ip_output.c

ARM 嵌入式 系统 设计与实例开发 实验教材 二源码

/*
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the  BSD Socket
 *		interface as the means of communication with the user level.
 *
 *		The Internet Protocol (IP) output module.
 *
 * Version:	$Id: ip_output.c,v 1.99 2001/10/15 12:34:50 davem Exp $
 *
 * Authors:	Ross Biro, <[e-mail]>
 *		Fred N. van Kempen, <[e-mail]>
 *		Donald Becker, <[e-mail]>
 *		Alan Cox, <[e-mail]>
 *		Richard Underwood
 *		Stefan Becker, <[e-mail]>
 *		Jorge Cwik, <[e-mail]>
 *		Arnt Gulbrandsen, <[e-mail]>
 *
 *	See ip_input.c for original log
 *
 *	Fixes:
 *		Alan Cox	:	Missing nonblock feature in ip_build_xmit.
 *		Mike Kilburn	:	htons() missing in ip_build_xmit.
 *		Bradford Johnson:	Fix faulty handling of some frames when 
 *					no route is found.
 *		Alexander Demenshin:	Missing sk/skb free in ip_queue_xmit
 *					(in case if packet not accepted by
 *					output firewall rules)
 *		Mike McLagan	:	Routing by source
 *		Alexey Kuznetsov:	use new route cache
 *		Andi Kleen:		Fix broken PMTU recovery and remove
 *					some redundant tests.
 *	Vitaly E. Lavrov	:	Transparent proxy revived after year coma.
 *		Andi Kleen	: 	Replace ip_reply with ip_send_reply.
 *		Andi Kleen	:	Split fast and slow ip_build_xmit path 
 *					for decreased register pressure on x86 
 *					and more readibility. 
 *		Marc Boucher	:	When call_out_firewall returns FW_QUEUE,
 *					silently drop skb instead of failing with -EPERM.
 *		Detlev Wengorz	:	Copy protocol for fragments.
 */

#include <asm/uaccess.h>
#include <asm/system.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/mm.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/config.h>

#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/in.h>
#include <linux/inet.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/proc_fs.h>
#include <linux/stat.h>
#include <linux/init.h>

#include <net/snmp.h>
#include <net/ip.h>
#include <net/protocol.h>
#include <net/route.h>
#include <net/tcp.h>
#include <net/udp.h>
#include <linux/skbuff.h>
#include <net/sock.h>
#include <net/arp.h>
#include <net/icmp.h>
#include <net/raw.h>
#include <net/checksum.h>
#include <net/inetpeer.h>
#include <linux/igmp.h>
#include <linux/netfilter_ipv4.h>
#include <linux/mroute.h>
#include <linux/netlink.h>

/*
 *      Shall we try to damage output packets if routing dev changes?
 */

int sysctl_ip_dynaddr = 0;
int sysctl_ip_default_ttl = IPDEFTTL;

/* Generate a checksum for an outgoing IP datagram. */
__inline__ void ip_send_check(struct iphdr *iph)
{
	iph->check = 0;
	iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
}

/* dev_loopback_xmit for use with netfilter. */
static int ip_dev_loopback_xmit(struct sk_buff *newskb)
{
	newskb->mac.raw = newskb->data;
	__skb_pull(newskb, newskb->nh.raw - newskb->data);
	newskb->pkt_type = PACKET_LOOPBACK;
	newskb->ip_summed = CHECKSUM_UNNECESSARY;
	BUG_TRAP(newskb->dst);

#ifdef CONFIG_NETFILTER_DEBUG
	nf_debug_ip_loopback_xmit(newskb);
#endif
	netif_rx(newskb);
	return 0;
}

/* Don't just hand NF_HOOK skb->dst->output, in case netfilter hook
   changes route */
static inline int
output_maybe_reroute(struct sk_buff *skb)
{
	return skb->dst->output(skb);
}

/* 
 *		Add an ip header to a skbuff and send it out.
 */
int ip_build_and_send_pkt(struct sk_buff *skb, struct sock *sk,
			  u32 saddr, u32 daddr, struct ip_options *opt)
{
	struct rtable *rt = (struct rtable *)skb->dst;
	struct iphdr *iph;

	/* Build the IP header. */
	if (opt)
		iph=(struct iphdr *)skb_push(skb,sizeof(struct iphdr) + opt->optlen);
	else
		iph=(struct iphdr *)skb_push(skb,sizeof(struct iphdr));

	iph->version  = 4;
	iph->ihl      = 5;
	iph->tos      = sk->protinfo.af_inet.tos;
	iph->frag_off = 0;
	if (ip_dont_fragment(sk, &rt->u.dst))
		iph->frag_off |= htons(IP_DF);
	iph->ttl      = sk->protinfo.af_inet.ttl;
	iph->daddr    = rt->rt_dst;
	iph->saddr    = rt->rt_src;
	iph->protocol = sk->protocol;
	iph->tot_len  = htons(skb->len);
	ip_select_ident(iph, &rt->u.dst, sk);
	skb->nh.iph   = iph;

	if (opt && opt->optlen) {
		iph->ihl += opt->optlen>>2;
		ip_options_build(skb, opt, daddr, rt, 0);
	}
	ip_send_check(iph);

	/* Send it out. */
	return NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
		       output_maybe_reroute);
}

static inline int ip_finish_output2(struct sk_buff *skb)
{
	struct dst_entry *dst = skb->dst;
	struct hh_cache *hh = dst->hh;

#ifdef CONFIG_NETFILTER_DEBUG
	nf_debug_ip_finish_output2(skb);
#endif /*CONFIG_NETFILTER_DEBUG*/

	if (hh) {
		read_lock_bh(&hh->hh_lock);
  		memcpy(skb->data - 16, hh->hh_data, 16);
		read_unlock_bh(&hh->hh_lock);
	        skb_push(skb, hh->hh_len);
		return hh->hh_output(skb);
	} else if (dst->neighbour)
		return dst->neighbour->output(skb);

	if (net_ratelimit())
		printk(KERN_DEBUG "ip_finish_output2: No header cache and no neighbour!\n");
	kfree_skb(skb);
	return -EINVAL;
}

__inline__ int ip_finish_output(struct sk_buff *skb)
{
	struct net_device *dev = skb->dst->dev;

	skb->dev = dev;
	skb->protocol = __constant_htons(ETH_P_IP);

	return NF_HOOK(PF_INET, NF_IP_POST_ROUTING, skb, NULL, dev,
		       ip_finish_output2);
}

int ip_mc_output(struct sk_buff *skb)
{
	struct sock *sk = skb->sk;
	struct rtable *rt = (struct rtable*)skb->dst;
	struct net_device *dev = rt->u.dst.dev;

	/*
	 *	If the indicated interface is up and running, send the packet.
	 */
	IP_INC_STATS(IpOutRequests);
#ifdef CONFIG_IP_ROUTE_NAT
	if (rt->rt_flags & RTCF_NAT)
		ip_do_nat(skb);
#endif

	skb->dev = dev;
	skb->protocol = __constant_htons(ETH_P_IP);

	/*
	 *	Multicasts are looped back for other local users
	 */

	if (rt->rt_flags&RTCF_MULTICAST) {
		if ((!sk || sk->protinfo.af_inet.mc_loop)
#ifdef CONFIG_IP_MROUTE
		/* Small optimization: do not loopback not local frames,
		   which returned after forwarding; they will be  dropped
		   by ip_mr_input in any case.
		   Note, that local frames are looped back to be delivered
		   to local recipients.

		   This check is duplicated in ip_mr_input at the moment.
		 */
		    && ((rt->rt_flags&RTCF_LOCAL) || !(IPCB(skb)->flags&IPSKB_FORWARDED))
#endif
		) {
			struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
			if (newskb)
				NF_HOOK(PF_INET, NF_IP_POST_ROUTING, newskb, NULL,
					newskb->dev, 
					ip_dev_loopback_xmit);
		}

		/* Multicasts with ttl 0 must not go beyond the host */

		if (skb->nh.iph->ttl == 0) {
			kfree_skb(skb);
			return 0;
		}
	}

	if (rt->rt_flags&RTCF_BROADCAST) {
		struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
		if (newskb)
			NF_HOOK(PF_INET, NF_IP_POST_ROUTING, newskb, NULL,
				newskb->dev, ip_dev_loopback_xmit);
	}

	return ip_finish_output(skb);
}

int ip_output(struct sk_buff *skb)
{
#ifdef CONFIG_IP_ROUTE_NAT
	struct rtable *rt = (struct rtable*)skb->dst;
#endif

	IP_INC_STATS(IpOutRequests);

#ifdef CONFIG_IP_ROUTE_NAT
	if (rt->rt_flags&RTCF_NAT)
		ip_do_nat(skb);
#endif

	return ip_finish_output(skb);
}

/* Queues a packet to be sent, and starts the transmitter if necessary.  
 * This routine also needs to put in the total length and compute the 
 * checksum.  We use to do this in two stages, ip_build_header() then
 * this, but that scheme created a mess when routes disappeared etc.
 * So we do it all here, and the TCP send engine has been changed to
 * match. (No more unroutable FIN disasters, etc. wheee...)  This will
 * most likely make other reliable transport layers above IP easier
 * to implement under Linux.
 */
static inline int ip_queue_xmit2(struct sk_buff *skb)
{
	struct sock *sk = skb->sk;
	struct rtable *rt = (struct rtable *)skb->dst;
	struct net_device *dev;
	struct iphdr *iph = skb->nh.iph;

	dev = rt->u.dst.dev;

	/* This can happen when the transport layer has segments queued
	 * with a cached route, and by the time we get here things are
	 * re-routed to a device with a different MTU than the original
	 * device.  Sick, but we must cover it.
	 */
	if (skb_headroom(skb) < dev->hard_header_len && dev->hard_header) {
		struct sk_buff *skb2;

		skb2 = skb_realloc_headroom(skb, (dev->hard_header_len + 15) & ~15);
		kfree_skb(skb);
		if (skb2 == NULL)
			return -ENOMEM;
		if (sk)
			skb_set_owner_w(skb2, sk);
		skb = skb2;
		iph = skb->nh.iph;
	}

	if (skb->len > rt->u.dst.pmtu)
		goto fragment;

	if (ip_dont_fragment(sk, &rt->u.dst))
		iph->frag_off |= __constant_htons(IP_DF);

	ip_select_ident(iph, &rt->u.dst, sk);

	/* Add an IP checksum. */
	ip_send_check(iph);

	skb->priority = sk->priority;
	return skb->dst->output(skb);

fragment:
	if (ip_dont_fragment(sk, &rt->u.dst)) {
		/* Reject packet ONLY if TCP might fragment
		 * it itself, if were careful enough.
		 */
		iph->frag_off |= __constant_htons(IP_DF);
		NETDEBUG(printk(KERN_DEBUG "sending pkt_too_big to self\n"));

		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
			  htonl(rt->u.dst.pmtu));
		kfree_skb(skb);
		return -EMSGSIZE;
	}
	ip_select_ident(iph, &rt->u.dst, sk);
	if (skb->ip_summed == CHECKSUM_HW &&
	    (skb = skb_checksum_help(skb)) == NULL)
		return -ENOMEM;
	return ip_fragment(skb, skb->dst->output);
}

int ip_queue_xmit(struct sk_buff *skb)
{
	struct sock *sk = skb->sk;
	struct ip_options *opt = sk->protinfo.af_inet.opt;
	struct rtable *rt;
	struct iphdr *iph;

	/* Skip all of this if the packet is already routed,
	 * f.e. by something like SCTP.
	 */
	rt = (struct rtable *) skb->dst;
	if (rt != NULL)
		goto packet_routed;

	/* Make sure we can route this packet. */
	rt = (struct rtable *)__sk_dst_check(sk, 0);
	if (rt == NULL) {
		u32 daddr;

		/* Use correct destination address if we have options. */
		daddr = sk->daddr;
		if(opt && opt->srr)
			daddr = opt->faddr;

		/* If this fails, retransmit mechanism of transport layer will
		 * keep trying until route appears or the connection times itself
		 * out.
		 */
		if (ip_route_output(&rt, daddr, sk->saddr,
				    RT_CONN_FLAGS(sk),
				    sk->bound_dev_if))
			goto no_route;
		__sk_dst_set(sk, &rt->u.dst);
		sk->route_caps = rt->u.dst.dev->features;
	}
	skb->dst = dst_clone(&rt->u.dst);

packet_routed:
	if (opt && opt->is_strictroute && rt->rt_dst != rt->rt_gateway)
		goto no_route;

	/* OK, we know where to send it, allocate and build IP header. */
	iph = (struct iphdr *) skb_push(skb, sizeof(struct iphdr) + (opt ? opt->optlen : 0));
	*((__u16 *)iph)	= htons((4 << 12) | (5 << 8) | (sk->protinfo.af_inet.tos & 0xff));
	iph->tot_len = htons(skb->len);
	iph->frag_off = 0;
	iph->ttl      = sk->protinfo.af_inet.ttl;
	iph->protocol = sk->protocol;
	iph->saddr    = rt->rt_src;
	iph->daddr    = rt->rt_dst;
	skb->nh.iph   = iph;
	/* Transport layer set skb->h.foo itself. */

	if(opt && opt->optlen) {
		iph->ihl += opt->optlen >> 2;
		ip_options_build(skb, opt, sk->daddr, rt, 0);
	}

	return NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
		       ip_queue_xmit2);

no_route:
	IP_INC_STATS(IpOutNoRoutes);
	kfree_skb(skb);
	return -EHOSTUNREACH;
}

/*
 *	Build and send a packet, with as little as one copy
 *
 *	Doesn't care much about ip options... option length can be
 *	different for fragment at 0 and other fragments.
 *
 *	Note that the fragment at the highest offset is sent first,
 *	so the getfrag routine can fill in the TCP/UDP checksum header
 *	field in the last fragment it sends... actually it also helps
 * 	the reassemblers, they can put most packets in at the head of
 *	the fragment queue, and they know the total size in advance. This
 *	last feature will measurably improve the Linux fragment handler one
 *	day.
 *
 *	The callback has five args, an arbitrary pointer (copy of frag),
 *	the source IP address (may depend on the routing table), the 
 *	destination address (char *), the offset to copy from, and the
 *	length to be copied.
 */

static int ip_build_xmit_slow(struct sock *sk,
		  int getfrag (const void *,
			       char *,
			       unsigned int,	
			       unsigned int),
		  const void *frag,
		  unsigned length,
		  struct ipcm_cookie *ipc,
		  struct rtable *rt,
		  int flags)
{
	unsigned int fraglen, maxfraglen, fragheaderlen;
	int err;
	int offset, mf;
	int mtu;
	u16 id;

	int hh_len = (rt->u.dst.dev->hard_header_len + 15)&~15;
	int nfrags=0;
	struct ip_options *opt = ipc->opt;
	int df = 0;

	mtu = rt->u.dst.pmtu;
	if (ip_dont_fragment(sk, &rt->u.dst))
		df = htons(IP_DF);

	length -= sizeof(struct iphdr);

	if (opt) {
		fragheaderlen = sizeof(struct iphdr) + opt->optlen;
		maxfraglen = ((mtu-sizeof(struct iphdr)-opt->optlen) & ~7) + fragheaderlen;
	} else {
		fragheaderlen = sizeof(struct iphdr);

		/*
		 *	Fragheaderlen is the size of 'overhead' on each buffer. Now work
		 *	out the size of the frames to send.
		 */

		maxfraglen = ((mtu-sizeof(struct iphdr)) & ~7) + fragheaderlen;
	}

	if (length + fragheaderlen > 0xFFFF) {
		ip_local_error(sk, EMSGSIZE, rt->rt_dst, sk->dport, mtu);
		return -EMSGSIZE;
	}

	/*
	 *	Start at the end of the frame by handling the remainder.
	 */

	offset = length - (length % (maxfraglen - fragheaderlen));

	/*
	 *	Amount of memory to allocate for final fragment.
	 */

	fraglen = length - offset + fragheaderlen;

	if (length-offset==0) {
		fraglen = maxfraglen;
		offset -= maxfraglen-fragheaderlen;
	}

	/*
	 *	The last fragment will not have MF (more fragments) set.
	 */

	mf = 0;

	/*
	 *	Don't fragment packets for path mtu discovery.
	 */

	if (offset > 0 && sk->protinfo.af_inet.pmtudisc==IP_PMTUDISC_DO) { 
		ip_local_error(sk, EMSGSIZE, rt->rt_dst, sk->dport, mtu);
 		return -EMSGSIZE;
	}
	if (flags&MSG_PROBE)
		goto out;