[deliverable/linux.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 <asm/unaligned.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, skb, rt);
		if (opt->ts_needaddr)
			ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, skb, 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)
{
	const 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)
		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 + 7 <= optlen) {
						__be32 addr;

						memcpy(&addr, dptr+soffset-1, 4);
						if (inet_addr_type(dev_net(skb_dst(skb)->dev), addr) != RTN_UNICAST) {
							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 (rt) {
					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) {
				unsigned char *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 = &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 (rt)  {
						memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
						timeptr = &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 = &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;
					u32  midtime;
					getnstimeofday(&tv);
					midtime = (tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC;
					put_unaligned_be32(midtime, timeptr);
					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_rcu *ip_options_get_alloc(const int optlen)
{
	return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
		       GFP_KERNEL);
}

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

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

	if (!opt)
		return -ENOMEM;
	if (optlen && copy_from_user(opt->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_rcu **optp,
		   unsigned char *data, int optlen)
{
	struct ip_options_rcu *opt = ip_options_get_alloc(optlen);

	if (!opt)
		return -ENOMEM;
	if (optlen)
		memcpy(opt->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], skb, 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(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
				break;
		}
		if (srrptr + 3 <= srrspace) {
			opt->is_changed = 1;
			ip_hdr(skb)->daddr = opt->nexthop;
			ip_rt_get_source(&optptr[srrptr-1], skb, rt);
			optptr[2] = srrptr+4;
		} else if (net_ratelimit())
			pr_crit("%s(): Argh! Destination lost!\n", __func__);
		if (opt->ts_needaddr) {
			optptr = raw + opt->ts;
			ip_rt_get_source(&optptr[optptr[2]-9], skb, 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 (!rt)
		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 */
		iph->daddr = nexthop;
		opt->is_changed = 1;
	}
	if (srrptr <= srrspace) {
		opt->srr_is_hit = 1;
		opt->nexthop = nexthop;
		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>
1da177e4 LT	16	#include <asm/uaccess.h>
48bdf072	17	#include <asm/unaligned.h>
1da177e4 LT	18	#include <linux/skbuff.h>
	19	#include <linux/ip.h>
	20	#include <linux/icmp.h>
	21	#include <linux/netdevice.h>
	22	#include <linux/rtnetlink.h>
	23	#include <net/sock.h>
	24	#include <net/ip.h>
	25	#include <net/icmp.h>
14c85021	26	#include <net/route.h>
11a03f78	27	#include <net/cipso_ipv4.h>
1da177e4	28
e905a9ed	29	/*
1da177e4 LT	30	* Write options to IP header, record destination address to
	31	* source route option, address of outgoing interface
	32	* (we should already know it, so that this function is allowed be
	33	* called only after routing decision) and timestamp,
	34	* if we originate this datagram.
	35	*
	36	* daddr is real destination address, next hop is recorded in IP header.
	37	* saddr is address of outgoing interface.
	38	*/
	39
f6d8bd05	40	void ip_options_build(struct sk_buff skb, struct ip_options opt,
8e36360a	41	__be32 daddr, struct rtable *rt, int is_frag)
1da177e4	42	{
d56f90a7	43	unsigned char *iph = skb_network_header(skb);
1da177e4 LT	44
	45	memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
	46	memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
	47	opt = &(IPCB(skb)->opt);
1da177e4 LT	48
	49	if (opt->srr)
	50	memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
	51
	52	if (!is_frag) {
	53	if (opt->rr_needaddr)
8e36360a	54	ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, skb, rt);
1da177e4	55	if (opt->ts_needaddr)
8e36360a	56	ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, skb, rt);
1da177e4	57	if (opt->ts_needtime) {
f25c3d61	58	struct timespec tv;
e25d2ca6	59	__be32 midtime;
f25c3d61 YH	60	getnstimeofday(&tv);
f25c3d61 YH	61	midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
1da177e4 LT	62	memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
	63	}
	64	return;
	65	}
	66	if (opt->rr) {
	67	memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
	68	opt->rr = 0;
	69	opt->rr_needaddr = 0;
	70	}
	71	if (opt->ts) {
	72	memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
	73	opt->ts = 0;
	74	opt->ts_needaddr = opt->ts_needtime = 0;
	75	}
	76	}
	77
e905a9ed	78	/*
1da177e4 LT	79	* Provided (sopt, skb) points to received options,
	80	* build in dopt compiled option set appropriate for answering.
	81	* i.e. invert SRR option, copy anothers,
	82	* and grab room in RR/TS options.
	83	*
	84	* NOTE: dopt cannot point to skb.
	85	*/
	86
f6d8bd05	87	int ip_options_echo(struct ip_options dopt, struct sk_buff skb)
1da177e4	88	{
f6d8bd05	89	const struct ip_options *sopt;
1da177e4 LT	90	unsigned char sptr, dptr;
	91	int soffset, doffset;
	92	int optlen;
e25d2ca6	93	__be32 daddr;
1da177e4 LT	94
	95	memset(dopt, 0, sizeof(struct ip_options));
	96
1da177e4 LT	97	sopt = &(IPCB(skb)->opt);
1da177e4 LT	98
f6d8bd05	99	if (sopt->optlen == 0)
1da177e4	100	return 0;
1da177e4	101
d56f90a7	102	sptr = skb_network_header(skb);
1da177e4 LT	103	dptr = dopt->__data;
1da177e4 LT	104
511c3f92	105	daddr = skb_rtable(skb)->rt_spec_dst;
1da177e4 LT	106
	107	if (sopt->rr) {
	108	optlen = sptr[sopt->rr+1];
	109	soffset = sptr[sopt->rr+2];
	110	dopt->rr = dopt->optlen + sizeof(struct iphdr);
	111	memcpy(dptr, sptr+sopt->rr, optlen);
	112	if (sopt->rr_needaddr && soffset <= optlen) {
	113	if (soffset + 3 > optlen)
	114	return -EINVAL;
	115	dptr[2] = soffset + 4;
	116	dopt->rr_needaddr = 1;
	117	}
	118	dptr += optlen;
	119	dopt->optlen += optlen;
	120	}
	121	if (sopt->ts) {
	122	optlen = sptr[sopt->ts+1];
	123	soffset = sptr[sopt->ts+2];
	124	dopt->ts = dopt->optlen + sizeof(struct iphdr);
	125	memcpy(dptr, sptr+sopt->ts, optlen);
	126	if (soffset <= optlen) {
	127	if (sopt->ts_needaddr) {
	128	if (soffset + 3 > optlen)
	129	return -EINVAL;
	130	dopt->ts_needaddr = 1;
	131	soffset += 4;
	132	}
	133	if (sopt->ts_needtime) {
	134	if (soffset + 3 > optlen)
	135	return -EINVAL;
	136	if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
	137	dopt->ts_needtime = 1;
	138	soffset += 4;
	139	} else {
	140	dopt->ts_needtime = 0;
	141
8628bd8a	142	if (soffset + 7 <= optlen) {
fd683222	143	__be32 addr;
1da177e4	144
8628bd8a JL	145	memcpy(&addr, dptr+soffset-1, 4);
8628bd8a JL	146	if (inet_addr_type(dev_net(skb_dst(skb)->dev), addr) != RTN_UNICAST) {
1da177e4 LT	147	dopt->ts_needtime = 1;
	148	soffset += 8;
	149	}
	150	}
	151	}
	152	}
	153	dptr[2] = soffset;
	154	}
	155	dptr += optlen;
	156	dopt->optlen += optlen;
	157	}
	158	if (sopt->srr) {
f6d8bd05	159	unsigned char *start = sptr+sopt->srr;
3ca3c68e	160	__be32 faddr;
1da177e4 LT	161
	162	optlen = start[1];
	163	soffset = start[2];
	164	doffset = 0;
	165	if (soffset > optlen)
	166	soffset = optlen + 1;
	167	soffset -= 4;
	168	if (soffset > 3) {
	169	memcpy(&faddr, &start[soffset-1], 4);
	170	for (soffset-=4, doffset=4; soffset > 3; soffset-=4, doffset+=4)
	171	memcpy(&dptr[doffset-1], &start[soffset-1], 4);
	172	/*
	173	* RFC1812 requires to fix illegal source routes.
	174	*/
eddc9ec5 ACM	175	if (memcmp(&ip_hdr(skb)->saddr,
eddc9ec5 ACM	176	&start[soffset + 3], 4) == 0)
1da177e4 LT	177	doffset -= 4;
	178	}
	179	if (doffset > 3) {
	180	memcpy(&start[doffset-1], &daddr, 4);
	181	dopt->faddr = faddr;
	182	dptr[0] = start[0];
	183	dptr[1] = doffset+3;
	184	dptr[2] = 4;
	185	dptr += doffset+3;
	186	dopt->srr = dopt->optlen + sizeof(struct iphdr);
	187	dopt->optlen += doffset+3;
	188	dopt->is_strictroute = sopt->is_strictroute;
	189	}
	190	}
11a03f78 PM	191	if (sopt->cipso) {
	192	optlen = sptr[sopt->cipso+1];
	193	dopt->cipso = dopt->optlen+sizeof(struct iphdr);
	194	memcpy(dptr, sptr+sopt->cipso, optlen);
	195	dptr += optlen;
	196	dopt->optlen += optlen;
	197	}
1da177e4 LT	198	while (dopt->optlen & 3) {
	199	*dptr++ = IPOPT_END;
	200	dopt->optlen++;
	201	}
	202	return 0;
	203	}
	204
	205	/*
	206	* Options "fragmenting", just fill options not
	207	* allowed in fragments with NOOPs.
	208	* Simple and stupid 8), but the most efficient way.
	209	*/
	210
e905a9ed	211	void ip_options_fragment(struct sk_buff * skb)
1da177e4	212	{
d56f90a7	213	unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
1da177e4 LT	214	struct ip_options * opt = &(IPCB(skb)->opt);
	215	int l = opt->optlen;
	216	int optlen;
	217
	218	while (l > 0) {
	219	switch (*optptr) {
	220	case IPOPT_END:
	221	return;
	222	case IPOPT_NOOP:
	223	l--;
	224	optptr++;
	225	continue;
	226	}
	227	optlen = optptr[1];
	228	if (optlen<2 \|\| optlen>l)
	229	return;
	230	if (!IPOPT_COPIED(*optptr))
	231	memset(optptr, IPOPT_NOOP, optlen);
	232	l -= optlen;
	233	optptr += optlen;
	234	}
	235	opt->ts = 0;
	236	opt->rr = 0;
	237	opt->rr_needaddr = 0;
	238	opt->ts_needaddr = 0;
	239	opt->ts_needtime = 0;
1da177e4 LT	240	}
	241
	242	/*
	243	* Verify options and fill pointers in struct options.
	244	* Caller should clear *opt, and set opt->data.
	245	* If opt == NULL, then skb->data should point to IP header.
	246	*/
	247
0e6bd4a1 DL	248	int ip_options_compile(struct net *net,
0e6bd4a1 DL	249	struct ip_options * opt, struct sk_buff * skb)
1da177e4 LT	250	{
	251	int l;
	252	unsigned char * iph;
	253	unsigned char * optptr;
	254	int optlen;
	255	unsigned char * pp_ptr = NULL;
22aba383	256	struct rtable *rt = NULL;
1da177e4	257
22aba383	258	if (skb != NULL) {
511c3f92	259	rt = skb_rtable(skb);
22aba383 DL	260	optptr = (unsigned char *)&(ip_hdr(skb)[1]);
22aba383 DL	261	} else
10fe7d85	262	optptr = opt->__data;
22aba383	263	iph = optptr - sizeof(struct iphdr);
1da177e4 LT	264
	265	for (l = opt->optlen; l > 0; ) {
	266	switch (*optptr) {
	267	case IPOPT_END:
	268	for (optptr++, l--; l>0; optptr++, l--) {
	269	if (*optptr != IPOPT_END) {
	270	*optptr = IPOPT_END;
	271	opt->is_changed = 1;
	272	}
	273	}
	274	goto eol;
	275	case IPOPT_NOOP:
	276	l--;
	277	optptr++;
	278	continue;
	279	}
	280	optlen = optptr[1];
	281	if (optlen<2 \|\| optlen>l) {
	282	pp_ptr = optptr;
	283	goto error;
	284	}
	285	switch (*optptr) {
	286	case IPOPT_SSRR:
	287	case IPOPT_LSRR:
	288	if (optlen < 3) {
	289	pp_ptr = optptr + 1;
	290	goto error;
	291	}
	292	if (optptr[2] < 4) {
	293	pp_ptr = optptr + 2;
	294	goto error;
	295	}
	296	/* NB: cf RFC-1812 5.2.4.1 */
	297	if (opt->srr) {
	298	pp_ptr = optptr;
	299	goto error;
	300	}
	301	if (!skb) {
	302	if (optptr[2] != 4 \|\| optlen < 7 \|\| ((optlen-3) & 3)) {
	303	pp_ptr = optptr + 1;
	304	goto error;
	305	}
	306	memcpy(&opt->faddr, &optptr[3], 4);
	307	if (optlen > 7)
	308	memmove(&optptr[3], &optptr[7], optlen-7);
	309	}
	310	opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
	311	opt->srr = optptr - iph;
	312	break;
	313	case IPOPT_RR:
	314	if (opt->rr) {
	315	pp_ptr = optptr;
	316	goto error;
	317	}
	318	if (optlen < 3) {
	319	pp_ptr = optptr + 1;
	320	goto error;
	321	}
	322	if (optptr[2] < 4) {
	323	pp_ptr = optptr + 2;
	324	goto error;
	325	}
	326	if (optptr[2] <= optlen) {
	327	if (optptr[2]+3 > optlen) {
328	pp_ptr = optptr + 2;
329	goto error;
330	}
c65353da	331	if (rt) {
1da177e4 LT	332	memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
	333	opt->is_changed = 1;
	334	}
	335	optptr[2] += 4;
	336	opt->rr_needaddr = 1;
	337	}
	338	opt->rr = optptr - iph;
	339	break;
	340	case IPOPT_TIMESTAMP:
	341	if (opt->ts) {
	342	pp_ptr = optptr;
	343	goto error;
	344	}
	345	if (optlen < 4) {
	346	pp_ptr = optptr + 1;
	347	goto error;
	348	}
	349	if (optptr[2] < 5) {
	350	pp_ptr = optptr + 2;
	351	goto error;
	352	}
	353	if (optptr[2] <= optlen) {
48bdf072	354	unsigned char *timeptr = NULL;
1da177e4 LT	355	if (optptr[2]+3 > optptr[1]) {
	356	pp_ptr = optptr + 2;
	357	goto error;
	358	}
	359	switch (optptr[3]&0xF) {
	360	case IPOPT_TS_TSONLY:
	361	opt->ts = optptr - iph;
e905a9ed	362	if (skb)
48bdf072	363	timeptr = &optptr[optptr[2]-1];
1da177e4 LT	364	opt->ts_needtime = 1;
	365	optptr[2] += 4;
	366	break;
	367	case IPOPT_TS_TSANDADDR:
	368	if (optptr[2]+7 > optptr[1]) {
	369	pp_ptr = optptr + 2;
	370	goto error;
	371	}
	372	opt->ts = optptr - iph;
c65353da	373	if (rt) {
1da177e4	374	memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
48bdf072	375	timeptr = &optptr[optptr[2]+3];
1da177e4 LT	376	}
	377	opt->ts_needaddr = 1;
	378	opt->ts_needtime = 1;
	379	optptr[2] += 8;
	380	break;
	381	case IPOPT_TS_PRESPEC:
	382	if (optptr[2]+7 > optptr[1]) {
	383	pp_ptr = optptr + 2;
	384	goto error;
	385	}
	386	opt->ts = optptr - iph;
	387	{
fd683222	388	__be32 addr;
1da177e4	389	memcpy(&addr, &optptr[optptr[2]-1], 4);
0e6bd4a1	390	if (inet_addr_type(net, addr) == RTN_UNICAST)
1da177e4 LT	391	break;
1da177e4 LT	392	if (skb)
48bdf072	393	timeptr = &optptr[optptr[2]+3];
1da177e4 LT	394	}
	395	opt->ts_needtime = 1;
	396	optptr[2] += 8;
	397	break;
	398	default:
	399	if (!skb && !capable(CAP_NET_RAW)) {
	400	pp_ptr = optptr + 3;
	401	goto error;
	402	}
	403	break;
	404	}
	405	if (timeptr) {
f25c3d61	406	struct timespec tv;
48bdf072	407	u32 midtime;
f25c3d61	408	getnstimeofday(&tv);
48bdf072 CM	409	midtime = (tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC;
48bdf072 CM	410	put_unaligned_be32(midtime, timeptr);
1da177e4 LT	411	opt->is_changed = 1;
	412	}
	413	} else {
	414	unsigned overflow = optptr[3]>>4;
	415	if (overflow == 15) {
	416	pp_ptr = optptr + 3;
	417	goto error;
	418	}
	419	opt->ts = optptr - iph;
	420	if (skb) {
	421	optptr[3] = (optptr[3]&0xF)\|((overflow+1)<<4);
	422	opt->is_changed = 1;
	423	}
	424	}
	425	break;
	426	case IPOPT_RA:
	427	if (optlen < 4) {
	428	pp_ptr = optptr + 1;
	429	goto error;
	430	}
	431	if (optptr[2] == 0 && optptr[3] == 0)
	432	opt->router_alert = optptr - iph;
	433	break;
11a03f78	434	case IPOPT_CIPSO:
f8687afe	435	if ((!skb && !capable(CAP_NET_RAW)) \|\| opt->cipso) {
11a03f78 PM	436	pp_ptr = optptr;
	437	goto error;
	438	}
	439	opt->cipso = optptr - iph;
15c45f7b	440	if (cipso_v4_validate(skb, &optptr)) {
11a03f78 PM	441	pp_ptr = optptr;
	442	goto error;
	443	}
	444	break;
1da177e4 LT	445	case IPOPT_SEC:
	446	case IPOPT_SID:
	447	default:
	448	if (!skb && !capable(CAP_NET_RAW)) {
	449	pp_ptr = optptr;
	450	goto error;
	451	}
	452	break;
	453	}
	454	l -= optlen;
	455	optptr += optlen;
	456	}
	457
	458	eol:
	459	if (!pp_ptr)
	460	return 0;
	461
	462	error:
	463	if (skb) {
	464	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
	465	}
	466	return -EINVAL;
	467	}
462fb2af	468	EXPORT_SYMBOL(ip_options_compile);
1da177e4 LT	469
	470	/*
	471	* Undo all the changes done by ip_options_compile().
	472	*/
	473
	474	void ip_options_undo(struct ip_options * opt)
	475	{
	476	if (opt->srr) {
	477	unsigned char * optptr = opt->__data+opt->srr-sizeof(struct iphdr);
	478	memmove(optptr+7, optptr+3, optptr[1]-7);
	479	memcpy(optptr+3, &opt->faddr, 4);
	480	}
	481	if (opt->rr_needaddr) {
	482	unsigned char * optptr = opt->__data+opt->rr-sizeof(struct iphdr);
	483	optptr[2] -= 4;
	484	memset(&optptr[optptr[2]-1], 0, 4);
	485	}
	486	if (opt->ts) {
	487	unsigned char * optptr = opt->__data+opt->ts-sizeof(struct iphdr);
	488	if (opt->ts_needtime) {
	489	optptr[2] -= 4;
	490	memset(&optptr[optptr[2]-1], 0, 4);
	491	if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
	492	optptr[2] -= 4;
	493	}
	494	if (opt->ts_needaddr) {
	495	optptr[2] -= 4;
	496	memset(&optptr[optptr[2]-1], 0, 4);
	497	}
	498	}
	499	}
	500
f6d8bd05	501	static struct ip_options_rcu *ip_options_get_alloc(const int optlen)
1da177e4	502	{
f6d8bd05	503	return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
37640703	504	GFP_KERNEL);
4c6ea29d	505	}
1da177e4	506
f6d8bd05 ED	507	static int ip_options_get_finish(struct net net, struct ip_options_rcu *optp,
f6d8bd05 ED	508	struct ip_options_rcu *opt, int optlen)
4c6ea29d	509	{
1da177e4	510	while (optlen & 3)
f6d8bd05 ED	511	opt->opt.__data[optlen++] = IPOPT_END;
	512	opt->opt.optlen = optlen;
	513	if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
1da177e4 LT	514	kfree(opt);
	515	return -EINVAL;
	516	}
a51482bd	517	kfree(*optp);
1da177e4 LT	518	*optp = opt;
	519	return 0;
	520	}
	521
f6d8bd05	522	int ip_options_get_from_user(struct net net, struct ip_options_rcu *optp,
f2c4802b	523	unsigned char __user *data, int optlen)
4c6ea29d	524	{
f6d8bd05	525	struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
4c6ea29d ACM	526
	527	if (!opt)
	528	return -ENOMEM;
f6d8bd05	529	if (optlen && copy_from_user(opt->opt.__data, data, optlen)) {
4c6ea29d ACM	530	kfree(opt);
	531	return -EFAULT;
	532	}
f2c4802b	533	return ip_options_get_finish(net, optp, opt, optlen);
4c6ea29d ACM	534	}
4c6ea29d ACM	535
f6d8bd05	536	int ip_options_get(struct net net, struct ip_options_rcu *optp,
f2c4802b	537	unsigned char *data, int optlen)
4c6ea29d	538	{
f6d8bd05	539	struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
4c6ea29d ACM	540
	541	if (!opt)
	542	return -ENOMEM;
	543	if (optlen)
f6d8bd05	544	memcpy(opt->opt.__data, data, optlen);
f2c4802b	545	return ip_options_get_finish(net, optp, opt, optlen);
4c6ea29d ACM	546	}
4c6ea29d ACM	547
1da177e4 LT	548	void ip_forward_options(struct sk_buff *skb)
	549	{
	550	struct ip_options * opt = &(IPCB(skb)->opt);
	551	unsigned char * optptr;
511c3f92	552	struct rtable *rt = skb_rtable(skb);
d56f90a7	553	unsigned char *raw = skb_network_header(skb);
1da177e4 LT	554
	555	if (opt->rr_needaddr) {
	556	optptr = (unsigned char *)raw + opt->rr;
8e36360a	557	ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
1da177e4 LT	558	opt->is_changed = 1;
	559	}
	560	if (opt->srr_is_hit) {
	561	int srrptr, srrspace;
	562
	563	optptr = raw + opt->srr;
	564
	565	for ( srrptr=optptr[2], srrspace = optptr[1];
	566	srrptr <= srrspace;
	567	srrptr += 4
	568	) {
	569	if (srrptr + 3 > srrspace)
	570	break;
ac8a4810	571	if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
1da177e4 LT	572	break;
	573	}
	574	if (srrptr + 3 <= srrspace) {
	575	opt->is_changed = 1;
ac8a4810	576	ip_hdr(skb)->daddr = opt->nexthop;
5dc7883f	577	ip_rt_get_source(&optptr[srrptr-1], skb, rt);
1da177e4 LT	578	optptr[2] = srrptr+4;
1da177e4 LT	579	} else if (net_ratelimit())
058bd4d2	580	pr_crit("%s(): Argh! Destination lost!\n", __func__);
1da177e4 LT	581	if (opt->ts_needaddr) {
1da177e4 LT	582	optptr = raw + opt->ts;
8e36360a	583	ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
1da177e4 LT	584	opt->is_changed = 1;
	585	}
	586	}
	587	if (opt->is_changed) {
	588	opt->is_changed = 0;
eddc9ec5	589	ip_send_check(ip_hdr(skb));
1da177e4 LT	590	}
	591	}
	592
	593	int ip_options_rcv_srr(struct sk_buff *skb)
	594	{
	595	struct ip_options *opt = &(IPCB(skb)->opt);
	596	int srrspace, srrptr;
9e12bb22	597	__be32 nexthop;
eddc9ec5	598	struct iphdr *iph = ip_hdr(skb);
d56f90a7	599	unsigned char *optptr = skb_network_header(skb) + opt->srr;
511c3f92	600	struct rtable *rt = skb_rtable(skb);
1da177e4	601	struct rtable *rt2;
7fee226a	602	unsigned long orefdst;
1da177e4 LT	603	int err;
1da177e4 LT	604
10949550	605	if (!rt)
1da177e4 LT	606	return 0;
	607
	608	if (skb->pkt_type != PACKET_HOST)
	609	return -EINVAL;
	610	if (rt->rt_type == RTN_UNICAST) {
	611	if (!opt->is_strictroute)
	612	return 0;
	613	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
	614	return -EINVAL;
	615	}
	616	if (rt->rt_type != RTN_LOCAL)
	617	return -EINVAL;
	618
	619	for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
	620	if (srrptr + 3 > srrspace) {
	621	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
	622	return -EINVAL;
	623	}
	624	memcpy(&nexthop, &optptr[srrptr-1], 4);
	625
7fee226a	626	orefdst = skb->_skb_refdst;
adf30907	627	skb_dst_set(skb, NULL);
1da177e4	628	err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
511c3f92	629	rt2 = skb_rtable(skb);
1da177e4	630	if (err \|\| (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
7fee226a ED	631	skb_dst_drop(skb);
7fee226a ED	632	skb->_skb_refdst = orefdst;
1da177e4 LT	633	return -EINVAL;
1da177e4 LT	634	}
7fee226a	635	refdst_drop(orefdst);
1da177e4 LT	636	if (rt2->rt_type != RTN_LOCAL)
	637	break;
	638	/* Superfast 8) loopback forward */
c30883bd	639	iph->daddr = nexthop;
1da177e4 LT	640	opt->is_changed = 1;
	641	}
	642	if (srrptr <= srrspace) {
	643	opt->srr_is_hit = 1;
ac8a4810	644	opt->nexthop = nexthop;
1da177e4 LT	645	opt->is_changed = 1;
	646	}
	647	return 0;
	648	}
462fb2af	649	EXPORT_SYMBOL(ip_options_rcv_srr);