[net-next-2.6.git] / net / ipv4 / ip_options.c

/*
 * 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 options processing module for ip.c
 *
 * Authors:	A.N.Kuznetsov
 *
 */

#include <linux/capability.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/types.h>
#include <asm/uaccess.h>
#include <linux/skbuff.h>
#include <linux/ip.h>
#include <linux/icmp.h>
#include <linux/netdevice.h>
#include <linux/rtnetlink.h>
#include <net/sock.h>
#include <net/ip.h>
#include <net/icmp.h>
#include <net/route.h>
#include <net/cipso_ipv4.h>

/*
 * Write options to IP header, record destination address to
 * source route option, address of outgoing interface
 * (we should already know it, so that this  function is allowed be
 * called only after routing decision) and timestamp,
 * if we originate this datagram.
 *
 * daddr is real destination address, next hop is recorded in IP header.
 * saddr is address of outgoing interface.
 */

void ip_options_build(struct sk_buff * skb, struct ip_options * opt,
			    __be32 daddr, struct rtable *rt, int is_frag)
{
	unsigned char *iph = skb_network_header(skb);

	memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
	memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
	opt = &(IPCB(skb)->opt);

	if (opt->srr)
		memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);

	if (!is_frag) {
		if (opt->rr_needaddr)
			ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, rt);
		if (opt->ts_needaddr)
			ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, rt);
		if (opt->ts_needtime) {
			struct timespec tv;
			__be32 midtime;
			getnstimeofday(&tv);
			midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
			memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
		}
		return;
	}
	if (opt->rr) {
		memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
		opt->rr = 0;
		opt->rr_needaddr = 0;
	}
	if (opt->ts) {
		memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
		opt->ts = 0;
		opt->ts_needaddr = opt->ts_needtime = 0;
	}
}

/*
 * Provided (sopt, skb) points to received options,
 * build in dopt compiled option set appropriate for answering.
 * i.e. invert SRR option, copy anothers,
 * and grab room in RR/TS options.
 *
 * NOTE: dopt cannot point to skb.
 */

int ip_options_echo(struct ip_options * dopt, struct sk_buff * skb)
{
	struct ip_options *sopt;
	unsigned char *sptr, *dptr;
	int soffset, doffset;
	int	optlen;
	__be32	daddr;

	memset(dopt, 0, sizeof(struct ip_options));

	sopt = &(IPCB(skb)->opt);

	if (sopt->optlen == 0) {
		dopt->optlen = 0;
		return 0;
	}

	sptr = skb_network_header(skb);
	dptr = dopt->__data;

	daddr = skb_rtable(skb)->rt_spec_dst;

	if (sopt->rr) {
		optlen  = sptr[sopt->rr+1];
		soffset = sptr[sopt->rr+2];
		dopt->rr = dopt->optlen + sizeof(struct iphdr);
		memcpy(dptr, sptr+sopt->rr, optlen);
		if (sopt->rr_needaddr && soffset <= optlen) {
			if (soffset + 3 > optlen)
				return -EINVAL;
			dptr[2] = soffset + 4;
			dopt->rr_needaddr = 1;
		}
		dptr += optlen;
		dopt->optlen += optlen;
	}
	if (sopt->ts) {
		optlen = sptr[sopt->ts+1];
		soffset = sptr[sopt->ts+2];
		dopt->ts = dopt->optlen + sizeof(struct iphdr);
		memcpy(dptr, sptr+sopt->ts, optlen);
		if (soffset <= optlen) {
			if (sopt->ts_needaddr) {
				if (soffset + 3 > optlen)
					return -EINVAL;
				dopt->ts_needaddr = 1;
				soffset += 4;
			}
			if (sopt->ts_needtime) {
				if (soffset + 3 > optlen)
					return -EINVAL;
				if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
					dopt->ts_needtime = 1;
					soffset += 4;
				} else {
					dopt->ts_needtime = 0;

					if (soffset + 8 <= optlen) {
						__be32 addr;

						memcpy(&addr, sptr+soffset-1, 4);
						if (inet_addr_type(dev_net(skb_dst(skb)->dev), addr) != RTN_LOCAL) {
							dopt->ts_needtime = 1;
							soffset += 8;
						}
					}
				}
			}
			dptr[2] = soffset;
		}
		dptr += optlen;
		dopt->optlen += optlen;
	}
	if (sopt->srr) {
		unsigned char * start = sptr+sopt->srr;
		__be32 faddr;

		optlen  = start[1];
		soffset = start[2];
		doffset = 0;
		if (soffset > optlen)
			soffset = optlen + 1;
		soffset -= 4;
		if (soffset > 3) {
			memcpy(&faddr, &start[soffset-1], 4);
			for (soffset-=4, doffset=4; soffset > 3; soffset-=4, doffset+=4)
				memcpy(&dptr[doffset-1], &start[soffset-1], 4);
			/*
			 * RFC1812 requires to fix illegal source routes.
			 */
			if (memcmp(&ip_hdr(skb)->saddr,
				   &start[soffset + 3], 4) == 0)
				doffset -= 4;
		}
		if (doffset > 3) {
			memcpy(&start[doffset-1], &daddr, 4);
			dopt->faddr = faddr;
			dptr[0] = start[0];
			dptr[1] = doffset+3;
			dptr[2] = 4;
			dptr += doffset+3;
			dopt->srr = dopt->optlen + sizeof(struct iphdr);
			dopt->optlen += doffset+3;
			dopt->is_strictroute = sopt->is_strictroute;
		}
	}
	if (sopt->cipso) {
		optlen  = sptr[sopt->cipso+1];
		dopt->cipso = dopt->optlen+sizeof(struct iphdr);
		memcpy(dptr, sptr+sopt->cipso, optlen);
		dptr += optlen;
		dopt->optlen += optlen;
	}
	while (dopt->optlen & 3) {
		*dptr++ = IPOPT_END;
		dopt->optlen++;
	}
	return 0;
}

/*
 *	Options "fragmenting", just fill options not
 *	allowed in fragments with NOOPs.
 *	Simple and stupid 8), but the most efficient way.
 */

void ip_options_fragment(struct sk_buff * skb)
{
	unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
	struct ip_options * opt = &(IPCB(skb)->opt);
	int  l = opt->optlen;
	int  optlen;

	while (l > 0) {
		switch (*optptr) {
		case IPOPT_END:
			return;
		case IPOPT_NOOP:
			l--;
			optptr++;
			continue;
		}
		optlen = optptr[1];
		if (optlen<2 || optlen>l)
		  return;
		if (!IPOPT_COPIED(*optptr))
			memset(optptr, IPOPT_NOOP, optlen);
		l -= optlen;
		optptr += optlen;
	}
	opt->ts = 0;
	opt->rr = 0;
	opt->rr_needaddr = 0;
	opt->ts_needaddr = 0;
	opt->ts_needtime = 0;
}

/*
 * Verify options and fill pointers in struct options.
 * Caller should clear *opt, and set opt->data.
 * If opt == NULL, then skb->data should point to IP header.
 */

int ip_options_compile(struct net *net,
		       struct ip_options * opt, struct sk_buff * skb)
{
	int l;
	unsigned char * iph;
	unsigned char * optptr;
	int optlen;
	unsigned char * pp_ptr = NULL;
	struct rtable *rt = NULL;

	if (skb != NULL) {
		rt = skb_rtable(skb);
		optptr = (unsigned char *)&(ip_hdr(skb)[1]);
	} else
		optptr = opt->__data;
	iph = optptr - sizeof(struct iphdr);

	for (l = opt->optlen; l > 0; ) {
		switch (*optptr) {
		      case IPOPT_END:
			for (optptr++, l--; l>0; optptr++, l--) {
				if (*optptr != IPOPT_END) {
					*optptr = IPOPT_END;
					opt->is_changed = 1;
				}
			}
			goto eol;
		      case IPOPT_NOOP:
			l--;
			optptr++;
			continue;
		}
		optlen = optptr[1];
		if (optlen<2 || optlen>l) {
			pp_ptr = optptr;
			goto error;
		}
		switch (*optptr) {
		      case IPOPT_SSRR:
		      case IPOPT_LSRR:
			if (optlen < 3) {
				pp_ptr = optptr + 1;
				goto error;
			}
			if (optptr[2] < 4) {
				pp_ptr = optptr + 2;
				goto error;
			}
			/* NB: cf RFC-1812 5.2.4.1 */
			if (opt->srr) {
				pp_ptr = optptr;
				goto error;
			}
			if (!skb) {
				if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
					pp_ptr = optptr + 1;
					goto error;
				}
				memcpy(&opt->faddr, &optptr[3], 4);
				if (optlen > 7)
					memmove(&optptr[3], &optptr[7], optlen-7);
			}
			opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
			opt->srr = optptr - iph;
			break;
		      case IPOPT_RR:
			if (opt->rr) {
				pp_ptr = optptr;
				goto error;
			}
			if (optlen < 3) {
				pp_ptr = optptr + 1;
				goto error;
			}
			if (optptr[2] < 4) {
				pp_ptr = optptr + 2;
				goto error;
			}
			if (optptr[2] <= optlen) {
				if (optptr[2]+3 > optlen) {
					pp_ptr = optptr + 2;
					goto error;
				}
				if (skb) {
					memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
					opt->is_changed = 1;
				}
				optptr[2] += 4;
				opt->rr_needaddr = 1;
			}
			opt->rr = optptr - iph;
			break;
		      case IPOPT_TIMESTAMP:
			if (opt->ts) {
				pp_ptr = optptr;
				goto error;
			}
			if (optlen < 4) {
				pp_ptr = optptr + 1;
				goto error;
			}
			if (optptr[2] < 5) {
				pp_ptr = optptr + 2;
				goto error;
			}
			if (optptr[2] <= optlen) {
				__be32 *timeptr = NULL;
				if (optptr[2]+3 > optptr[1]) {
					pp_ptr = optptr + 2;
					goto error;
				}
				switch (optptr[3]&0xF) {
				      case IPOPT_TS_TSONLY:
					opt->ts = optptr - iph;
					if (skb)
						timeptr = (__be32*)&optptr[optptr[2]-1];
					opt->ts_needtime = 1;
					optptr[2] += 4;
					break;
				      case IPOPT_TS_TSANDADDR:
					if (optptr[2]+7 > optptr[1]) {
						pp_ptr = optptr + 2;
						goto error;
					}
					opt->ts = optptr - iph;
					if (skb) {
						memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
						timeptr = (__be32*)&optptr[optptr[2]+3];
					}
					opt->ts_needaddr = 1;
					opt->ts_needtime = 1;
					optptr[2] += 8;
					break;
				      case IPOPT_TS_PRESPEC:
					if (optptr[2]+7 > optptr[1]) {
						pp_ptr = optptr + 2;
						goto error;
					}
					opt->ts = optptr - iph;
					{
						__be32 addr;
						memcpy(&addr, &optptr[optptr[2]-1], 4);
						if (inet_addr_type(net, addr) == RTN_UNICAST)
							break;
						if (skb)
							timeptr = (__be32*)&optptr[optptr[2]+3];
					}
					opt->ts_needtime = 1;
					optptr[2] += 8;
					break;
				      default:
					if (!skb && !capable(CAP_NET_RAW)) {
						pp_ptr = optptr + 3;
						goto error;
					}
					break;
				}
				if (timeptr) {
					struct timespec tv;
					__be32  midtime;
					getnstimeofday(&tv);
					midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
					memcpy(timeptr, &midtime, sizeof(__be32));
					opt->is_changed = 1;
				}
			} else {
				unsigned overflow = optptr[3]>>4;
				if (overflow == 15) {
					pp_ptr = optptr + 3;
					goto error;
				}
				opt->ts = optptr - iph;
				if (skb) {
					optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
					opt->is_changed = 1;
				}
			}
			break;
		      case IPOPT_RA:
			if (optlen < 4) {
				pp_ptr = optptr + 1;
				goto error;
			}
			if (optptr[2] == 0 && optptr[3] == 0)
				opt->router_alert = optptr - iph;
			break;
		      case IPOPT_CIPSO:
			if ((!skb && !capable(CAP_NET_RAW)) || opt->cipso) {
				pp_ptr = optptr;
				goto error;
			}
			opt->cipso = optptr - iph;
			if (cipso_v4_validate(skb, &optptr)) {
				pp_ptr = optptr;
				goto error;
			}
			break;
		      case IPOPT_SEC:
		      case IPOPT_SID:
		      default:
			if (!skb && !capable(CAP_NET_RAW)) {
				pp_ptr = optptr;
				goto error;
			}
			break;
		}
		l -= optlen;
		optptr += optlen;
	}

eol:
	if (!pp_ptr)
		return 0;

error:
	if (skb) {
		icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
	}
	return -EINVAL;
}
EXPORT_SYMBOL(ip_options_compile);

/*
 *	Undo all the changes done by ip_options_compile().
 */

void ip_options_undo(struct ip_options * opt)
{
	if (opt->srr) {
		unsigned  char * optptr = opt->__data+opt->srr-sizeof(struct  iphdr);
		memmove(optptr+7, optptr+3, optptr[1]-7);
		memcpy(optptr+3, &opt->faddr, 4);
	}
	if (opt->rr_needaddr) {
		unsigned  char * optptr = opt->__data+opt->rr-sizeof(struct  iphdr);
		optptr[2] -= 4;
		memset(&optptr[optptr[2]-1], 0, 4);
	}
	if (opt->ts) {
		unsigned  char * optptr = opt->__data+opt->ts-sizeof(struct  iphdr);
		if (opt->ts_needtime) {
			optptr[2] -= 4;
			memset(&optptr[optptr[2]-1], 0, 4);
			if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
				optptr[2] -= 4;
		}
		if (opt->ts_needaddr) {
			optptr[2] -= 4;
			memset(&optptr[optptr[2]-1], 0, 4);
		}
	}
}

static struct ip_options *ip_options_get_alloc(const int optlen)
{
	return kzalloc(sizeof(struct ip_options) + ((optlen + 3) & ~3),
		       GFP_KERNEL);
}

static int ip_options_get_finish(struct net *net, struct ip_options **optp,
				 struct ip_options *opt, int optlen)
{
	while (optlen & 3)
		opt->__data[optlen++] = IPOPT_END;
	opt->optlen = optlen;
	if (optlen && ip_options_compile(net, opt, NULL)) {
		kfree(opt);
		return -EINVAL;
	}
	kfree(*optp);
	*optp = opt;
	return 0;
}

int ip_options_get_from_user(struct net *net, struct ip_options **optp,
			     unsigned char __user *data, int optlen)
{
	struct ip_options *opt = ip_options_get_alloc(optlen);

	if (!opt)
		return -ENOMEM;
	if (optlen && copy_from_user(opt->__data, data, optlen)) {
		kfree(opt);
		return -EFAULT;
	}
	return ip_options_get_finish(net, optp, opt, optlen);
}

int ip_options_get(struct net *net, struct ip_options **optp,
		   unsigned char *data, int optlen)
{
	struct ip_options *opt = ip_options_get_alloc(optlen);

	if (!opt)
		return -ENOMEM;
	if (optlen)
		memcpy(opt->__data, data, optlen);
	return ip_options_get_finish(net, optp, opt, optlen);
}

void ip_forward_options(struct sk_buff *skb)
{
	struct   ip_options * opt	= &(IPCB(skb)->opt);
	unsigned char * optptr;
	struct rtable *rt = skb_rtable(skb);
	unsigned char *raw = skb_network_header(skb);

	if (opt->rr_needaddr) {
		optptr = (unsigned char *)raw + opt->rr;
		ip_rt_get_source(&optptr[optptr[2]-5], rt);
		opt->is_changed = 1;
	}
	if (opt->srr_is_hit) {
		int srrptr, srrspace;

		optptr = raw + opt->srr;

		for ( srrptr=optptr[2], srrspace = optptr[1];
		     srrptr <= srrspace;
		     srrptr += 4
		     ) {
			if (srrptr + 3 > srrspace)
				break;
			if (memcmp(&rt->rt_dst, &optptr[srrptr-1], 4) == 0)
				break;
		}
		if (srrptr + 3 <= srrspace) {
			opt->is_changed = 1;
			ip_rt_get_source(&optptr[srrptr-1], rt);
			ip_hdr(skb)->daddr = rt->rt_dst;
			optptr[2] = srrptr+4;
		} else if (net_ratelimit())
			printk(KERN_CRIT "ip_forward(): Argh! Destination lost!\n");
		if (opt->ts_needaddr) {
			optptr = raw + opt->ts;
			ip_rt_get_source(&optptr[optptr[2]-9], rt);
			opt->is_changed = 1;
		}
	}
	if (opt->is_changed) {
		opt->is_changed = 0;
		ip_send_check(ip_hdr(skb));
	}
}

int ip_options_rcv_srr(struct sk_buff *skb)
{
	struct ip_options *opt = &(IPCB(skb)->opt);
	int srrspace, srrptr;
	__be32 nexthop;
	struct iphdr *iph = ip_hdr(skb);
	unsigned char *optptr = skb_network_header(skb) + opt->srr;
	struct rtable *rt = skb_rtable(skb);
	struct rtable *rt2;
	unsigned long orefdst;
	int err;

	if (!opt->srr)
		return 0;

	if (skb->pkt_type != PACKET_HOST)
		return -EINVAL;
	if (rt->rt_type == RTN_UNICAST) {
		if (!opt->is_strictroute)
			return 0;
		icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
		return -EINVAL;
	}
	if (rt->rt_type != RTN_LOCAL)
		return -EINVAL;

	for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
		if (srrptr + 3 > srrspace) {
			icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
			return -EINVAL;
		}
		memcpy(&nexthop, &optptr[srrptr-1], 4);

		orefdst = skb->_skb_refdst;
		skb_dst_set(skb, NULL);
		err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
		rt2 = skb_rtable(skb);
		if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
			skb_dst_drop(skb);
			skb->_skb_refdst = orefdst;
			return -EINVAL;
		}
		refdst_drop(orefdst);
		if (rt2->rt_type != RTN_LOCAL)
			break;
		/* Superfast 8) loopback forward */
		memcpy(&iph->daddr, &optptr[srrptr-1], 4);
		opt->is_changed = 1;
	}
	if (srrptr <= srrspace) {
		opt->srr_is_hit = 1;
		opt->is_changed = 1;
	}
	return 0;
}
EXPORT_SYMBOL(ip_options_rcv_srr);
Commit	Line	Data
1da177e4 LT	1	/*
	2	* INET An implementation of the TCP/IP protocol suite for the LINUX
	3	* operating system. INET is implemented using the BSD Socket
	4	* interface as the means of communication with the user level.
	5	*
	6	* The options processing module for ip.c
	7	*
1da177e4	8	* Authors: A.N.Kuznetsov
e905a9ed	9	*
1da177e4 LT	10	*/
1da177e4 LT	11
4fc268d2	12	#include <linux/capability.h>
1da177e4	13	#include <linux/module.h>
5a0e3ad6	14	#include <linux/slab.h>
1da177e4 LT	15	#include <linux/types.h>
	16	#include <asm/uaccess.h>
	17	#include <linux/skbuff.h>
	18	#include <linux/ip.h>
	19	#include <linux/icmp.h>
	20	#include <linux/netdevice.h>
	21	#include <linux/rtnetlink.h>
	22	#include <net/sock.h>
	23	#include <net/ip.h>
	24	#include <net/icmp.h>
14c85021	25	#include <net/route.h>
11a03f78	26	#include <net/cipso_ipv4.h>
1da177e4	27
e905a9ed	28	/*
1da177e4 LT	29	* Write options to IP header, record destination address to
	30	* source route option, address of outgoing interface
	31	* (we should already know it, so that this function is allowed be
	32	* called only after routing decision) and timestamp,
	33	* if we originate this datagram.
	34	*
	35	* daddr is real destination address, next hop is recorded in IP header.
	36	* saddr is address of outgoing interface.
	37	*/
	38
	39	void ip_options_build(struct sk_buff * skb, struct ip_options * opt,
8712f774	40	__be32 daddr, struct rtable *rt, int is_frag)
1da177e4	41	{
d56f90a7	42	unsigned char *iph = skb_network_header(skb);
1da177e4 LT	43
	44	memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
	45	memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
	46	opt = &(IPCB(skb)->opt);
1da177e4 LT	47
	48	if (opt->srr)
	49	memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
	50
	51	if (!is_frag) {
	52	if (opt->rr_needaddr)
	53	ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, rt);
	54	if (opt->ts_needaddr)
	55	ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, rt);
	56	if (opt->ts_needtime) {
f25c3d61	57	struct timespec tv;
e25d2ca6	58	__be32 midtime;
f25c3d61 YH	59	getnstimeofday(&tv);
f25c3d61 YH	60	midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
1da177e4 LT	61	memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
	62	}
	63	return;
	64	}
	65	if (opt->rr) {
	66	memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
	67	opt->rr = 0;
	68	opt->rr_needaddr = 0;
	69	}
	70	if (opt->ts) {
	71	memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
	72	opt->ts = 0;
	73	opt->ts_needaddr = opt->ts_needtime = 0;
	74	}
	75	}
	76
e905a9ed	77	/*
1da177e4 LT	78	* Provided (sopt, skb) points to received options,
	79	* build in dopt compiled option set appropriate for answering.
	80	* i.e. invert SRR option, copy anothers,
	81	* and grab room in RR/TS options.
	82	*
	83	* NOTE: dopt cannot point to skb.
	84	*/
	85
e905a9ed	86	int ip_options_echo(struct ip_options * dopt, struct sk_buff * skb)
1da177e4 LT	87	{
	88	struct ip_options *sopt;
	89	unsigned char sptr, dptr;
	90	int soffset, doffset;
	91	int optlen;
e25d2ca6	92	__be32 daddr;
1da177e4 LT	93
	94	memset(dopt, 0, sizeof(struct ip_options));
	95
1da177e4 LT	96	sopt = &(IPCB(skb)->opt);
	97
	98	if (sopt->optlen == 0) {
	99	dopt->optlen = 0;
	100	return 0;
	101	}
	102
d56f90a7	103	sptr = skb_network_header(skb);
1da177e4 LT	104	dptr = dopt->__data;
1da177e4 LT	105
511c3f92	106	daddr = skb_rtable(skb)->rt_spec_dst;
1da177e4 LT	107
	108	if (sopt->rr) {
	109	optlen = sptr[sopt->rr+1];
	110	soffset = sptr[sopt->rr+2];
	111	dopt->rr = dopt->optlen + sizeof(struct iphdr);
	112	memcpy(dptr, sptr+sopt->rr, optlen);
	113	if (sopt->rr_needaddr && soffset <= optlen) {
	114	if (soffset + 3 > optlen)
	115	return -EINVAL;
	116	dptr[2] = soffset + 4;
	117	dopt->rr_needaddr = 1;
	118	}
	119	dptr += optlen;
	120	dopt->optlen += optlen;
	121	}
	122	if (sopt->ts) {
	123	optlen = sptr[sopt->ts+1];
	124	soffset = sptr[sopt->ts+2];
	125	dopt->ts = dopt->optlen + sizeof(struct iphdr);
	126	memcpy(dptr, sptr+sopt->ts, optlen);
	127	if (soffset <= optlen) {
	128	if (sopt->ts_needaddr) {
	129	if (soffset + 3 > optlen)
	130	return -EINVAL;
	131	dopt->ts_needaddr = 1;
	132	soffset += 4;
	133	}
	134	if (sopt->ts_needtime) {
	135	if (soffset + 3 > optlen)
	136	return -EINVAL;
	137	if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
	138	dopt->ts_needtime = 1;
	139	soffset += 4;
	140	} else {
	141	dopt->ts_needtime = 0;
	142
	143	if (soffset + 8 <= optlen) {
fd683222	144	__be32 addr;
1da177e4 LT	145
1da177e4 LT	146	memcpy(&addr, sptr+soffset-1, 4);
adf30907	147	if (inet_addr_type(dev_net(skb_dst(skb)->dev), addr) != RTN_LOCAL) {
1da177e4 LT	148	dopt->ts_needtime = 1;
	149	soffset += 8;
	150	}
	151	}
	152	}
	153	}
	154	dptr[2] = soffset;
	155	}
	156	dptr += optlen;
	157	dopt->optlen += optlen;
	158	}
	159	if (sopt->srr) {
	160	unsigned char * start = sptr+sopt->srr;
3ca3c68e	161	__be32 faddr;
1da177e4 LT	162
	163	optlen = start[1];
	164	soffset = start[2];
	165	doffset = 0;
	166	if (soffset > optlen)
	167	soffset = optlen + 1;
	168	soffset -= 4;
	169	if (soffset > 3) {
	170	memcpy(&faddr, &start[soffset-1], 4);
	171	for (soffset-=4, doffset=4; soffset > 3; soffset-=4, doffset+=4)
	172	memcpy(&dptr[doffset-1], &start[soffset-1], 4);
	173	/*
	174	* RFC1812 requires to fix illegal source routes.
	175	*/
eddc9ec5 ACM	176	if (memcmp(&ip_hdr(skb)->saddr,
eddc9ec5 ACM	177	&start[soffset + 3], 4) == 0)
1da177e4 LT	178	doffset -= 4;
	179	}
	180	if (doffset > 3) {
	181	memcpy(&start[doffset-1], &daddr, 4);
	182	dopt->faddr = faddr;
	183	dptr[0] = start[0];
	184	dptr[1] = doffset+3;
	185	dptr[2] = 4;
	186	dptr += doffset+3;
	187	dopt->srr = dopt->optlen + sizeof(struct iphdr);
	188	dopt->optlen += doffset+3;
	189	dopt->is_strictroute = sopt->is_strictroute;
	190	}
	191	}
11a03f78 PM	192	if (sopt->cipso) {
	193	optlen = sptr[sopt->cipso+1];
	194	dopt->cipso = dopt->optlen+sizeof(struct iphdr);
	195	memcpy(dptr, sptr+sopt->cipso, optlen);
	196	dptr += optlen;
	197	dopt->optlen += optlen;
	198	}
1da177e4 LT	199	while (dopt->optlen & 3) {
	200	*dptr++ = IPOPT_END;
	201	dopt->optlen++;
	202	}
	203	return 0;
	204	}
	205
	206	/*
	207	* Options "fragmenting", just fill options not
	208	* allowed in fragments with NOOPs.
	209	* Simple and stupid 8), but the most efficient way.
	210	*/
	211
e905a9ed	212	void ip_options_fragment(struct sk_buff * skb)
1da177e4	213	{
d56f90a7	214	unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
1da177e4 LT	215	struct ip_options * opt = &(IPCB(skb)->opt);
	216	int l = opt->optlen;
	217	int optlen;
	218
	219	while (l > 0) {
	220	switch (*optptr) {
	221	case IPOPT_END:
	222	return;
	223	case IPOPT_NOOP:
	224	l--;
	225	optptr++;
	226	continue;
	227	}
	228	optlen = optptr[1];
	229	if (optlen<2 \|\| optlen>l)
	230	return;
	231	if (!IPOPT_COPIED(*optptr))
	232	memset(optptr, IPOPT_NOOP, optlen);
	233	l -= optlen;
	234	optptr += optlen;
	235	}
	236	opt->ts = 0;
	237	opt->rr = 0;
	238	opt->rr_needaddr = 0;
	239	opt->ts_needaddr = 0;
	240	opt->ts_needtime = 0;
1da177e4 LT	241	}
	242
	243	/*
	244	* Verify options and fill pointers in struct options.
	245	* Caller should clear *opt, and set opt->data.
	246	* If opt == NULL, then skb->data should point to IP header.
	247	*/
	248
0e6bd4a1 DL	249	int ip_options_compile(struct net *net,
0e6bd4a1 DL	250	struct ip_options * opt, struct sk_buff * skb)
1da177e4 LT	251	{
	252	int l;
	253	unsigned char * iph;
	254	unsigned char * optptr;
	255	int optlen;
	256	unsigned char * pp_ptr = NULL;
22aba383	257	struct rtable *rt = NULL;
1da177e4	258
22aba383	259	if (skb != NULL) {
511c3f92	260	rt = skb_rtable(skb);
22aba383 DL	261	optptr = (unsigned char *)&(ip_hdr(skb)[1]);
22aba383 DL	262	} else
10fe7d85	263	optptr = opt->__data;
22aba383	264	iph = optptr - sizeof(struct iphdr);
1da177e4 LT	265
	266	for (l = opt->optlen; l > 0; ) {
	267	switch (*optptr) {
	268	case IPOPT_END:
	269	for (optptr++, l--; l>0; optptr++, l--) {
	270	if (*optptr != IPOPT_END) {
	271	*optptr = IPOPT_END;
	272	opt->is_changed = 1;
	273	}
	274	}
	275	goto eol;
	276	case IPOPT_NOOP:
	277	l--;
	278	optptr++;
	279	continue;
	280	}
	281	optlen = optptr[1];
	282	if (optlen<2 \|\| optlen>l) {
	283	pp_ptr = optptr;
	284	goto error;
	285	}
	286	switch (*optptr) {
	287	case IPOPT_SSRR:
	288	case IPOPT_LSRR:
	289	if (optlen < 3) {
	290	pp_ptr = optptr + 1;
	291	goto error;
	292	}
	293	if (optptr[2] < 4) {
	294	pp_ptr = optptr + 2;
	295	goto error;
	296	}
	297	/* NB: cf RFC-1812 5.2.4.1 */
	298	if (opt->srr) {
	299	pp_ptr = optptr;
	300	goto error;
	301	}
	302	if (!skb) {
	303	if (optptr[2] != 4 \|\| optlen < 7 \|\| ((optlen-3) & 3)) {
	304	pp_ptr = optptr + 1;
	305	goto error;
	306	}
	307	memcpy(&opt->faddr, &optptr[3], 4);
	308	if (optlen > 7)
	309	memmove(&optptr[3], &optptr[7], optlen-7);
	310	}
	311	opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
	312	opt->srr = optptr - iph;
	313	break;
	314	case IPOPT_RR:
	315	if (opt->rr) {
	316	pp_ptr = optptr;
	317	goto error;
	318	}
	319	if (optlen < 3) {
	320	pp_ptr = optptr + 1;
	321	goto error;
	322	}
	323	if (optptr[2] < 4) {
	324	pp_ptr = optptr + 2;
	325	goto error;
	326	}
	327	if (optptr[2] <= optlen) {
	328	if (optptr[2]+3 > optlen) {
329	pp_ptr = optptr + 2;
330	goto error;
331	}
332	if (skb) {
333	memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
334	opt->is_changed = 1;
335	}
336	optptr[2] += 4;
337	opt->rr_needaddr = 1;
338	}
339	opt->rr = optptr - iph;
340	break;
341	case IPOPT_TIMESTAMP:
342	if (opt->ts) {
343	pp_ptr = optptr;
344	goto error;
345	}
346	if (optlen < 4) {
347	pp_ptr = optptr + 1;
348	goto error;
349	}
350	if (optptr[2] < 5) {
351	pp_ptr = optptr + 2;
352	goto error;
353	}
354	if (optptr[2] <= optlen) {
e25d2ca6	355	__be32 *timeptr = NULL;
1da177e4 LT	356	if (optptr[2]+3 > optptr[1]) {
	357	pp_ptr = optptr + 2;
	358	goto error;
	359	}
	360	switch (optptr[3]&0xF) {
	361	case IPOPT_TS_TSONLY:
	362	opt->ts = optptr - iph;
e905a9ed	363	if (skb)
e25d2ca6	364	timeptr = (__be32*)&optptr[optptr[2]-1];
1da177e4 LT	365	opt->ts_needtime = 1;
	366	optptr[2] += 4;
	367	break;
	368	case IPOPT_TS_TSANDADDR:
	369	if (optptr[2]+7 > optptr[1]) {
	370	pp_ptr = optptr + 2;
	371	goto error;
	372	}
	373	opt->ts = optptr - iph;
	374	if (skb) {
	375	memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
e25d2ca6	376	timeptr = (__be32*)&optptr[optptr[2]+3];
1da177e4 LT	377	}
	378	opt->ts_needaddr = 1;
	379	opt->ts_needtime = 1;
	380	optptr[2] += 8;
	381	break;
	382	case IPOPT_TS_PRESPEC:
	383	if (optptr[2]+7 > optptr[1]) {
	384	pp_ptr = optptr + 2;
	385	goto error;
	386	}
	387	opt->ts = optptr - iph;
	388	{
fd683222	389	__be32 addr;
1da177e4	390	memcpy(&addr, &optptr[optptr[2]-1], 4);
0e6bd4a1	391	if (inet_addr_type(net, addr) == RTN_UNICAST)
1da177e4 LT	392	break;
1da177e4 LT	393	if (skb)
e25d2ca6	394	timeptr = (__be32*)&optptr[optptr[2]+3];
1da177e4 LT	395	}
	396	opt->ts_needtime = 1;
	397	optptr[2] += 8;
	398	break;
	399	default:
	400	if (!skb && !capable(CAP_NET_RAW)) {
	401	pp_ptr = optptr + 3;
	402	goto error;
	403	}
	404	break;
	405	}
	406	if (timeptr) {
f25c3d61	407	struct timespec tv;
e25d2ca6	408	__be32 midtime;
f25c3d61 YH	409	getnstimeofday(&tv);
f25c3d61 YH	410	midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
e25d2ca6	411	memcpy(timeptr, &midtime, sizeof(__be32));
1da177e4 LT	412	opt->is_changed = 1;
	413	}
	414	} else {
	415	unsigned overflow = optptr[3]>>4;
	416	if (overflow == 15) {
	417	pp_ptr = optptr + 3;
	418	goto error;
	419	}
	420	opt->ts = optptr - iph;
	421	if (skb) {
	422	optptr[3] = (optptr[3]&0xF)\|((overflow+1)<<4);
	423	opt->is_changed = 1;
	424	}
	425	}
	426	break;
	427	case IPOPT_RA:
	428	if (optlen < 4) {
	429	pp_ptr = optptr + 1;
	430	goto error;
	431	}
	432	if (optptr[2] == 0 && optptr[3] == 0)
	433	opt->router_alert = optptr - iph;
	434	break;
11a03f78	435	case IPOPT_CIPSO:
f8687afe	436	if ((!skb && !capable(CAP_NET_RAW)) \|\| opt->cipso) {
11a03f78 PM	437	pp_ptr = optptr;
	438	goto error;
	439	}
	440	opt->cipso = optptr - iph;
15c45f7b	441	if (cipso_v4_validate(skb, &optptr)) {
11a03f78 PM	442	pp_ptr = optptr;
	443	goto error;
	444	}
	445	break;
1da177e4 LT	446	case IPOPT_SEC:
	447	case IPOPT_SID:
	448	default:
	449	if (!skb && !capable(CAP_NET_RAW)) {
	450	pp_ptr = optptr;
	451	goto error;
	452	}
	453	break;
	454	}
	455	l -= optlen;
	456	optptr += optlen;
	457	}
	458
	459	eol:
	460	if (!pp_ptr)
	461	return 0;
	462
	463	error:
	464	if (skb) {
	465	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
	466	}
	467	return -EINVAL;
	468	}
462fb2af	469	EXPORT_SYMBOL(ip_options_compile);
1da177e4 LT	470
	471	/*
	472	* Undo all the changes done by ip_options_compile().
	473	*/
	474
	475	void ip_options_undo(struct ip_options * opt)
	476	{
	477	if (opt->srr) {
	478	unsigned char * optptr = opt->__data+opt->srr-sizeof(struct iphdr);
	479	memmove(optptr+7, optptr+3, optptr[1]-7);
	480	memcpy(optptr+3, &opt->faddr, 4);
	481	}
	482	if (opt->rr_needaddr) {
	483	unsigned char * optptr = opt->__data+opt->rr-sizeof(struct iphdr);
	484	optptr[2] -= 4;
	485	memset(&optptr[optptr[2]-1], 0, 4);
	486	}
	487	if (opt->ts) {
	488	unsigned char * optptr = opt->__data+opt->ts-sizeof(struct iphdr);
	489	if (opt->ts_needtime) {
	490	optptr[2] -= 4;
	491	memset(&optptr[optptr[2]-1], 0, 4);
	492	if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
	493	optptr[2] -= 4;
	494	}
	495	if (opt->ts_needaddr) {
	496	optptr[2] -= 4;
	497	memset(&optptr[optptr[2]-1], 0, 4);
	498	}
	499	}
	500	}
	501
4c6ea29d	502	static struct ip_options *ip_options_get_alloc(const int optlen)
1da177e4	503	{
37640703 MK	504	return kzalloc(sizeof(struct ip_options) + ((optlen + 3) & ~3),
37640703 MK	505	GFP_KERNEL);
4c6ea29d	506	}
1da177e4	507
f2c4802b	508	static int ip_options_get_finish(struct net net, struct ip_options *optp,
4c6ea29d ACM	509	struct ip_options *opt, int optlen)
4c6ea29d ACM	510	{
1da177e4 LT	511	while (optlen & 3)
	512	opt->__data[optlen++] = IPOPT_END;
	513	opt->optlen = optlen;
f2c4802b	514	if (optlen && ip_options_compile(net, opt, NULL)) {
1da177e4 LT	515	kfree(opt);
	516	return -EINVAL;
	517	}
a51482bd	518	kfree(*optp);
1da177e4 LT	519	*optp = opt;
	520	return 0;
	521	}
	522
f2c4802b DL	523	int ip_options_get_from_user(struct net net, struct ip_options *optp,
f2c4802b DL	524	unsigned char __user *data, int optlen)
4c6ea29d ACM	525	{
	526	struct ip_options *opt = ip_options_get_alloc(optlen);
	527
	528	if (!opt)
	529	return -ENOMEM;
	530	if (optlen && copy_from_user(opt->__data, data, optlen)) {
	531	kfree(opt);
	532	return -EFAULT;
	533	}
f2c4802b	534	return ip_options_get_finish(net, optp, opt, optlen);
4c6ea29d ACM	535	}
4c6ea29d ACM	536
f2c4802b DL	537	int ip_options_get(struct net net, struct ip_options *optp,
f2c4802b DL	538	unsigned char *data, int optlen)
4c6ea29d ACM	539	{
	540	struct ip_options *opt = ip_options_get_alloc(optlen);
	541
	542	if (!opt)
	543	return -ENOMEM;
	544	if (optlen)
	545	memcpy(opt->__data, data, optlen);
f2c4802b	546	return ip_options_get_finish(net, optp, opt, optlen);
4c6ea29d ACM	547	}
4c6ea29d ACM	548
1da177e4 LT	549	void ip_forward_options(struct sk_buff *skb)
	550	{
	551	struct ip_options * opt = &(IPCB(skb)->opt);
	552	unsigned char * optptr;
511c3f92	553	struct rtable *rt = skb_rtable(skb);
d56f90a7	554	unsigned char *raw = skb_network_header(skb);
1da177e4 LT	555
	556	if (opt->rr_needaddr) {
	557	optptr = (unsigned char *)raw + opt->rr;
	558	ip_rt_get_source(&optptr[optptr[2]-5], rt);
	559	opt->is_changed = 1;
	560	}
	561	if (opt->srr_is_hit) {
	562	int srrptr, srrspace;
	563
	564	optptr = raw + opt->srr;
	565
	566	for ( srrptr=optptr[2], srrspace = optptr[1];
	567	srrptr <= srrspace;
	568	srrptr += 4
	569	) {
	570	if (srrptr + 3 > srrspace)
	571	break;
	572	if (memcmp(&rt->rt_dst, &optptr[srrptr-1], 4) == 0)
	573	break;
	574	}
	575	if (srrptr + 3 <= srrspace) {
	576	opt->is_changed = 1;
	577	ip_rt_get_source(&optptr[srrptr-1], rt);
eddc9ec5	578	ip_hdr(skb)->daddr = rt->rt_dst;
1da177e4 LT	579	optptr[2] = srrptr+4;
	580	} else if (net_ratelimit())
	581	printk(KERN_CRIT "ip_forward(): Argh! Destination lost!\n");
	582	if (opt->ts_needaddr) {
	583	optptr = raw + opt->ts;
	584	ip_rt_get_source(&optptr[optptr[2]-9], rt);
	585	opt->is_changed = 1;
	586	}
	587	}
	588	if (opt->is_changed) {
	589	opt->is_changed = 0;
eddc9ec5	590	ip_send_check(ip_hdr(skb));
1da177e4 LT	591	}
	592	}
	593
	594	int ip_options_rcv_srr(struct sk_buff *skb)
	595	{
	596	struct ip_options *opt = &(IPCB(skb)->opt);
	597	int srrspace, srrptr;
9e12bb22	598	__be32 nexthop;
eddc9ec5	599	struct iphdr *iph = ip_hdr(skb);
d56f90a7	600	unsigned char *optptr = skb_network_header(skb) + opt->srr;
511c3f92	601	struct rtable *rt = skb_rtable(skb);
1da177e4	602	struct rtable *rt2;
7fee226a	603	unsigned long orefdst;
1da177e4 LT	604	int err;
	605
	606	if (!opt->srr)
	607	return 0;
	608
	609	if (skb->pkt_type != PACKET_HOST)
	610	return -EINVAL;
	611	if (rt->rt_type == RTN_UNICAST) {
	612	if (!opt->is_strictroute)
	613	return 0;
	614	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
	615	return -EINVAL;
	616	}
	617	if (rt->rt_type != RTN_LOCAL)
	618	return -EINVAL;
	619
	620	for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
	621	if (srrptr + 3 > srrspace) {
	622	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
	623	return -EINVAL;
	624	}
	625	memcpy(&nexthop, &optptr[srrptr-1], 4);
	626
7fee226a	627	orefdst = skb->_skb_refdst;
adf30907	628	skb_dst_set(skb, NULL);
1da177e4	629	err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
511c3f92	630	rt2 = skb_rtable(skb);
1da177e4	631	if (err \|\| (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
7fee226a ED	632	skb_dst_drop(skb);
7fee226a ED	633	skb->_skb_refdst = orefdst;
1da177e4 LT	634	return -EINVAL;
1da177e4 LT	635	}
7fee226a	636	refdst_drop(orefdst);
1da177e4 LT	637	if (rt2->rt_type != RTN_LOCAL)
	638	break;
	639	/* Superfast 8) loopback forward */
	640	memcpy(&iph->daddr, &optptr[srrptr-1], 4);
	641	opt->is_changed = 1;
	642	}
	643	if (srrptr <= srrspace) {
	644	opt->srr_is_hit = 1;
	645	opt->is_changed = 1;
	646	}
	647	return 0;
	648	}
462fb2af	649	EXPORT_SYMBOL(ip_options_rcv_srr);