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>
30 #include <net/ip_fib.h>
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.
39 * daddr is real destination address, next hop is recorded in IP header.
40 * saddr is address of outgoing interface.
43 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
44 __be32 daddr, struct rtable *rt, int is_frag)
46 unsigned char *iph = skb_network_header(skb);
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);
53 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
57 ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, skb, rt);
59 ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, skb, rt);
60 if (opt->ts_needtime) {
64 midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
65 memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
70 memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
75 memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
77 opt->ts_needaddr = opt->ts_needtime = 0;
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.
87 * NOTE: dopt cannot point to skb.
90 int __ip_options_echo(struct ip_options *dopt, struct sk_buff *skb,
91 const struct ip_options *sopt)
93 unsigned char *sptr, *dptr;
97 memset(dopt, 0, sizeof(struct ip_options));
99 if (sopt->optlen == 0)
102 sptr = skb_network_header(skb);
106 optlen = sptr[sopt->rr+1];
107 soffset = sptr[sopt->rr+2];
108 dopt->rr = dopt->optlen + sizeof(struct iphdr);
109 memcpy(dptr, sptr+sopt->rr, optlen);
110 if (sopt->rr_needaddr && soffset <= optlen) {
111 if (soffset + 3 > optlen)
113 dptr[2] = soffset + 4;
114 dopt->rr_needaddr = 1;
117 dopt->optlen += optlen;
120 optlen = sptr[sopt->ts+1];
121 soffset = sptr[sopt->ts+2];
122 dopt->ts = dopt->optlen + sizeof(struct iphdr);
123 memcpy(dptr, sptr+sopt->ts, optlen);
124 if (soffset <= optlen) {
125 if (sopt->ts_needaddr) {
126 if (soffset + 3 > optlen)
128 dopt->ts_needaddr = 1;
131 if (sopt->ts_needtime) {
132 if (soffset + 3 > optlen)
134 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
135 dopt->ts_needtime = 1;
138 dopt->ts_needtime = 0;
140 if (soffset + 7 <= optlen) {
143 memcpy(&addr, dptr+soffset-1, 4);
144 if (inet_addr_type(dev_net(skb_dst(skb)->dev), addr) != RTN_UNICAST) {
145 dopt->ts_needtime = 1;
154 dopt->optlen += optlen;
157 unsigned char *start = sptr+sopt->srr;
163 if (soffset > optlen)
164 soffset = optlen + 1;
167 memcpy(&faddr, &start[soffset-1], 4);
168 for (soffset -= 4, doffset = 4; soffset > 3; soffset -= 4, doffset += 4)
169 memcpy(&dptr[doffset-1], &start[soffset-1], 4);
171 * RFC1812 requires to fix illegal source routes.
173 if (memcmp(&ip_hdr(skb)->saddr,
174 &start[soffset + 3], 4) == 0)
178 __be32 daddr = fib_compute_spec_dst(skb);
180 memcpy(&start[doffset-1], &daddr, 4);
186 dopt->srr = dopt->optlen + sizeof(struct iphdr);
187 dopt->optlen += doffset+3;
188 dopt->is_strictroute = sopt->is_strictroute;
192 optlen = sptr[sopt->cipso+1];
193 dopt->cipso = dopt->optlen+sizeof(struct iphdr);
194 memcpy(dptr, sptr+sopt->cipso, optlen);
196 dopt->optlen += optlen;
198 while (dopt->optlen & 3) {
206 * Options "fragmenting", just fill options not
207 * allowed in fragments with NOOPs.
208 * Simple and stupid 8), but the most efficient way.
211 void ip_options_fragment(struct sk_buff *skb)
213 unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
214 struct ip_options *opt = &(IPCB(skb)->opt);
228 if (optlen < 2 || optlen > l)
230 if (!IPOPT_COPIED(*optptr))
231 memset(optptr, IPOPT_NOOP, optlen);
237 opt->rr_needaddr = 0;
238 opt->ts_needaddr = 0;
239 opt->ts_needtime = 0;
242 /* helper used by ip_options_compile() to call fib_compute_spec_dst()
245 static void spec_dst_fill(__be32 *spec_dst, struct sk_buff *skb)
247 if (*spec_dst == htonl(INADDR_ANY))
248 *spec_dst = fib_compute_spec_dst(skb);
252 * Verify options and fill pointers in struct options.
253 * Caller should clear *opt, and set opt->data.
254 * If opt == NULL, then skb->data should point to IP header.
257 int __ip_options_compile(struct net *net,
258 struct ip_options *opt, struct sk_buff *skb,
261 __be32 spec_dst = htonl(INADDR_ANY);
262 unsigned char *pp_ptr = NULL;
263 struct rtable *rt = NULL;
264 unsigned char *optptr;
269 rt = skb_rtable(skb);
270 optptr = (unsigned char *)&(ip_hdr(skb)[1]);
272 optptr = opt->__data;
273 iph = optptr - sizeof(struct iphdr);
275 for (l = opt->optlen; l > 0; ) {
278 for (optptr++, l--; l > 0; optptr++, l--) {
279 if (*optptr != IPOPT_END) {
290 if (unlikely(l < 2)) {
295 if (optlen < 2 || optlen > l) {
310 /* NB: cf RFC-1812 5.2.4.1 */
316 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
320 memcpy(&opt->faddr, &optptr[3], 4);
322 memmove(&optptr[3], &optptr[7], optlen-7);
324 opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
325 opt->srr = optptr - iph;
340 if (optptr[2] <= optlen) {
341 if (optptr[2]+3 > optlen) {
346 spec_dst_fill(&spec_dst, skb);
347 memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
351 opt->rr_needaddr = 1;
353 opt->rr = optptr - iph;
355 case IPOPT_TIMESTAMP:
368 if (optptr[2] <= optlen) {
369 unsigned char *timeptr = NULL;
370 if (optptr[2]+3 > optlen) {
374 switch (optptr[3]&0xF) {
375 case IPOPT_TS_TSONLY:
377 timeptr = &optptr[optptr[2]-1];
378 opt->ts_needtime = 1;
381 case IPOPT_TS_TSANDADDR:
382 if (optptr[2]+7 > optlen) {
387 spec_dst_fill(&spec_dst, skb);
388 memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
389 timeptr = &optptr[optptr[2]+3];
391 opt->ts_needaddr = 1;
392 opt->ts_needtime = 1;
395 case IPOPT_TS_PRESPEC:
396 if (optptr[2]+7 > optlen) {
402 memcpy(&addr, &optptr[optptr[2]-1], 4);
403 if (inet_addr_type(net, addr) == RTN_UNICAST)
406 timeptr = &optptr[optptr[2]+3];
408 opt->ts_needtime = 1;
412 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
422 midtime = (tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC;
423 put_unaligned_be32(midtime, timeptr);
426 } else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) {
427 unsigned int overflow = optptr[3]>>4;
428 if (overflow == 15) {
433 optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
437 opt->ts = optptr - iph;
444 if (optptr[2] == 0 && optptr[3] == 0)
445 opt->router_alert = optptr - iph;
448 if ((!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) || opt->cipso) {
452 opt->cipso = optptr - iph;
453 if (cipso_v4_validate(skb, &optptr)) {
461 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
477 *info = htonl((pp_ptr-iph)<<24);
481 int ip_options_compile(struct net *net,
482 struct ip_options *opt, struct sk_buff *skb)
487 ret = __ip_options_compile(net, opt, skb, &info);
489 icmp_send(skb, ICMP_PARAMETERPROB, 0, info);
492 EXPORT_SYMBOL(ip_options_compile);
495 * Undo all the changes done by ip_options_compile().
498 void ip_options_undo(struct ip_options *opt)
501 unsigned char *optptr = opt->__data+opt->srr-sizeof(struct iphdr);
502 memmove(optptr+7, optptr+3, optptr[1]-7);
503 memcpy(optptr+3, &opt->faddr, 4);
505 if (opt->rr_needaddr) {
506 unsigned char *optptr = opt->__data+opt->rr-sizeof(struct iphdr);
508 memset(&optptr[optptr[2]-1], 0, 4);
511 unsigned char *optptr = opt->__data+opt->ts-sizeof(struct iphdr);
512 if (opt->ts_needtime) {
514 memset(&optptr[optptr[2]-1], 0, 4);
515 if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
518 if (opt->ts_needaddr) {
520 memset(&optptr[optptr[2]-1], 0, 4);
525 static struct ip_options_rcu *ip_options_get_alloc(const int optlen)
527 return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
531 static int ip_options_get_finish(struct net *net, struct ip_options_rcu **optp,
532 struct ip_options_rcu *opt, int optlen)
535 opt->opt.__data[optlen++] = IPOPT_END;
536 opt->opt.optlen = optlen;
537 if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
546 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
547 unsigned char __user *data, int optlen)
549 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
553 if (optlen && copy_from_user(opt->opt.__data, data, optlen)) {
557 return ip_options_get_finish(net, optp, opt, optlen);
560 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
561 unsigned char *data, int optlen)
563 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
568 memcpy(opt->opt.__data, data, optlen);
569 return ip_options_get_finish(net, optp, opt, optlen);
572 void ip_forward_options(struct sk_buff *skb)
574 struct ip_options *opt = &(IPCB(skb)->opt);
575 unsigned char *optptr;
576 struct rtable *rt = skb_rtable(skb);
577 unsigned char *raw = skb_network_header(skb);
579 if (opt->rr_needaddr) {
580 optptr = (unsigned char *)raw + opt->rr;
581 ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
584 if (opt->srr_is_hit) {
585 int srrptr, srrspace;
587 optptr = raw + opt->srr;
589 for ( srrptr = optptr[2], srrspace = optptr[1];
593 if (srrptr + 3 > srrspace)
595 if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
598 if (srrptr + 3 <= srrspace) {
600 ip_hdr(skb)->daddr = opt->nexthop;
601 ip_rt_get_source(&optptr[srrptr-1], skb, rt);
602 optptr[2] = srrptr+4;
604 net_crit_ratelimited("%s(): Argh! Destination lost!\n",
607 if (opt->ts_needaddr) {
608 optptr = raw + opt->ts;
609 ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
613 if (opt->is_changed) {
615 ip_send_check(ip_hdr(skb));
619 int ip_options_rcv_srr(struct sk_buff *skb)
621 struct ip_options *opt = &(IPCB(skb)->opt);
622 int srrspace, srrptr;
624 struct iphdr *iph = ip_hdr(skb);
625 unsigned char *optptr = skb_network_header(skb) + opt->srr;
626 struct rtable *rt = skb_rtable(skb);
628 unsigned long orefdst;
634 if (skb->pkt_type != PACKET_HOST)
636 if (rt->rt_type == RTN_UNICAST) {
637 if (!opt->is_strictroute)
639 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
642 if (rt->rt_type != RTN_LOCAL)
645 for (srrptr = optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
646 if (srrptr + 3 > srrspace) {
647 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
650 memcpy(&nexthop, &optptr[srrptr-1], 4);
652 orefdst = skb->_skb_refdst;
653 skb_dst_set(skb, NULL);
654 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
655 rt2 = skb_rtable(skb);
656 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
658 skb->_skb_refdst = orefdst;
661 refdst_drop(orefdst);
662 if (rt2->rt_type != RTN_LOCAL)
664 /* Superfast 8) loopback forward */
665 iph->daddr = nexthop;
668 if (srrptr <= srrspace) {
670 opt->nexthop = nexthop;
675 EXPORT_SYMBOL(ip_options_rcv_srr);