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)
92 const struct ip_options *sopt;
93 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);
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)
115 dptr[2] = soffset + 4;
116 dopt->rr_needaddr = 1;
119 dopt->optlen += optlen;
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)
130 dopt->ts_needaddr = 1;
133 if (sopt->ts_needtime) {
134 if (soffset + 3 > optlen)
136 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
137 dopt->ts_needtime = 1;
140 dopt->ts_needtime = 0;
142 if (soffset + 7 <= optlen) {
145 memcpy(&addr, dptr+soffset-1, 4);
146 if (inet_addr_type(dev_net(skb_dst(skb)->dev), addr) != RTN_UNICAST) {
147 dopt->ts_needtime = 1;
156 dopt->optlen += optlen;
159 unsigned char *start = sptr+sopt->srr;
165 if (soffset > optlen)
166 soffset = optlen + 1;
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);
173 * RFC1812 requires to fix illegal source routes.
175 if (memcmp(&ip_hdr(skb)->saddr,
176 &start[soffset + 3], 4) == 0)
180 __be32 daddr = fib_compute_spec_dst(skb);
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;
244 /* helper used by ip_options_compile() to call fib_compute_spec_dst()
247 static void spec_dst_fill(__be32 *spec_dst, struct sk_buff *skb)
249 if (*spec_dst == htonl(INADDR_ANY))
250 *spec_dst = fib_compute_spec_dst(skb);
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.
259 int ip_options_compile(struct net *net,
260 struct ip_options *opt, struct sk_buff *skb)
262 __be32 spec_dst = htonl(INADDR_ANY);
263 unsigned char *pp_ptr = NULL;
264 struct rtable *rt = NULL;
265 unsigned char *optptr;
270 rt = skb_rtable(skb);
271 optptr = (unsigned char *)&(ip_hdr(skb)[1]);
273 optptr = opt->__data;
274 iph = optptr - sizeof(struct iphdr);
276 for (l = opt->optlen; l > 0; ) {
279 for (optptr++, l--; l>0; optptr++, l--) {
280 if (*optptr != IPOPT_END) {
292 if (optlen<2 || optlen>l) {
307 /* NB: cf RFC-1812 5.2.4.1 */
313 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
317 memcpy(&opt->faddr, &optptr[3], 4);
319 memmove(&optptr[3], &optptr[7], optlen-7);
321 opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
322 opt->srr = optptr - iph;
337 if (optptr[2] <= optlen) {
338 if (optptr[2]+3 > optlen) {
343 spec_dst_fill(&spec_dst, skb);
344 memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
348 opt->rr_needaddr = 1;
350 opt->rr = optptr - iph;
352 case IPOPT_TIMESTAMP:
365 if (optptr[2] <= optlen) {
366 unsigned char *timeptr = NULL;
367 if (optptr[2]+3 > optptr[1]) {
371 switch (optptr[3]&0xF) {
372 case IPOPT_TS_TSONLY:
374 timeptr = &optptr[optptr[2]-1];
375 opt->ts_needtime = 1;
378 case IPOPT_TS_TSANDADDR:
379 if (optptr[2]+7 > optptr[1]) {
384 spec_dst_fill(&spec_dst, skb);
385 memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
386 timeptr = &optptr[optptr[2]+3];
388 opt->ts_needaddr = 1;
389 opt->ts_needtime = 1;
392 case IPOPT_TS_PRESPEC:
393 if (optptr[2]+7 > optptr[1]) {
399 memcpy(&addr, &optptr[optptr[2]-1], 4);
400 if (inet_addr_type(net, addr) == RTN_UNICAST)
403 timeptr = &optptr[optptr[2]+3];
405 opt->ts_needtime = 1;
409 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
419 midtime = (tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC;
420 put_unaligned_be32(midtime, timeptr);
423 } else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) {
424 unsigned int overflow = optptr[3]>>4;
425 if (overflow == 15) {
430 optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
434 opt->ts = optptr - iph;
441 if (optptr[2] == 0 && optptr[3] == 0)
442 opt->router_alert = optptr - iph;
445 if ((!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) || opt->cipso) {
449 opt->cipso = optptr - iph;
450 if (cipso_v4_validate(skb, &optptr)) {
458 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
474 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
478 EXPORT_SYMBOL(ip_options_compile);
481 * Undo all the changes done by ip_options_compile().
484 void ip_options_undo(struct ip_options *opt)
487 unsigned char *optptr = opt->__data+opt->srr-sizeof(struct iphdr);
488 memmove(optptr+7, optptr+3, optptr[1]-7);
489 memcpy(optptr+3, &opt->faddr, 4);
491 if (opt->rr_needaddr) {
492 unsigned char *optptr = opt->__data+opt->rr-sizeof(struct iphdr);
494 memset(&optptr[optptr[2]-1], 0, 4);
497 unsigned char *optptr = opt->__data+opt->ts-sizeof(struct iphdr);
498 if (opt->ts_needtime) {
500 memset(&optptr[optptr[2]-1], 0, 4);
501 if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
504 if (opt->ts_needaddr) {
506 memset(&optptr[optptr[2]-1], 0, 4);
511 static struct ip_options_rcu *ip_options_get_alloc(const int optlen)
513 return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
517 static int ip_options_get_finish(struct net *net, struct ip_options_rcu **optp,
518 struct ip_options_rcu *opt, int optlen)
521 opt->opt.__data[optlen++] = IPOPT_END;
522 opt->opt.optlen = optlen;
523 if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
532 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
533 unsigned char __user *data, int optlen)
535 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
539 if (optlen && copy_from_user(opt->opt.__data, data, optlen)) {
543 return ip_options_get_finish(net, optp, opt, optlen);
546 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
547 unsigned char *data, int optlen)
549 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
554 memcpy(opt->opt.__data, data, optlen);
555 return ip_options_get_finish(net, optp, opt, optlen);
558 void ip_forward_options(struct sk_buff *skb)
560 struct ip_options *opt = &(IPCB(skb)->opt);
561 unsigned char *optptr;
562 struct rtable *rt = skb_rtable(skb);
563 unsigned char *raw = skb_network_header(skb);
565 if (opt->rr_needaddr) {
566 optptr = (unsigned char *)raw + opt->rr;
567 ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
570 if (opt->srr_is_hit) {
571 int srrptr, srrspace;
573 optptr = raw + opt->srr;
575 for ( srrptr=optptr[2], srrspace = optptr[1];
579 if (srrptr + 3 > srrspace)
581 if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
584 if (srrptr + 3 <= srrspace) {
586 ip_hdr(skb)->daddr = opt->nexthop;
587 ip_rt_get_source(&optptr[srrptr-1], skb, rt);
588 optptr[2] = srrptr+4;
590 net_crit_ratelimited("%s(): Argh! Destination lost!\n",
593 if (opt->ts_needaddr) {
594 optptr = raw + opt->ts;
595 ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
599 if (opt->is_changed) {
601 ip_send_check(ip_hdr(skb));
605 int ip_options_rcv_srr(struct sk_buff *skb)
607 struct ip_options *opt = &(IPCB(skb)->opt);
608 int srrspace, srrptr;
610 struct iphdr *iph = ip_hdr(skb);
611 unsigned char *optptr = skb_network_header(skb) + opt->srr;
612 struct rtable *rt = skb_rtable(skb);
614 unsigned long orefdst;
620 if (skb->pkt_type != PACKET_HOST)
622 if (rt->rt_type == RTN_UNICAST) {
623 if (!opt->is_strictroute)
625 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
628 if (rt->rt_type != RTN_LOCAL)
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));
636 memcpy(&nexthop, &optptr[srrptr-1], 4);
638 orefdst = skb->_skb_refdst;
639 skb_dst_set(skb, NULL);
640 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
641 rt2 = skb_rtable(skb);
642 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
644 skb->_skb_refdst = orefdst;
647 refdst_drop(orefdst);
648 if (rt2->rt_type != RTN_LOCAL)
650 /* Superfast 8) loopback forward */
651 iph->daddr = nexthop;
654 if (srrptr <= srrspace) {
656 opt->nexthop = nexthop;
661 EXPORT_SYMBOL(ip_options_rcv_srr);