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.
6 * The options processing module for ip.c
8 * Authors: A.N.Kuznetsov
12 #define pr_fmt(fmt) "IPv4: " fmt
14 #include <linux/capability.h>
15 #include <linux/module.h>
16 #include <linux/slab.h>
17 #include <linux/types.h>
18 #include <asm/uaccess.h>
19 #include <asm/unaligned.h>
20 #include <linux/skbuff.h>
22 #include <linux/icmp.h>
23 #include <linux/netdevice.h>
24 #include <linux/rtnetlink.h>
28 #include <net/route.h>
29 #include <net/cipso_ipv4.h>
32 * Write options to IP header, record destination address to
33 * source route option, address of outgoing interface
34 * (we should already know it, so that this function is allowed be
35 * called only after routing decision) and timestamp,
36 * if we originate this datagram.
38 * daddr is real destination address, next hop is recorded in IP header.
39 * saddr is address of outgoing interface.
42 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
43 __be32 daddr, struct rtable *rt, int is_frag)
45 unsigned char *iph = skb_network_header(skb);
47 memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
48 memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
49 opt = &(IPCB(skb)->opt);
52 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
56 ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, skb, rt);
58 ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, skb, rt);
59 if (opt->ts_needtime) {
63 midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
64 memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
69 memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
74 memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
76 opt->ts_needaddr = opt->ts_needtime = 0;
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.
86 * NOTE: dopt cannot point to skb.
89 int ip_options_echo(struct ip_options *dopt, struct sk_buff *skb)
91 const struct ip_options *sopt;
92 unsigned char *sptr, *dptr;
97 memset(dopt, 0, sizeof(struct ip_options));
99 sopt = &(IPCB(skb)->opt);
101 if (sopt->optlen == 0)
104 sptr = skb_network_header(skb);
107 daddr = skb_rtable(skb)->rt_spec_dst;
110 optlen = sptr[sopt->rr+1];
111 soffset = sptr[sopt->rr+2];
112 dopt->rr = dopt->optlen + sizeof(struct iphdr);
113 memcpy(dptr, sptr+sopt->rr, optlen);
114 if (sopt->rr_needaddr && soffset <= optlen) {
115 if (soffset + 3 > optlen)
117 dptr[2] = soffset + 4;
118 dopt->rr_needaddr = 1;
121 dopt->optlen += optlen;
124 optlen = sptr[sopt->ts+1];
125 soffset = sptr[sopt->ts+2];
126 dopt->ts = dopt->optlen + sizeof(struct iphdr);
127 memcpy(dptr, sptr+sopt->ts, optlen);
128 if (soffset <= optlen) {
129 if (sopt->ts_needaddr) {
130 if (soffset + 3 > optlen)
132 dopt->ts_needaddr = 1;
135 if (sopt->ts_needtime) {
136 if (soffset + 3 > optlen)
138 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
139 dopt->ts_needtime = 1;
142 dopt->ts_needtime = 0;
144 if (soffset + 7 <= optlen) {
147 memcpy(&addr, dptr+soffset-1, 4);
148 if (inet_addr_type(dev_net(skb_dst(skb)->dev), addr) != RTN_UNICAST) {
149 dopt->ts_needtime = 1;
158 dopt->optlen += optlen;
161 unsigned char *start = sptr+sopt->srr;
167 if (soffset > optlen)
168 soffset = optlen + 1;
171 memcpy(&faddr, &start[soffset-1], 4);
172 for (soffset-=4, doffset=4; soffset > 3; soffset-=4, doffset+=4)
173 memcpy(&dptr[doffset-1], &start[soffset-1], 4);
175 * RFC1812 requires to fix illegal source routes.
177 if (memcmp(&ip_hdr(skb)->saddr,
178 &start[soffset + 3], 4) == 0)
182 memcpy(&start[doffset-1], &daddr, 4);
188 dopt->srr = dopt->optlen + sizeof(struct iphdr);
189 dopt->optlen += doffset+3;
190 dopt->is_strictroute = sopt->is_strictroute;
194 optlen = sptr[sopt->cipso+1];
195 dopt->cipso = dopt->optlen+sizeof(struct iphdr);
196 memcpy(dptr, sptr+sopt->cipso, optlen);
198 dopt->optlen += optlen;
200 while (dopt->optlen & 3) {
208 * Options "fragmenting", just fill options not
209 * allowed in fragments with NOOPs.
210 * Simple and stupid 8), but the most efficient way.
213 void ip_options_fragment(struct sk_buff *skb)
215 unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
216 struct ip_options *opt = &(IPCB(skb)->opt);
230 if (optlen<2 || optlen>l)
232 if (!IPOPT_COPIED(*optptr))
233 memset(optptr, IPOPT_NOOP, optlen);
239 opt->rr_needaddr = 0;
240 opt->ts_needaddr = 0;
241 opt->ts_needtime = 0;
245 * Verify options and fill pointers in struct options.
246 * Caller should clear *opt, and set opt->data.
247 * If opt == NULL, then skb->data should point to IP header.
250 int ip_options_compile(struct net *net,
251 struct ip_options *opt, struct sk_buff *skb)
255 unsigned char *optptr;
257 unsigned char *pp_ptr = NULL;
258 struct rtable *rt = NULL;
261 rt = skb_rtable(skb);
262 optptr = (unsigned char *)&(ip_hdr(skb)[1]);
264 optptr = opt->__data;
265 iph = optptr - sizeof(struct iphdr);
267 for (l = opt->optlen; l > 0; ) {
270 for (optptr++, l--; l>0; optptr++, l--) {
271 if (*optptr != IPOPT_END) {
283 if (optlen<2 || optlen>l) {
298 /* NB: cf RFC-1812 5.2.4.1 */
304 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
308 memcpy(&opt->faddr, &optptr[3], 4);
310 memmove(&optptr[3], &optptr[7], optlen-7);
312 opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
313 opt->srr = optptr - iph;
328 if (optptr[2] <= optlen) {
329 if (optptr[2]+3 > optlen) {
334 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
338 opt->rr_needaddr = 1;
340 opt->rr = optptr - iph;
342 case IPOPT_TIMESTAMP:
355 if (optptr[2] <= optlen) {
356 unsigned char *timeptr = NULL;
357 if (optptr[2]+3 > optptr[1]) {
361 switch (optptr[3]&0xF) {
362 case IPOPT_TS_TSONLY:
363 opt->ts = optptr - iph;
365 timeptr = &optptr[optptr[2]-1];
366 opt->ts_needtime = 1;
369 case IPOPT_TS_TSANDADDR:
370 if (optptr[2]+7 > optptr[1]) {
374 opt->ts = optptr - iph;
376 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
377 timeptr = &optptr[optptr[2]+3];
379 opt->ts_needaddr = 1;
380 opt->ts_needtime = 1;
383 case IPOPT_TS_PRESPEC:
384 if (optptr[2]+7 > optptr[1]) {
388 opt->ts = optptr - iph;
391 memcpy(&addr, &optptr[optptr[2]-1], 4);
392 if (inet_addr_type(net, addr) == RTN_UNICAST)
395 timeptr = &optptr[optptr[2]+3];
397 opt->ts_needtime = 1;
401 if (!skb && !capable(CAP_NET_RAW)) {
411 midtime = (tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC;
412 put_unaligned_be32(midtime, timeptr);
416 unsigned int overflow = optptr[3]>>4;
417 if (overflow == 15) {
421 opt->ts = optptr - iph;
423 optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
433 if (optptr[2] == 0 && optptr[3] == 0)
434 opt->router_alert = optptr - iph;
437 if ((!skb && !capable(CAP_NET_RAW)) || opt->cipso) {
441 opt->cipso = optptr - iph;
442 if (cipso_v4_validate(skb, &optptr)) {
450 if (!skb && !capable(CAP_NET_RAW)) {
466 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
470 EXPORT_SYMBOL(ip_options_compile);
473 * Undo all the changes done by ip_options_compile().
476 void ip_options_undo(struct ip_options *opt)
479 unsigned char *optptr = opt->__data+opt->srr-sizeof(struct iphdr);
480 memmove(optptr+7, optptr+3, optptr[1]-7);
481 memcpy(optptr+3, &opt->faddr, 4);
483 if (opt->rr_needaddr) {
484 unsigned char *optptr = opt->__data+opt->rr-sizeof(struct iphdr);
486 memset(&optptr[optptr[2]-1], 0, 4);
489 unsigned char *optptr = opt->__data+opt->ts-sizeof(struct iphdr);
490 if (opt->ts_needtime) {
492 memset(&optptr[optptr[2]-1], 0, 4);
493 if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
496 if (opt->ts_needaddr) {
498 memset(&optptr[optptr[2]-1], 0, 4);
503 static struct ip_options_rcu *ip_options_get_alloc(const int optlen)
505 return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
509 static int ip_options_get_finish(struct net *net, struct ip_options_rcu **optp,
510 struct ip_options_rcu *opt, int optlen)
513 opt->opt.__data[optlen++] = IPOPT_END;
514 opt->opt.optlen = optlen;
515 if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
524 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
525 unsigned char __user *data, int optlen)
527 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
531 if (optlen && copy_from_user(opt->opt.__data, data, optlen)) {
535 return ip_options_get_finish(net, optp, opt, optlen);
538 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
539 unsigned char *data, int optlen)
541 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
546 memcpy(opt->opt.__data, data, optlen);
547 return ip_options_get_finish(net, optp, opt, optlen);
550 void ip_forward_options(struct sk_buff *skb)
552 struct ip_options *opt = &(IPCB(skb)->opt);
553 unsigned char *optptr;
554 struct rtable *rt = skb_rtable(skb);
555 unsigned char *raw = skb_network_header(skb);
557 if (opt->rr_needaddr) {
558 optptr = (unsigned char *)raw + opt->rr;
559 ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
562 if (opt->srr_is_hit) {
563 int srrptr, srrspace;
565 optptr = raw + opt->srr;
567 for ( srrptr=optptr[2], srrspace = optptr[1];
571 if (srrptr + 3 > srrspace)
573 if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
576 if (srrptr + 3 <= srrspace) {
578 ip_hdr(skb)->daddr = opt->nexthop;
579 ip_rt_get_source(&optptr[srrptr-1], skb, rt);
580 optptr[2] = srrptr+4;
582 net_crit_ratelimited("%s(): Argh! Destination lost!\n",
585 if (opt->ts_needaddr) {
586 optptr = raw + opt->ts;
587 ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
591 if (opt->is_changed) {
593 ip_send_check(ip_hdr(skb));
597 int ip_options_rcv_srr(struct sk_buff *skb)
599 struct ip_options *opt = &(IPCB(skb)->opt);
600 int srrspace, srrptr;
602 struct iphdr *iph = ip_hdr(skb);
603 unsigned char *optptr = skb_network_header(skb) + opt->srr;
604 struct rtable *rt = skb_rtable(skb);
606 unsigned long orefdst;
612 if (skb->pkt_type != PACKET_HOST)
614 if (rt->rt_type == RTN_UNICAST) {
615 if (!opt->is_strictroute)
617 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
620 if (rt->rt_type != RTN_LOCAL)
623 for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
624 if (srrptr + 3 > srrspace) {
625 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
628 memcpy(&nexthop, &optptr[srrptr-1], 4);
630 orefdst = skb->_skb_refdst;
631 skb_dst_set(skb, NULL);
632 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
633 rt2 = skb_rtable(skb);
634 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
636 skb->_skb_refdst = orefdst;
639 refdst_drop(orefdst);
640 if (rt2->rt_type != RTN_LOCAL)
642 /* Superfast 8) loopback forward */
643 iph->daddr = nexthop;
646 if (srrptr <= srrspace) {
648 opt->nexthop = nexthop;
653 EXPORT_SYMBOL(ip_options_rcv_srr);