[linux-2.6-block.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 <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>
#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) {
			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;

	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;

	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) {
			__be32 daddr = fib_compute_spec_dst(skb);

			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;
}

/* 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 != 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) {
					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 > optptr[1]) {
					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 > optptr[1]) {
						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 > optptr[1]) {
						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) {
					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 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 LT	17	#include <linux/types.h>
1da177e4 LT	18	#include <asm/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) {
f25c3d61	61	struct timespec tv;
e25d2ca6	62	__be32 midtime;
f25c3d61 YH	63	getnstimeofday(&tv);
f25c3d61 YH	64	midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
1da177e4 LT	65	memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
	66	}
	67	return;
	68	}
	69	if (opt->rr) {
	70	memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
	71	opt->rr = 0;
	72	opt->rr_needaddr = 0;
	73	}
	74	if (opt->ts) {
	75	memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
	76	opt->ts = 0;
	77	opt->ts_needaddr = opt->ts_needtime = 0;
	78	}
	79	}
	80
e905a9ed	81	/*
1da177e4 LT	82	* Provided (sopt, skb) points to received options,
	83	* build in dopt compiled option set appropriate for answering.
	84	* i.e. invert SRR option, copy anothers,
	85	* and grab room in RR/TS options.
	86	*
	87	* NOTE: dopt cannot point to skb.
	88	*/
	89
f6d8bd05	90	int ip_options_echo(struct ip_options dopt, struct sk_buff skb)
1da177e4	91	{
f6d8bd05	92	const struct ip_options *sopt;
1da177e4 LT	93	unsigned char sptr, dptr;
	94	int soffset, doffset;
	95	int optlen;
1da177e4 LT	96
	97	memset(dopt, 0, sizeof(struct ip_options));
	98
1da177e4 LT	99	sopt = &(IPCB(skb)->opt);
1da177e4 LT	100
f6d8bd05	101	if (sopt->optlen == 0)
1da177e4	102	return 0;
1da177e4	103
d56f90a7	104	sptr = skb_network_header(skb);
1da177e4 LT	105	dptr = dopt->__data;
1da177e4 LT	106
1da177e4 LT	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) {
6255e5ea JA	180	__be32 daddr = fib_compute_spec_dst(skb);
6255e5ea JA	181
1da177e4 LT	182	memcpy(&start[doffset-1], &daddr, 4);
	183	dopt->faddr = faddr;
	184	dptr[0] = start[0];
	185	dptr[1] = doffset+3;
	186	dptr[2] = 4;
	187	dptr += doffset+3;
	188	dopt->srr = dopt->optlen + sizeof(struct iphdr);
	189	dopt->optlen += doffset+3;
	190	dopt->is_strictroute = sopt->is_strictroute;
	191	}
	192	}
11a03f78 PM	193	if (sopt->cipso) {
	194	optlen = sptr[sopt->cipso+1];
	195	dopt->cipso = dopt->optlen+sizeof(struct iphdr);
	196	memcpy(dptr, sptr+sopt->cipso, optlen);
	197	dptr += optlen;
	198	dopt->optlen += optlen;
	199	}
1da177e4 LT	200	while (dopt->optlen & 3) {
	201	*dptr++ = IPOPT_END;
	202	dopt->optlen++;
	203	}
	204	return 0;
	205	}
	206
	207	/*
	208	* Options "fragmenting", just fill options not
	209	* allowed in fragments with NOOPs.
	210	* Simple and stupid 8), but the most efficient way.
	211	*/
	212
5e73ea1a	213	void ip_options_fragment(struct sk_buff *skb)
1da177e4	214	{
d56f90a7	215	unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
5e73ea1a	216	struct ip_options *opt = &(IPCB(skb)->opt);
1da177e4 LT	217	int l = opt->optlen;
	218	int optlen;
	219
	220	while (l > 0) {
	221	switch (*optptr) {
	222	case IPOPT_END:
	223	return;
	224	case IPOPT_NOOP:
	225	l--;
	226	optptr++;
	227	continue;
	228	}
	229	optlen = optptr[1];
	230	if (optlen<2 \|\| optlen>l)
	231	return;
	232	if (!IPOPT_COPIED(*optptr))
	233	memset(optptr, IPOPT_NOOP, optlen);
	234	l -= optlen;
	235	optptr += optlen;
	236	}
	237	opt->ts = 0;
	238	opt->rr = 0;
	239	opt->rr_needaddr = 0;
	240	opt->ts_needaddr = 0;
	241	opt->ts_needtime = 0;
1da177e4 LT	242	}
1da177e4 LT	243
bf5e53e3 ED	244	/* helper used by ip_options_compile() to call fib_compute_spec_dst()
	245	* at most one time.
	246	*/
	247	static void spec_dst_fill(__be32 spec_dst, struct sk_buff skb)
	248	{
	249	if (*spec_dst == htonl(INADDR_ANY))
	250	*spec_dst = fib_compute_spec_dst(skb);
	251	}
	252
1da177e4 LT	253	/*
	254	* Verify options and fill pointers in struct options.
	255	* Caller should clear *opt, and set opt->data.
	256	* If opt == NULL, then skb->data should point to IP header.
	257	*/
	258
0e6bd4a1	259	int ip_options_compile(struct net *net,
5e73ea1a	260	struct ip_options opt, struct sk_buff skb)
1da177e4	261	{
bf5e53e3	262	__be32 spec_dst = htonl(INADDR_ANY);
5e73ea1a	263	unsigned char *pp_ptr = NULL;
11604721	264	struct rtable *rt = NULL;
35ebf65e DM	265	unsigned char *optptr;
	266	unsigned char *iph;
	267	int optlen, l;
1da177e4	268
22aba383	269	if (skb != NULL) {
11604721	270	rt = skb_rtable(skb);
22aba383 DL	271	optptr = (unsigned char *)&(ip_hdr(skb)[1]);
22aba383 DL	272	} else
10fe7d85	273	optptr = opt->__data;
22aba383	274	iph = optptr - sizeof(struct iphdr);
1da177e4 LT	275
	276	for (l = opt->optlen; l > 0; ) {
	277	switch (*optptr) {
	278	case IPOPT_END:
	279	for (optptr++, l--; l>0; optptr++, l--) {
	280	if (*optptr != IPOPT_END) {
	281	*optptr = IPOPT_END;
	282	opt->is_changed = 1;
	283	}
	284	}
	285	goto eol;
	286	case IPOPT_NOOP:
	287	l--;
	288	optptr++;
	289	continue;
	290	}
	291	optlen = optptr[1];
	292	if (optlen<2 \|\| optlen>l) {
	293	pp_ptr = optptr;
	294	goto error;
	295	}
	296	switch (*optptr) {
	297	case IPOPT_SSRR:
	298	case IPOPT_LSRR:
	299	if (optlen < 3) {
	300	pp_ptr = optptr + 1;
	301	goto error;
	302	}
	303	if (optptr[2] < 4) {
	304	pp_ptr = optptr + 2;
	305	goto error;
	306	}
	307	/* NB: cf RFC-1812 5.2.4.1 */
	308	if (opt->srr) {
	309	pp_ptr = optptr;
	310	goto error;
	311	}
	312	if (!skb) {
	313	if (optptr[2] != 4 \|\| optlen < 7 \|\| ((optlen-3) & 3)) {
	314	pp_ptr = optptr + 1;
	315	goto error;
	316	}
	317	memcpy(&opt->faddr, &optptr[3], 4);
	318	if (optlen > 7)
	319	memmove(&optptr[3], &optptr[7], optlen-7);
	320	}
	321	opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
	322	opt->srr = optptr - iph;
	323	break;
	324	case IPOPT_RR:
	325	if (opt->rr) {
	326	pp_ptr = optptr;
	327	goto error;
	328	}
	329	if (optlen < 3) {
	330	pp_ptr = optptr + 1;
	331	goto error;
	332	}
	333	if (optptr[2] < 4) {
	334	pp_ptr = optptr + 2;
	335	goto error;
	336	}
	337	if (optptr[2] <= optlen) {
	338	if (optptr[2]+3 > optlen) {
339	pp_ptr = optptr + 2;
340	goto error;
341	}
11604721	342	if (rt) {
bf5e53e3	343	spec_dst_fill(&spec_dst, skb);
35ebf65e	344	memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
1da177e4 LT	345	opt->is_changed = 1;
	346	}
	347	optptr[2] += 4;
	348	opt->rr_needaddr = 1;
	349	}
	350	opt->rr = optptr - iph;
	351	break;
	352	case IPOPT_TIMESTAMP:
	353	if (opt->ts) {
	354	pp_ptr = optptr;
	355	goto error;
	356	}
	357	if (optlen < 4) {
	358	pp_ptr = optptr + 1;
	359	goto error;
	360	}
	361	if (optptr[2] < 5) {
	362	pp_ptr = optptr + 2;
	363	goto error;
	364	}
	365	if (optptr[2] <= optlen) {
48bdf072	366	unsigned char *timeptr = NULL;
1da177e4 LT	367	if (optptr[2]+3 > optptr[1]) {
	368	pp_ptr = optptr + 2;
	369	goto error;
	370	}
	371	switch (optptr[3]&0xF) {
	372	case IPOPT_TS_TSONLY:
e905a9ed	373	if (skb)
48bdf072	374	timeptr = &optptr[optptr[2]-1];
1da177e4 LT	375	opt->ts_needtime = 1;
	376	optptr[2] += 4;
	377	break;
	378	case IPOPT_TS_TSANDADDR:
	379	if (optptr[2]+7 > optptr[1]) {
	380	pp_ptr = optptr + 2;
	381	goto error;
	382	}
11604721	383	if (rt) {
bf5e53e3	384	spec_dst_fill(&spec_dst, skb);
35ebf65e	385	memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
48bdf072	386	timeptr = &optptr[optptr[2]+3];
1da177e4 LT	387	}
	388	opt->ts_needaddr = 1;
	389	opt->ts_needtime = 1;
	390	optptr[2] += 8;
	391	break;
	392	case IPOPT_TS_PRESPEC:
	393	if (optptr[2]+7 > optptr[1]) {
	394	pp_ptr = optptr + 2;
	395	goto error;
	396	}
1da177e4	397	{
fd683222	398	__be32 addr;
1da177e4	399	memcpy(&addr, &optptr[optptr[2]-1], 4);
0e6bd4a1	400	if (inet_addr_type(net, addr) == RTN_UNICAST)
1da177e4 LT	401	break;
1da177e4 LT	402	if (skb)
48bdf072	403	timeptr = &optptr[optptr[2]+3];
1da177e4 LT	404	}
	405	opt->ts_needtime = 1;
	406	optptr[2] += 8;
	407	break;
	408	default:
52e804c6	409	if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
1da177e4 LT	410	pp_ptr = optptr + 3;
	411	goto error;
	412	}
	413	break;
	414	}
	415	if (timeptr) {
f25c3d61	416	struct timespec tv;
48bdf072	417	u32 midtime;
f25c3d61	418	getnstimeofday(&tv);
48bdf072 CM	419	midtime = (tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC;
48bdf072 CM	420	put_unaligned_be32(midtime, timeptr);
1da177e4 LT	421	opt->is_changed = 1;
1da177e4 LT	422	}
fa2b04f4	423	} else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) {
95c96174	424	unsigned int overflow = optptr[3]>>4;
1da177e4 LT	425	if (overflow == 15) {
	426	pp_ptr = optptr + 3;
	427	goto error;
	428	}
1da177e4 LT	429	if (skb) {
	430	optptr[3] = (optptr[3]&0xF)\|((overflow+1)<<4);
	431	opt->is_changed = 1;
	432	}
	433	}
4660c7f4	434	opt->ts = optptr - iph;
1da177e4 LT	435	break;
	436	case IPOPT_RA:
	437	if (optlen < 4) {
	438	pp_ptr = optptr + 1;
	439	goto error;
	440	}
	441	if (optptr[2] == 0 && optptr[3] == 0)
	442	opt->router_alert = optptr - iph;
	443	break;
11a03f78	444	case IPOPT_CIPSO:
52e804c6	445	if ((!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) \|\| opt->cipso) {
11a03f78 PM	446	pp_ptr = optptr;
	447	goto error;
	448	}
	449	opt->cipso = optptr - iph;
15c45f7b	450	if (cipso_v4_validate(skb, &optptr)) {
11a03f78 PM	451	pp_ptr = optptr;
	452	goto error;
	453	}
	454	break;
1da177e4 LT	455	case IPOPT_SEC:
	456	case IPOPT_SID:
	457	default:
52e804c6	458	if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
1da177e4 LT	459	pp_ptr = optptr;
	460	goto error;
	461	}
	462	break;
	463	}
	464	l -= optlen;
	465	optptr += optlen;
	466	}
	467
	468	eol:
	469	if (!pp_ptr)
	470	return 0;
	471
	472	error:
	473	if (skb) {
	474	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
	475	}
	476	return -EINVAL;
	477	}
462fb2af	478	EXPORT_SYMBOL(ip_options_compile);
1da177e4 LT	479
	480	/*
	481	* Undo all the changes done by ip_options_compile().
	482	*/
	483
5e73ea1a	484	void ip_options_undo(struct ip_options *opt)
1da177e4 LT	485	{
1da177e4 LT	486	if (opt->srr) {
5e73ea1a	487	unsigned char *optptr = opt->__data+opt->srr-sizeof(struct iphdr);
1da177e4 LT	488	memmove(optptr+7, optptr+3, optptr[1]-7);
	489	memcpy(optptr+3, &opt->faddr, 4);
	490	}
	491	if (opt->rr_needaddr) {
5e73ea1a	492	unsigned char *optptr = opt->__data+opt->rr-sizeof(struct iphdr);
1da177e4 LT	493	optptr[2] -= 4;
	494	memset(&optptr[optptr[2]-1], 0, 4);
	495	}
	496	if (opt->ts) {
5e73ea1a	497	unsigned char *optptr = opt->__data+opt->ts-sizeof(struct iphdr);
1da177e4 LT	498	if (opt->ts_needtime) {
	499	optptr[2] -= 4;
	500	memset(&optptr[optptr[2]-1], 0, 4);
	501	if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
	502	optptr[2] -= 4;
	503	}
	504	if (opt->ts_needaddr) {
	505	optptr[2] -= 4;
	506	memset(&optptr[optptr[2]-1], 0, 4);
	507	}
	508	}
	509	}
	510
f6d8bd05	511	static struct ip_options_rcu *ip_options_get_alloc(const int optlen)
1da177e4	512	{
f6d8bd05	513	return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
37640703	514	GFP_KERNEL);
4c6ea29d	515	}
1da177e4	516
f6d8bd05 ED	517	static int ip_options_get_finish(struct net net, struct ip_options_rcu *optp,
f6d8bd05 ED	518	struct ip_options_rcu *opt, int optlen)
4c6ea29d	519	{
1da177e4	520	while (optlen & 3)
f6d8bd05 ED	521	opt->opt.__data[optlen++] = IPOPT_END;
	522	opt->opt.optlen = optlen;
	523	if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
1da177e4 LT	524	kfree(opt);
	525	return -EINVAL;
	526	}
a51482bd	527	kfree(*optp);
1da177e4 LT	528	*optp = opt;
	529	return 0;
	530	}
	531
f6d8bd05	532	int ip_options_get_from_user(struct net net, struct ip_options_rcu *optp,
f2c4802b	533	unsigned char __user *data, int optlen)
4c6ea29d	534	{
f6d8bd05	535	struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
4c6ea29d ACM	536
	537	if (!opt)
	538	return -ENOMEM;
f6d8bd05	539	if (optlen && copy_from_user(opt->opt.__data, data, optlen)) {
4c6ea29d ACM	540	kfree(opt);
	541	return -EFAULT;
	542	}
f2c4802b	543	return ip_options_get_finish(net, optp, opt, optlen);
4c6ea29d ACM	544	}
4c6ea29d ACM	545
f6d8bd05	546	int ip_options_get(struct net net, struct ip_options_rcu *optp,
f2c4802b	547	unsigned char *data, int optlen)
4c6ea29d	548	{
f6d8bd05	549	struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
4c6ea29d ACM	550
	551	if (!opt)
	552	return -ENOMEM;
	553	if (optlen)
f6d8bd05	554	memcpy(opt->opt.__data, data, optlen);
f2c4802b	555	return ip_options_get_finish(net, optp, opt, optlen);
4c6ea29d ACM	556	}
4c6ea29d ACM	557
1da177e4 LT	558	void ip_forward_options(struct sk_buff *skb)
1da177e4 LT	559	{
5e73ea1a DB	560	struct ip_options *opt = &(IPCB(skb)->opt);
5e73ea1a DB	561	unsigned char *optptr;
511c3f92	562	struct rtable *rt = skb_rtable(skb);
d56f90a7	563	unsigned char *raw = skb_network_header(skb);
1da177e4 LT	564
	565	if (opt->rr_needaddr) {
	566	optptr = (unsigned char *)raw + opt->rr;
8e36360a	567	ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
1da177e4 LT	568	opt->is_changed = 1;
	569	}
	570	if (opt->srr_is_hit) {
	571	int srrptr, srrspace;
	572
	573	optptr = raw + opt->srr;
	574
	575	for ( srrptr=optptr[2], srrspace = optptr[1];
	576	srrptr <= srrspace;
	577	srrptr += 4
	578	) {
	579	if (srrptr + 3 > srrspace)
	580	break;
ac8a4810	581	if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
1da177e4 LT	582	break;
	583	}
	584	if (srrptr + 3 <= srrspace) {
	585	opt->is_changed = 1;
ac8a4810	586	ip_hdr(skb)->daddr = opt->nexthop;
5dc7883f	587	ip_rt_get_source(&optptr[srrptr-1], skb, rt);
1da177e4	588	optptr[2] = srrptr+4;
e87cc472 JP	589	} else {
	590	net_crit_ratelimited("%s(): Argh! Destination lost!\n",
	591	__func__);
	592	}
1da177e4 LT	593	if (opt->ts_needaddr) {
1da177e4 LT	594	optptr = raw + opt->ts;
8e36360a	595	ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
1da177e4 LT	596	opt->is_changed = 1;
	597	}
	598	}
	599	if (opt->is_changed) {
	600	opt->is_changed = 0;
eddc9ec5	601	ip_send_check(ip_hdr(skb));
1da177e4 LT	602	}
	603	}
	604
	605	int ip_options_rcv_srr(struct sk_buff *skb)
	606	{
	607	struct ip_options *opt = &(IPCB(skb)->opt);
	608	int srrspace, srrptr;
9e12bb22	609	__be32 nexthop;
eddc9ec5	610	struct iphdr *iph = ip_hdr(skb);
d56f90a7	611	unsigned char *optptr = skb_network_header(skb) + opt->srr;
511c3f92	612	struct rtable *rt = skb_rtable(skb);
1da177e4	613	struct rtable *rt2;
7fee226a	614	unsigned long orefdst;
1da177e4 LT	615	int err;
1da177e4 LT	616
10949550	617	if (!rt)
1da177e4 LT	618	return 0;
	619
	620	if (skb->pkt_type != PACKET_HOST)
	621	return -EINVAL;
	622	if (rt->rt_type == RTN_UNICAST) {
	623	if (!opt->is_strictroute)
	624	return 0;
	625	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
	626	return -EINVAL;
	627	}
	628	if (rt->rt_type != RTN_LOCAL)
	629	return -EINVAL;
	630
	631	for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
	632	if (srrptr + 3 > srrspace) {
	633	icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
	634	return -EINVAL;
	635	}
	636	memcpy(&nexthop, &optptr[srrptr-1], 4);
	637
7fee226a	638	orefdst = skb->_skb_refdst;
adf30907	639	skb_dst_set(skb, NULL);
1da177e4	640	err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
511c3f92	641	rt2 = skb_rtable(skb);
1da177e4	642	if (err \|\| (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
7fee226a ED	643	skb_dst_drop(skb);
7fee226a ED	644	skb->_skb_refdst = orefdst;
1da177e4 LT	645	return -EINVAL;
1da177e4 LT	646	}
7fee226a	647	refdst_drop(orefdst);
1da177e4 LT	648	if (rt2->rt_type != RTN_LOCAL)
	649	break;
	650	/* Superfast 8) loopback forward */
c30883bd	651	iph->daddr = nexthop;
1da177e4 LT	652	opt->is_changed = 1;
	653	}
	654	if (srrptr <= srrspace) {
	655	opt->srr_is_hit = 1;
ac8a4810	656	opt->nexthop = nexthop;
1da177e4 LT	657	opt->is_changed = 1;
	658	}
	659	return 0;
	660	}
462fb2af	661	EXPORT_SYMBOL(ip_options_rcv_srr);