_index(iif)) == NULL) {
			kfree_skb(skb);
			return -ENODEV;
		}
		if (!dev->dn_ptr) {
			dev_put(dev);
			kfree_skb(skb);
			return -ENODEV;
		}
		skb->protocol = __constant_htons(ETH_P_DNA_RT);
		skb->dev = dev;
		cb->src = src;
		cb->dst = dst;
		local_bh_disable();
		err = dn_route_input(skb);
		local_bh_enable();
		memset(cb, 0, sizeof(struct dn_skb_cb));
		rt = (struct dn_route *)skb->dst;
	} else {
		err = dn_route_output((struct dst_entry **)&rt, dst, src, 0);
	}

	if (!err && rt->u.dst.error)
		err = rt->u.dst.error;
	if (skb->dev)
		dev_put(skb->dev);
	skb->dev = NULL;
	if (err)
		goto out_free;
	skb->dst = &rt->u.dst;

	NETLINK_CB(skb).dst_pid = NETLINK_CB(in_skb).pid;

	err = dn_rt_fill_info(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, RTM_NEWROUTE, 0);

	if (err == 0)
		goto out_free;
	if (err < 0) {
		err = -EMSGSIZE;
		goto out_free;
	}

	err = netlink_unicast(rtnl, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);

	return err;

out_free:
	kfree_skb(skb);
	return err;
}

/*
 * For routers, this is called from dn_fib_dump, but for endnodes its
 * called directly from the rtnetlink dispatch table.
 */
int dn_cache_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
	struct dn_route *rt;
	int h, s_h;
	int idx, s_idx;

	if (NLMSG_PAYLOAD(cb->nlh, 0) < sizeof(struct rtmsg))
		return -EINVAL;
	if (!(((struct rtmsg *)NLMSG_DATA(cb->nlh))->rtm_flags&RTM_F_CLONED))
		return 0;

	s_h = cb->args[0];
	s_idx = idx = cb->args[1];
	for(h = 0; h <= dn_rt_hash_mask; h++) {
		if (h < s_h)
			continue;
		if (h > s_h)
			s_idx = 0;
		read_lock_bh(&dn_rt_hash_table[h].lock);
		for(rt = dn_rt_hash_table[h].chain, idx = 0; rt; rt = rt->u.rt_next, idx++) {
			if (idx < s_idx)
				continue;
			skb->dst = dst_clone(&rt->u.dst);
			if (dn_rt_fill_info(skb, NETLINK_CB(cb->skb).pid,
					cb->nlh->nlmsg_seq, RTM_NEWROUTE, 1) <= 0) {
				dst_release(xchg(&skb->dst, NULL));
				read_unlock_bh(&dn_rt_hash_table[h].lock);
				goto done;
			}
			dst_release(xchg(&skb->dst, NULL));
		}
		read_unlock_bh(&dn_rt_hash_table[h].lock);
	}

done:
	cb->args[0] = h;
	cb->args[1] = idx;
	return skb->len;
}
#endif /* CONFIG_RTNETLINK */

#ifdef CONFIG_PROC_FS

static int decnet_cache_get_info(char *buffer, char **start, off_t offset, int length)
{
        int len = 0;
        off_t pos = 0;
        off_t begin = 0;
	struct dn_route *rt;
	int i;
	char buf1[DN_ASCBUF_LEN], buf2[DN_ASCBUF_LEN];

	for(i = 0; i <= dn_rt_hash_mask; i++) {
		read_lock_bh(&dn_rt_hash_table[i].lock);
		rt = dn_rt_hash_table[i].chain;
		for(; rt != NULL; rt = rt->u.rt_next) {
			len += sprintf(buffer + len, "%-8s %-7s %-7s %04d %04d %04d\n",
					rt->u.dst.dev ? rt->u.dst.dev->name : "*",
					dn_addr2asc(dn_ntohs(rt->rt_daddr), buf1),
					dn_addr2asc(dn_ntohs(rt->rt_saddr), buf2),
					atomic_read(&rt->u.dst.__refcnt),
					rt->u.dst.__use,
					(int)rt->u.dst.rtt
					);



	                pos = begin + len;
	
        	        if (pos < offset) {
                	        len   = 0;
                        	begin = pos;
                	}
              		if (pos > offset + length)
                	        break;
		}
		read_unlock_bh(&dn_rt_hash_table[i].lock);
		if (pos > offset + length)
			break;
	}

        *start = buffer + (offset - begin);
        len   -= (offset - begin);

        if (len > length) len = length;

        return(len);
} 

#endif /* CONFIG_PROC_FS */

void __init dn_route_init(void)
{
	int i, goal, order;

	dn_dst_ops.kmem_cachep = kmem_cache_create("dn_dst_cache",
						   sizeof(struct dn_route),
						   0, SLAB_HWCACHE_ALIGN,
						   NULL, NULL);

	if (!dn_dst_ops.kmem_cachep)
		panic("DECnet: Failed to allocate dn_dst_cache\n");

	dn_route_timer.function = dn_dst_check_expire;
	dn_route_timer.expires = jiffies + decnet_dst_gc_interval * HZ;
	add_timer(&dn_route_timer);

	goal = num_physpages >> (26 - PAGE_SHIFT);

	for(order = 0; (1UL << order) < goal; order++)
		/* NOTHING */;

        /*
         * Only want 1024 entries max, since the table is very, very unlikely
         * to be larger than that.
         */
        while(order && ((((1UL << order) * PAGE_SIZE) / 
                                sizeof(struct dn_rt_hash_bucket)) >= 2048))
                order--;

        do {
                dn_rt_hash_mask = (1UL << order) * PAGE_SIZE /
                        sizeof(struct dn_rt_hash_bucket);
                while(dn_rt_hash_mask & (dn_rt_hash_mask - 1))
                        dn_rt_hash_mask--;
                dn_rt_hash_table = (struct dn_rt_hash_bucket *)
                        __get_free_pages(GFP_ATOMIC, order);
        } while (dn_rt_hash_table == NULL && --order > 0);

	if (!dn_rt_hash_table)
                panic("Failed to allocate DECnet route cache hash table\n");

	printk(KERN_INFO 
		"DECnet: Routing cache hash table of %u buckets, %ldKbytes\n", 
		dn_rt_hash_mask, 
		(long)(dn_rt_hash_mask*sizeof(struct dn_rt_hash_bucket))/1024);

	dn_rt_hash_mask--;
        for(i = 0; i <= dn_rt_hash_mask; i++) {
                dn_rt_hash_table[i].lock = RW_LOCK_UNLOCKED;
                dn_rt_hash_table[i].chain = NULL;
        }

        dn_dst_ops.gc_thresh = (dn_rt_hash_mask + 1);

#ifdef CONFIG_PROC_FS
	proc_net_create("decnet_cache",0,decnet_cache_get_info);
#endif /* CONFIG_PROC_FS */
}

void __exit dn_route_cleanup(void)
{
	del_timer(&dn_route_timer);
	dn_run_flush(0);

#ifdef CONFIG_PROC_FS
	proc_net_remove("decnet_cache");
#endif /* CONFIG_PROC_FS */
}