[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
 *
 */

#define pr_fmt(fmt) "IPv4: " fmt

#include <linux/capability.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/types.h>
#include <linux/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>
#include <net/ip_fib.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) {
			__be32 midtime;

			midtime = inet_current_timestamp();
			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 net *net, struct ip_options *dopt,
		      struct sk_buff *skb, const struct ip_options *sopt)
{
	unsigned char *sptr, *dptr;
	int soffset, doffset;
	int	optlen;

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

	if (sopt->optlen == 0)
		return 0;

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

	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(net, 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) {
			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;
}

/* helper used by ip_options_compile() to call fib_compute_spec_dst()
 * at most one time.
 */
static void spec_dst_fill(__be32 *spec_dst, struct sk_buff *skb)
{
	if (*spec_dst == htonl(INADDR_ANY))
		*spec_dst = fib_compute_spec_dst(skb);
}

/*
 * 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)
{
	__be32 spec_dst = htonl(INADDR_ANY);
	unsigned char *pp_ptr = NULL;
	struct rtable *rt = NULL;
	unsigned char *optptr;
	unsigned char *iph;
	int optlen, l;

	if (skb) {
		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;
		}
		if (unlikely(l < 2)) {
			pp_ptr = optptr;
			goto error;
		}
		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) {
					spec_dst_fill(&spec_dst, skb);
					memcpy(&optptr[optptr[2]-1], &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 > optlen) {
					pp_ptr = optptr + 2;
					goto error;
				}
				switch (optptr[3]&0xF) {
				case IPOPT_TS_TSONLY:
					if (skb)
						timeptr = &optptr[optptr[2]-1];
					opt->ts_needtime = 1;
					optptr[2] += 4;
					break;
				case IPOPT_TS_TSANDADDR:
					if (optptr[2]+7 > optlen) {
						pp_ptr = optptr + 2;
						goto error;
					}
					if (rt)  {
						spec_dst_fill(&spec_dst, skb);
						memcpy(&optptr[optptr[2]-1], &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 > optlen) {
						pp_ptr = optptr + 2;
						goto error;
					}
					{
						__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 && !ns_capable(net->user_ns, CAP_NET_RAW)) {
						pp_ptr = optptr + 3;
						goto error;
					}
					break;
				}
				if (timeptr) {
					__be32 midtime;

					midtime = inet_current_timestamp();
					memcpy(timeptr, &midtime, 4);
					opt->is_changed = 1;
				}
			} else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) {
				unsigned int overflow = optptr[3]>>4;
				if (overflow == 15) {
					pp_ptr = optptr + 3;
					goto error;
				}
				if (skb) {
					optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
					opt->is_changed = 1;
				}
			}
			opt->ts = optptr - iph;
			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 && !ns_capable(net->user_ns, 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 && !ns_capable(net->user_ns, 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 {
			net_crit_ratelimited("%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
afd46503 JP	12	#define pr_fmt(fmt) "IPv4: " fmt
afd46503 JP	13
4fc268d2	14	#include <linux/capability.h>
1da177e4	15	#include <linux/module.h>
5a0e3ad6	16	#include <linux/slab.h>
1da177e4	17	#include <linux/types.h>
7c0f6ba6	18	#include <linux/uaccess.h>
48bdf072	19	#include <asm/unaligned.h>
1da177e4 LT	20	#include <linux/skbuff.h>
	21	#include <linux/ip.h>
	22	#include <linux/icmp.h>
	23	#include <linux/netdevice.h>
	24	#include <linux/rtnetlink.h>
	25	#include <net/sock.h>
	26	#include <net/ip.h>
	27	#include <net/icmp.h>
14c85021	28	#include <net/route.h>
11a03f78	29	#include <net/cipso_ipv4.h>
35ebf65e	30	#include <net/ip_fib.h>
1da177e4	31
e905a9ed	32	/*
1da177e4 LT	33	* Write options to IP header, record destination address to
	34	* source route option, address of outgoing interface
	35	* (we should already know it, so that this function is allowed be
	36	* called only after routing decision) and timestamp,
	37	* if we originate this datagram.
	38	*
	39	* daddr is real destination address, next hop is recorded in IP header.
	40	* saddr is address of outgoing interface.
	41	*/
	42
f6d8bd05	43	void ip_options_build(struct sk_buff skb, struct ip_options opt,
8e36360a	44	__be32 daddr, struct rtable *rt, int is_frag)
1da177e4	45	{
d56f90a7	46	unsigned char *iph = skb_network_header(skb);
1da177e4 LT	47
	48	memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
	49	memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
	50	opt = &(IPCB(skb)->opt);
1da177e4 LT	51
	52	if (opt->srr)
	53	memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
	54
	55	if (!is_frag) {
	56	if (opt->rr_needaddr)
8e36360a	57	ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, skb, rt);
1da177e4	58	if (opt->ts_needaddr)
8e36360a	59	ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, skb, rt);
1da177e4	60	if (opt->ts_needtime) {
e25d2ca6	61	__be32 midtime;
822c8685 DD	62
822c8685 DD	63	midtime = inet_current_timestamp();
1da177e4 LT	64	memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
	65	}
	66	return;
	67	}
	68	if (opt->rr) {
	69	memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
	70	opt->rr = 0;
	71	opt->rr_needaddr = 0;
	72	}
	73	if (opt->ts) {
	74	memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
	75	opt->ts = 0;
	76	opt->ts_needaddr = opt->ts_needtime = 0;
	77	}
	78	}
	79
e905a9ed	80	/*
1da177e4 LT	81	* Provided (sopt, skb) points to received options,
	82	* build in dopt compiled option set appropriate for answering.
	83	* i.e. invert SRR option, copy anothers,
	84	* and grab room in RR/TS options.
	85	*
	86	* NOTE: dopt cannot point to skb.
	87	*/
	88
91ed1e66 PA	89	int __ip_options_echo(struct net net, struct ip_options dopt,
91ed1e66 PA	90	struct sk_buff skb, const struct ip_options sopt)
1da177e4	91	{
1da177e4 LT	92	unsigned char sptr, dptr;
	93	int soffset, doffset;
	94	int optlen;
1da177e4 LT	95
	96	memset(dopt, 0, sizeof(struct ip_options));
	97
f6d8bd05	98	if (sopt->optlen == 0)
1da177e4	99	return 0;
1da177e4	100
d56f90a7	101	sptr = skb_network_header(skb);
1da177e4 LT	102	dptr = dopt->__data;
1da177e4 LT	103
1da177e4 LT	104	if (sopt->rr) {
	105	optlen = sptr[sopt->rr+1];
	106	soffset = sptr[sopt->rr+2];
	107	dopt->rr = dopt->optlen + sizeof(struct iphdr);
	108	memcpy(dptr, sptr+sopt->rr, optlen);
	109	if (sopt->rr_needaddr && soffset <= optlen) {
	110	if (soffset + 3 > optlen)
	111	return -EINVAL;
	112	dptr[2] = soffset + 4;
	113	dopt->rr_needaddr = 1;
	114	}
	115	dptr += optlen;
	116	dopt->optlen += optlen;
	117	}
	118	if (sopt->ts) {
	119	optlen = sptr[sopt->ts+1];
	120	soffset = sptr[sopt->ts+2];
	121	dopt->ts = dopt->optlen + sizeof(struct iphdr);
	122	memcpy(dptr, sptr+sopt->ts, optlen);
	123	if (soffset <= optlen) {
	124	if (sopt->ts_needaddr) {
	125	if (soffset + 3 > optlen)
	126	return -EINVAL;
	127	dopt->ts_needaddr = 1;
	128	soffset += 4;
	129	}
	130	if (sopt->ts_needtime) {
	131	if (soffset + 3 > optlen)
	132	return -EINVAL;
	133	if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
	134	dopt->ts_needtime = 1;
	135	soffset += 4;
	136	} else {
	137	dopt->ts_needtime = 0;
	138
8628bd8a	139	if (soffset + 7 <= optlen) {
fd683222	140	__be32 addr;
1da177e4	141
8628bd8a	142	memcpy(&addr, dptr+soffset-1, 4);
91ed1e66	143	if (inet_addr_type(net, addr) != RTN_UNICAST) {
1da177e4 LT	144	dopt->ts_needtime = 1;
	145	soffset += 8;
	146	}
	147	}
	148	}
	149	}
	150	dptr[2] = soffset;
	151	}
	152	dptr += optlen;
	153	dopt->optlen += optlen;
	154	}
	155	if (sopt->srr) {
f6d8bd05	156	unsigned char *start = sptr+sopt->srr;
3ca3c68e	157	__be32 faddr;
1da177e4 LT	158
	159	optlen = start[1];
	160	soffset = start[2];
	161	doffset = 0;
	162	if (soffset > optlen)
	163	soffset = optlen + 1;
	164	soffset -= 4;
	165	if (soffset > 3) {
	166	memcpy(&faddr, &start[soffset-1], 4);
a22318e8	167	for (soffset -= 4, doffset = 4; soffset > 3; soffset -= 4, doffset += 4)
1da177e4 LT	168	memcpy(&dptr[doffset-1], &start[soffset-1], 4);
	169	/*
	170	* RFC1812 requires to fix illegal source routes.
	171	*/
eddc9ec5 ACM	172	if (memcmp(&ip_hdr(skb)->saddr,
eddc9ec5 ACM	173	&start[soffset + 3], 4) == 0)
1da177e4 LT	174	doffset -= 4;
	175	}
	176	if (doffset > 3) {
1da177e4 LT	177	dopt->faddr = faddr;
	178	dptr[0] = start[0];
	179	dptr[1] = doffset+3;
	180	dptr[2] = 4;
	181	dptr += doffset+3;
	182	dopt->srr = dopt->optlen + sizeof(struct iphdr);
	183	dopt->optlen += doffset+3;
	184	dopt->is_strictroute = sopt->is_strictroute;
	185	}
	186	}
11a03f78 PM	187	if (sopt->cipso) {
	188	optlen = sptr[sopt->cipso+1];
	189	dopt->cipso = dopt->optlen+sizeof(struct iphdr);
	190	memcpy(dptr, sptr+sopt->cipso, optlen);
	191	dptr += optlen;
	192	dopt->optlen += optlen;
	193	}
1da177e4 LT	194	while (dopt->optlen & 3) {
	195	*dptr++ = IPOPT_END;
	196	dopt->optlen++;
	197	}
	198	return 0;
	199	}
	200
	201	/*
	202	* Options "fragmenting", just fill options not
	203	* allowed in fragments with NOOPs.
	204	* Simple and stupid 8), but the most efficient way.
	205	*/
	206
5e73ea1a	207	void ip_options_fragment(struct sk_buff *skb)
1da177e4	208	{
d56f90a7	209	unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
5e73ea1a	210	struct ip_options *opt = &(IPCB(skb)->opt);
1da177e4 LT	211	int l = opt->optlen;
	212	int optlen;
	213
	214	while (l > 0) {
	215	switch (*optptr) {
	216	case IPOPT_END:
	217	return;
	218	case IPOPT_NOOP:
	219	l--;
	220	optptr++;
	221	continue;
	222	}
	223	optlen = optptr[1];
a22318e8	224	if (optlen < 2 \|\| optlen > l)
1da177e4 LT	225	return;
	226	if (!IPOPT_COPIED(*optptr))
	227	memset(optptr, IPOPT_NOOP, optlen);
	228	l -= optlen;
	229	optptr += optlen;
	230	}
	231	opt->ts = 0;
	232	opt->rr = 0;
	233	opt->rr_needaddr = 0;
	234	opt->ts_needaddr = 0;
	235	opt->ts_needtime = 0;
1da177e4 LT	236	}
1da177e4 LT	237
bf5e53e3 ED	238	/* helper used by ip_options_compile() to call fib_compute_spec_dst()
	239	* at most one time.
	240	*/
	241	static void spec_dst_fill(__be32 spec_dst, struct sk_buff skb)
	242	{
	243	if (*spec_dst == htonl(INADDR_ANY))
	244	*spec_dst = fib_compute_spec_dst(skb);
	245	}
	246
1da177e4 LT	247	/*
	248	* Verify options and fill pointers in struct options.
	249	* Caller should clear *opt, and set opt->data.
	250	* If opt == NULL, then skb->data should point to IP header.
	251	*/
	252
0e6bd4a1	253	int ip_options_compile(struct net *net,
5e73ea1a	254	struct ip_options opt, struct sk_buff skb)
1da177e4	255	{
bf5e53e3	256	__be32 spec_dst = htonl(INADDR_ANY);
5e73ea1a	257	unsigned char *pp_ptr = NULL;
11604721	258	struct rtable *rt = NULL;
35ebf65e DM	259	unsigned char *optptr;
	260	unsigned char *iph;
	261	int optlen, l;
1da177e4	262
00db4124	263	if (skb) {
11604721	264	rt = skb_rtable(skb);
22aba383 DL	265	optptr = (unsigned char *)&(ip_hdr(skb)[1]);
22aba383 DL	266	} else
10fe7d85	267	optptr = opt->__data;
22aba383	268	iph = optptr - sizeof(struct iphdr);
1da177e4 LT	269
	270	for (l = opt->optlen; l > 0; ) {
	271	switch (*optptr) {
dd9b4559	272	case IPOPT_END:
a22318e8	273	for (optptr++, l--; l > 0; optptr++, l--) {
1da177e4 LT	274	if (*optptr != IPOPT_END) {
	275	*optptr = IPOPT_END;
	276	opt->is_changed = 1;
	277	}
	278	}
	279	goto eol;
dd9b4559	280	case IPOPT_NOOP:
1da177e4 LT	281	l--;
	282	optptr++;
	283	continue;
	284	}
10ec9472 ED	285	if (unlikely(l < 2)) {
	286	pp_ptr = optptr;
	287	goto error;
	288	}
1da177e4	289	optlen = optptr[1];
a22318e8	290	if (optlen < 2 \|\| optlen > l) {
1da177e4 LT	291	pp_ptr = optptr;
	292	goto error;
	293	}
	294	switch (*optptr) {
dd9b4559 WC	295	case IPOPT_SSRR:
dd9b4559 WC	296	case IPOPT_LSRR:
1da177e4 LT	297	if (optlen < 3) {
	298	pp_ptr = optptr + 1;
	299	goto error;
	300	}
	301	if (optptr[2] < 4) {
	302	pp_ptr = optptr + 2;
	303	goto error;
	304	}
	305	/* NB: cf RFC-1812 5.2.4.1 */
	306	if (opt->srr) {
	307	pp_ptr = optptr;
	308	goto error;
	309	}
	310	if (!skb) {
	311	if (optptr[2] != 4 \|\| optlen < 7 \|\| ((optlen-3) & 3)) {
	312	pp_ptr = optptr + 1;
	313	goto error;
	314	}
	315	memcpy(&opt->faddr, &optptr[3], 4);
	316	if (optlen > 7)
	317	memmove(&optptr[3], &optptr[7], optlen-7);
	318	}
	319	opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
	320	opt->srr = optptr - iph;
	321	break;
dd9b4559	322	case IPOPT_RR:
1da177e4 LT	323	if (opt->rr) {
	324	pp_ptr = optptr;
	325	goto error;
	326	}
	327	if (optlen < 3) {
	328	pp_ptr = optptr + 1;
	329	goto error;
	330	}
	331	if (optptr[2] < 4) {
	332	pp_ptr = optptr + 2;
	333	goto error;
	334	}
	335	if (optptr[2] <= optlen) {
	336	if (optptr[2]+3 > optlen) {
	337	pp_ptr = optptr + 2;
	338	goto error;
	339	}
11604721	340	if (rt) {
bf5e53e3	341	spec_dst_fill(&spec_dst, skb);
35ebf65e	342	memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
1da177e4 LT	343	opt->is_changed = 1;
	344	}
	345	optptr[2] += 4;
	346	opt->rr_needaddr = 1;
	347	}
	348	opt->rr = optptr - iph;
	349	break;
dd9b4559	350	case IPOPT_TIMESTAMP:
1da177e4 LT	351	if (opt->ts) {
	352	pp_ptr = optptr;
	353	goto error;
	354	}
	355	if (optlen < 4) {
	356	pp_ptr = optptr + 1;
	357	goto error;
	358	}
	359	if (optptr[2] < 5) {
	360	pp_ptr = optptr + 2;
	361	goto error;
	362	}
	363	if (optptr[2] <= optlen) {
48bdf072	364	unsigned char *timeptr = NULL;
5a2b646f	365	if (optptr[2]+3 > optlen) {
1da177e4 LT	366	pp_ptr = optptr + 2;
	367	goto error;
	368	}
	369	switch (optptr[3]&0xF) {
dd9b4559	370	case IPOPT_TS_TSONLY:
e905a9ed	371	if (skb)
48bdf072	372	timeptr = &optptr[optptr[2]-1];
1da177e4 LT	373	opt->ts_needtime = 1;
	374	optptr[2] += 4;
	375	break;
dd9b4559	376	case IPOPT_TS_TSANDADDR:
5a2b646f	377	if (optptr[2]+7 > optlen) {
1da177e4 LT	378	pp_ptr = optptr + 2;
	379	goto error;
	380	}
11604721	381	if (rt) {
bf5e53e3	382	spec_dst_fill(&spec_dst, skb);
35ebf65e	383	memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
48bdf072	384	timeptr = &optptr[optptr[2]+3];
1da177e4 LT	385	}
	386	opt->ts_needaddr = 1;
	387	opt->ts_needtime = 1;
	388	optptr[2] += 8;
	389	break;
dd9b4559	390	case IPOPT_TS_PRESPEC:
5a2b646f	391	if (optptr[2]+7 > optlen) {
1da177e4 LT	392	pp_ptr = optptr + 2;
	393	goto error;
	394	}
1da177e4	395	{
fd683222	396	__be32 addr;
1da177e4	397	memcpy(&addr, &optptr[optptr[2]-1], 4);
0e6bd4a1	398	if (inet_addr_type(net, addr) == RTN_UNICAST)
1da177e4 LT	399	break;
1da177e4 LT	400	if (skb)
48bdf072	401	timeptr = &optptr[optptr[2]+3];
1da177e4 LT	402	}
	403	opt->ts_needtime = 1;
	404	optptr[2] += 8;
	405	break;
dd9b4559	406	default:
52e804c6	407	if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
1da177e4 LT	408	pp_ptr = optptr + 3;
	409	goto error;
	410	}
	411	break;
	412	}
	413	if (timeptr) {
822c8685 DD	414	__be32 midtime;
	415
	416	midtime = inet_current_timestamp();
	417	memcpy(timeptr, &midtime, 4);
1da177e4 LT	418	opt->is_changed = 1;
1da177e4 LT	419	}
fa2b04f4	420	} else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) {
95c96174	421	unsigned int overflow = optptr[3]>>4;
1da177e4 LT	422	if (overflow == 15) {
	423	pp_ptr = optptr + 3;
	424	goto error;
	425	}
1da177e4 LT	426	if (skb) {
	427	optptr[3] = (optptr[3]&0xF)\|((overflow+1)<<4);
	428	opt->is_changed = 1;
	429	}
	430	}
4660c7f4	431	opt->ts = optptr - iph;
1da177e4	432	break;
dd9b4559	433	case IPOPT_RA:
1da177e4 LT	434	if (optlen < 4) {
	435	pp_ptr = optptr + 1;
	436	goto error;
	437	}
	438	if (optptr[2] == 0 && optptr[3] == 0)
	439	opt->router_alert = optptr - iph;
	440	break;
dd9b4559	441	case IPOPT_CIPSO:
52e804c6	442	if ((!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) \|\| opt->cipso) {
11a03f78 PM	443	pp_ptr = optptr;
	444	goto error;
	445	}
	446	opt->cipso = optptr - iph;
15c45f7b	447	if (cipso_v4_validate(skb, &optptr)) {
11a03f78 PM	448	pp_ptr = optptr;
	449	goto error;
	450	}
	451	break;
dd9b4559 WC	452	case IPOPT_SEC:
	453	case IPOPT_SID:
	454	default:
52e804c6	455	if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
1da177e4 LT	456	pp_ptr = optptr;
	457	goto error;
	458	}
	459	break;
	460	}
	461	l -= optlen;
	462	optptr += optlen;
	463	}
	464
	465	eol:
	466	if (!pp_ptr)
	467	return 0;
	468
	469	error:
	470	if (skb) {
	471	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
	472	}
	473	return -EINVAL;
	474	}
462fb2af	475	EXPORT_SYMBOL(ip_options_compile);
1da177e4 LT	476
	477	/*
	478	* Undo all the changes done by ip_options_compile().
	479	*/
	480
5e73ea1a	481	void ip_options_undo(struct ip_options *opt)
1da177e4 LT	482	{
1da177e4 LT	483	if (opt->srr) {
5e73ea1a	484	unsigned char *optptr = opt->__data+opt->srr-sizeof(struct iphdr);
1da177e4 LT	485	memmove(optptr+7, optptr+3, optptr[1]-7);
	486	memcpy(optptr+3, &opt->faddr, 4);
	487	}
	488	if (opt->rr_needaddr) {
5e73ea1a	489	unsigned char *optptr = opt->__data+opt->rr-sizeof(struct iphdr);
1da177e4 LT	490	optptr[2] -= 4;
	491	memset(&optptr[optptr[2]-1], 0, 4);
	492	}
	493	if (opt->ts) {
5e73ea1a	494	unsigned char *optptr = opt->__data+opt->ts-sizeof(struct iphdr);
1da177e4 LT	495	if (opt->ts_needtime) {
	496	optptr[2] -= 4;
	497	memset(&optptr[optptr[2]-1], 0, 4);
	498	if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
	499	optptr[2] -= 4;
	500	}
	501	if (opt->ts_needaddr) {
	502	optptr[2] -= 4;
	503	memset(&optptr[optptr[2]-1], 0, 4);
	504	}
	505	}
	506	}
	507
f6d8bd05	508	static struct ip_options_rcu *ip_options_get_alloc(const int optlen)
1da177e4	509	{
f6d8bd05	510	return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
37640703	511	GFP_KERNEL);
4c6ea29d	512	}
1da177e4	513
f6d8bd05 ED	514	static int ip_options_get_finish(struct net net, struct ip_options_rcu *optp,
f6d8bd05 ED	515	struct ip_options_rcu *opt, int optlen)
4c6ea29d	516	{
1da177e4	517	while (optlen & 3)
f6d8bd05 ED	518	opt->opt.__data[optlen++] = IPOPT_END;
	519	opt->opt.optlen = optlen;
	520	if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
1da177e4 LT	521	kfree(opt);
	522	return -EINVAL;
	523	}
a51482bd	524	kfree(*optp);
1da177e4 LT	525	*optp = opt;
	526	return 0;
	527	}
	528
f6d8bd05	529	int ip_options_get_from_user(struct net net, struct ip_options_rcu *optp,
f2c4802b	530	unsigned char __user *data, int optlen)
4c6ea29d	531	{
f6d8bd05	532	struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
4c6ea29d ACM	533
	534	if (!opt)
	535	return -ENOMEM;
f6d8bd05	536	if (optlen && copy_from_user(opt->opt.__data, data, optlen)) {
4c6ea29d ACM	537	kfree(opt);
	538	return -EFAULT;
	539	}
f2c4802b	540	return ip_options_get_finish(net, optp, opt, optlen);
4c6ea29d ACM	541	}
4c6ea29d ACM	542
f6d8bd05	543	int ip_options_get(struct net net, struct ip_options_rcu *optp,
f2c4802b	544	unsigned char *data, int optlen)
4c6ea29d	545	{
f6d8bd05	546	struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
4c6ea29d ACM	547
	548	if (!opt)
	549	return -ENOMEM;
	550	if (optlen)
f6d8bd05	551	memcpy(opt->opt.__data, data, optlen);
f2c4802b	552	return ip_options_get_finish(net, optp, opt, optlen);
4c6ea29d ACM	553	}
4c6ea29d ACM	554
1da177e4 LT	555	void ip_forward_options(struct sk_buff *skb)
1da177e4 LT	556	{
5e73ea1a DB	557	struct ip_options *opt = &(IPCB(skb)->opt);
5e73ea1a DB	558	unsigned char *optptr;
511c3f92	559	struct rtable *rt = skb_rtable(skb);
d56f90a7	560	unsigned char *raw = skb_network_header(skb);
1da177e4 LT	561
	562	if (opt->rr_needaddr) {
	563	optptr = (unsigned char *)raw + opt->rr;
8e36360a	564	ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
1da177e4 LT	565	opt->is_changed = 1;
	566	}
	567	if (opt->srr_is_hit) {
	568	int srrptr, srrspace;
	569
	570	optptr = raw + opt->srr;
	571
a22318e8	572	for ( srrptr = optptr[2], srrspace = optptr[1];
1da177e4 LT	573	srrptr <= srrspace;
	574	srrptr += 4
	575	) {
	576	if (srrptr + 3 > srrspace)
	577	break;
ac8a4810	578	if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
1da177e4 LT	579	break;
	580	}
	581	if (srrptr + 3 <= srrspace) {
	582	opt->is_changed = 1;
ac8a4810	583	ip_hdr(skb)->daddr = opt->nexthop;
5dc7883f	584	ip_rt_get_source(&optptr[srrptr-1], skb, rt);
1da177e4	585	optptr[2] = srrptr+4;
e87cc472 JP	586	} else {
	587	net_crit_ratelimited("%s(): Argh! Destination lost!\n",
	588	__func__);
	589	}
1da177e4 LT	590	if (opt->ts_needaddr) {
1da177e4 LT	591	optptr = raw + opt->ts;
8e36360a	592	ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
1da177e4 LT	593	opt->is_changed = 1;
	594	}
	595	}
	596	if (opt->is_changed) {
	597	opt->is_changed = 0;
eddc9ec5	598	ip_send_check(ip_hdr(skb));
1da177e4 LT	599	}
	600	}
	601
	602	int ip_options_rcv_srr(struct sk_buff *skb)
	603	{
	604	struct ip_options *opt = &(IPCB(skb)->opt);
	605	int srrspace, srrptr;
9e12bb22	606	__be32 nexthop;
eddc9ec5	607	struct iphdr *iph = ip_hdr(skb);
d56f90a7	608	unsigned char *optptr = skb_network_header(skb) + opt->srr;
511c3f92	609	struct rtable *rt = skb_rtable(skb);
1da177e4	610	struct rtable *rt2;
7fee226a	611	unsigned long orefdst;
1da177e4 LT	612	int err;
1da177e4 LT	613
10949550	614	if (!rt)
1da177e4 LT	615	return 0;
	616
	617	if (skb->pkt_type != PACKET_HOST)
	618	return -EINVAL;
	619	if (rt->rt_type == RTN_UNICAST) {
	620	if (!opt->is_strictroute)
	621	return 0;
	622	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
	623	return -EINVAL;
	624	}
	625	if (rt->rt_type != RTN_LOCAL)
	626	return -EINVAL;
	627
a22318e8	628	for (srrptr = optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
1da177e4 LT	629	if (srrptr + 3 > srrspace) {
	630	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
	631	return -EINVAL;
	632	}
	633	memcpy(&nexthop, &optptr[srrptr-1], 4);
	634
7fee226a	635	orefdst = skb->_skb_refdst;
adf30907	636	skb_dst_set(skb, NULL);
1da177e4	637	err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
511c3f92	638	rt2 = skb_rtable(skb);
1da177e4	639	if (err \|\| (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
7fee226a ED	640	skb_dst_drop(skb);
7fee226a ED	641	skb->_skb_refdst = orefdst;
1da177e4 LT	642	return -EINVAL;
1da177e4 LT	643	}
7fee226a	644	refdst_drop(orefdst);
1da177e4 LT	645	if (rt2->rt_type != RTN_LOCAL)
	646	break;
	647	/* Superfast 8) loopback forward */
c30883bd	648	iph->daddr = nexthop;
1da177e4 LT	649	opt->is_changed = 1;
	650	}
	651	if (srrptr <= srrspace) {
	652	opt->srr_is_hit = 1;
ac8a4810	653	opt->nexthop = nexthop;
1da177e4 LT	654	opt->is_changed = 1;
	655	}
	656	return 0;
	657	}
462fb2af	658	EXPORT_SYMBOL(ip_options_rcv_srr);