ip_output.c

GNU Hurd 源代码

		skb = sock_alloc_send_skb(sk, fraglen+hh_len+15, 0, flags&MSG_DONTWAIT, &err);
		if (skb == NULL)
			goto error;

		/*
		 *	Fill in the control structures
		 */
		 
		skb->priority = sk->priority;
		skb->dst = dst_clone(&rt->u.dst);
		skb_reserve(skb, hh_len);

		/*
		 *	Find where to start putting bytes.
		 */
		 
		data = skb_put(skb, fraglen);
		skb->nh.iph = (struct iphdr *)data;

		/*
		 *	Only write IP header onto non-raw packets 
		 */
		 
		{
			struct iphdr *iph = (struct iphdr *)data;

			iph->version = 4;
			iph->ihl = 5;
			if (opt) {
				iph->ihl += opt->optlen>>2;
				ip_options_build(skb, opt,
						 ipc->addr, rt, offset);
			}
			iph->tos = sk->ip_tos;
			iph->tot_len = htons(fraglen - fragheaderlen + iph->ihl*4);
			iph->id = id;
			iph->frag_off = htons(offset>>3);
			iph->frag_off |= mf|df;
			if (rt->rt_type == RTN_MULTICAST)
				iph->ttl = sk->ip_mc_ttl;
			else
				iph->ttl = sk->ip_ttl;
			iph->protocol = sk->protocol;
			iph->check = 0;
			iph->saddr = rt->rt_src;
			iph->daddr = rt->rt_dst;
			iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
			data += iph->ihl*4;
			
			/*
			 *	Any further fragments will have MF set.
			 */
			 
			mf = htons(IP_MF);
		}
		
		/*
		 *	User data callback
		 */

		if (getfrag(frag, data, offset, fraglen-fragheaderlen)) {
			err = -EFAULT;
			kfree_skb(skb);
			goto error;
		}

		offset -= (maxfraglen-fragheaderlen);
		fraglen = maxfraglen;

		nfrags++;

#ifdef CONFIG_FIREWALL
		switch (call_out_firewall(PF_INET, rt->u.dst.dev, skb->nh.iph, NULL, &skb)) {
		case FW_QUEUE:
			kfree_skb(skb);
			continue;
		case FW_BLOCK:
		case FW_REJECT:
			kfree_skb(skb);
			err = -EPERM;
			goto error;
		}
#endif

		err = -ENETDOWN;
		if (rt->u.dst.output(skb))
			goto error;
	} while (offset >= 0);

	if (nfrags>1)
		ip_statistics.IpFragCreates += nfrags;
	dev_unlock_list();
	return 0;

error:
	ip_statistics.IpOutDiscards++;
	if (nfrags>1)
		ip_statistics.IpFragCreates += nfrags;
	dev_unlock_list();
	return err; 
}


/*
 *	Fast path for unfragmented packets.
 */
int ip_build_xmit(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)
{
	int err;
	struct sk_buff *skb;
	int df;
	struct iphdr *iph;

	/*
	 *	Try the simple case first. This leaves fragmented frames, and by
	 *	choice RAW frames within 20 bytes of maximum size(rare) to the long path
	 */

	if (!sk->ip_hdrincl) {
		length += sizeof(struct iphdr);

		/*
		 * 	Check for slow path.
		 */
		if (length > rt->u.dst.pmtu || ipc->opt != NULL)  
			return ip_build_xmit_slow(sk,getfrag,frag,length,ipc,rt,flags); 
	} else {
		if (length > rt->u.dst.dev->mtu) {
			ip_local_error(sk, EMSGSIZE, rt->rt_dst, sk->dport, rt->u.dst.dev->mtu);
			return -EMSGSIZE;
		}
	}

	/*
	 *	Do path mtu discovery if needed.
	 */
	df = 0;
	if (ip_dont_fragment(sk, &rt->u.dst))
		df = htons(IP_DF);

	/* 
	 *	Fast path for unfragmented frames without options. 
	 */ 
	{
	int hh_len = (rt->u.dst.dev->hard_header_len + 15)&~15;

	skb = sock_alloc_send_skb(sk, length+hh_len+15,
				  0, flags&MSG_DONTWAIT, &err);
	if(skb==NULL)
		goto error; 
	skb_reserve(skb, hh_len);
	}
	
	skb->priority = sk->priority;
	skb->dst = dst_clone(&rt->u.dst);

	skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
	
	dev_lock_list();
	
	if(!sk->ip_hdrincl) {
		iph->version=4;
		iph->ihl=5;
		iph->tos=sk->ip_tos;
		iph->tot_len = htons(length);
		iph->id=htons(ip_id_count++);
		iph->frag_off = df;
		iph->ttl=sk->ip_mc_ttl;
		if (rt->rt_type != RTN_MULTICAST)
			iph->ttl=sk->ip_ttl;
		iph->protocol=sk->protocol;
		iph->saddr=rt->rt_src;
		iph->daddr=rt->rt_dst;
		iph->check=0;
		iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
		err = getfrag(frag, ((char *)iph)+iph->ihl*4,0, length-iph->ihl*4);
	}
	else
		err = getfrag(frag, (void *)iph, 0, length);

	dev_unlock_list();

	if (err)
		goto error_fault;

#ifdef CONFIG_FIREWALL
	switch (call_out_firewall(PF_INET, rt->u.dst.dev, iph, NULL, &skb)) {
	case FW_QUEUE:
		kfree_skb(skb);
		return 0;
	case FW_BLOCK:
	case FW_REJECT:
		kfree_skb(skb);
		err = -EPERM;
		goto error;
	}
#endif

	return rt->u.dst.output(skb);

error_fault:
	err = -EFAULT;
	kfree_skb(skb);
error:
	ip_statistics.IpOutDiscards++;
	return err; 
}
		       


/*
 *	This IP datagram is too large to be sent in one piece.  Break it up into
 *	smaller pieces (each of size equal to IP header plus
 *	a block of the data of the original IP data part) that will yet fit in a
 *	single device frame, and queue such a frame for sending.
 *
 *	Yes this is inefficient, feel free to submit a quicker one.
 */

void ip_fragment(struct sk_buff *skb, int (*output)(struct sk_buff*))
{
	struct iphdr *iph;
	unsigned char *raw;
	unsigned char *ptr;
	struct device *dev;
	struct sk_buff *skb2;
	unsigned int mtu, hlen, left, len; 
	int offset;
	int not_last_frag;
	struct rtable *rt = (struct rtable*)skb->dst;

	dev = rt->u.dst.dev;

	/*
	 *	Point into the IP datagram header.
	 */

	raw = skb->nh.raw;
	iph = (struct iphdr*)raw;

	/*
	 *	Setup starting values.
	 */

	hlen = iph->ihl * 4;
	left = ntohs(iph->tot_len) - hlen;	/* Space per frame */
	mtu = rt->u.dst.pmtu - hlen;	/* Size of data space */
	ptr = raw + hlen;			/* Where to start from */

	/*
	 *	The protocol doesn't seem to say what to do in the case that the
	 *	frame + options doesn't fit the mtu. As it used to fall down dead
	 *	in this case we were fortunate it didn't happen
	 *
	 *	It is impossible, because mtu>=68. --ANK (980801)
	 */

#ifdef CONFIG_NET_PARANOIA
	if (mtu<8) 
		goto fail;
#endif

	/*
	 *	Fragment the datagram.
	 */

	offset = (ntohs(iph->frag_off) & IP_OFFSET) << 3;
	not_last_frag = iph->frag_off & htons(IP_MF);

	/*
	 *	Keep copying data until we run out.
	 */

	while(left > 0)	{
		len = left;
		/* IF: it doesn't fit, use 'mtu' - the data space left */
		if (len > mtu)
			len = mtu;
		/* IF: we are not sending upto and including the packet end
		   then align the next start on an eight byte boundary */
		if (len < left)	{
			len &= ~7;
		}
		/*
		 *	Allocate buffer.
		 */

		if ((skb2 = alloc_skb(len+hlen+dev->hard_header_len+15,GFP_ATOMIC)) == NULL) {
			NETDEBUG(printk(KERN_INFO "IP: frag: no memory for new fragment!\n"));
			goto fail;
		}

		/*
		 *	Set up data on packet
		 */

		skb2->pkt_type = skb->pkt_type;
		skb2->priority = skb->priority;
		skb_reserve(skb2, (dev->hard_header_len+15)&~15);
		skb_put(skb2, len + hlen);
		skb2->nh.raw = skb2->data;
		skb2->h.raw = skb2->data + hlen;

		/*
		 *	Charge the memory for the fragment to any owner
		 *	it might possess
		 */

		if (skb->sk)
			skb_set_owner_w(skb2, skb->sk);
		skb2->dst = dst_clone(skb->dst);

		/*
		 *	Copy the packet header into the new buffer.
		 */

		memcpy(skb2->nh.raw, raw, hlen);

		/*
		 *	Copy a block of the IP datagram.
		 */
		memcpy(skb2->h.raw, ptr, len);
		left -= len;

		/*
		 *	Fill in the new header fields.
		 */
		iph = skb2->nh.iph;
		iph->frag_off = htons((offset >> 3));

		/* ANK: dirty, but effective trick. Upgrade options only if
		 * the segment to be fragmented was THE FIRST (otherwise,
		 * options are already fixed) and make it ONCE
		 * on the initial skb, so that all the following fragments
		 * will inherit fixed options.
		 */
		if (offset == 0)
			ip_options_fragment(skb);

		/*
		 *	Added AC : If we are fragmenting a fragment that's not the
		 *		   last fragment then keep MF on each bit
		 */
		if (left > 0 || not_last_frag)
			iph->frag_off |= htons(IP_MF);
		ptr += len;
		offset += len;

		/*
		 *	Put this fragment into the sending queue.
		 */

		ip_statistics.IpFragCreates++;

		iph->tot_len = htons(len + hlen);

		ip_send_check(iph);

		output(skb2);
	}
	kfree_skb(skb);
	ip_statistics.IpFragOKs++;
	return;
	
fail:
	kfree_skb(skb); 
	ip_statistics.IpFragFails++; 
}

/*
 *	Fetch data from kernel space and fill in checksum if needed.
 */
static int ip_reply_glue_bits(const void *dptr, char *to, unsigned int offset, 
			      unsigned int fraglen)
{
        struct ip_reply_arg *dp = (struct ip_reply_arg*)dptr;
	u16 *pktp = (u16 *)to;
	struct iovec *iov; 
	int len; 
	int hdrflag = 1; 

	iov = &dp->iov[0]; 
	if (offset >= iov->iov_len) { 
		offset -= iov->iov_len;
		iov++; 
		hdrflag = 0; 
	}
	len = iov->iov_len - offset;
	if (fraglen > len) { /* overlapping. */ 
		dp->csum = csum_partial_copy_nocheck(iov->iov_base+offset, to, len,
					     dp->csum);
		offset = 0;
		fraglen -= len; 
		to += len; 
		iov++;
	}

	dp->csum = csum_partial_copy_nocheck(iov->iov_base+offset, to, fraglen, 
					     dp->csum); 

	if (hdrflag && dp->csumoffset)
		*(pktp + dp->csumoffset) = csum_fold(dp->csum); /* fill in checksum */
	return 0;	       
}

/* 
 *	Generic function to send a packet as reply to another packet.
 *	Used to send TCP resets so far. ICMP should use this function too.
 *
 *	Should run single threaded per socket because it uses the sock 
 *     	structure to pass arguments.
 */
void ip_send_reply(struct sock *sk, struct sk_buff *skb, struct ip_reply_arg *arg,
		   unsigned int len)
{
	struct {
		struct ip_options	opt;
		char			data[40];
	} replyopts;
	struct ipcm_cookie ipc;
	u32 daddr;
	struct rtable *rt = (struct rtable*)skb->dst;
	
	if (ip_options_echo(&replyopts.opt, skb))
		return;
	
	sk->ip_tos = skb->nh.iph->tos;
	sk->priority = skb->priority;
	sk->protocol = skb->nh.iph->protocol;

	daddr = ipc.addr = rt->rt_src;
	ipc.opt = &replyopts.opt;
	
	if (ipc.opt->srr)
		daddr = replyopts.opt.faddr;
	if (ip_route_output(&rt, daddr, rt->rt_spec_dst, RT_TOS(skb->nh.iph->tos), 0))
		return;

	/* And let IP do all the hard work. */
	ip_build_xmit(sk, ip_reply_glue_bits, arg, len, &ipc, rt, MSG_DONTWAIT);
	ip_rt_put(rt);
}

/*
 *	IP protocol layer initialiser
 */

static struct packet_type ip_packet_type =
{
	__constant_htons(ETH_P_IP),
	NULL,	/* All devices */
	ip_rcv,
	NULL,
	NULL,
};



#ifdef CONFIG_PROC_FS
#ifdef CONFIG_IP_MULTICAST
static struct proc_dir_entry proc_net_igmp = {
	PROC_NET_IGMP, 4, "igmp",
	S_IFREG | S_IRUGO, 1, 0, 0,
	0, &proc_net_inode_operations,
	ip_mc_procinfo
};
#endif
#endif	

/*
 *	IP registers the packet type and then calls the subprotocol initialisers
 */

__initfunc(void ip_init(void))
{
	dev_add_pack(&ip_packet_type);

	ip_rt_init();

#ifdef CONFIG_PROC_FS
#ifdef CONFIG_IP_MULTICAST
	proc_net_register(&proc_net_igmp);
#endif
#endif	
}