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 *
8 * Authors: A.N.Kuznetsov
9 *
10 */
11
12 #define pr_fmt(fmt) "IPv4: " fmt
13
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>
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>
28 #include <net/route.h>
29 #include <net/cipso_ipv4.h>
30
31 /*
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.
37 *
38 * daddr is real destination address, next hop is recorded in IP header.
39 * saddr is address of outgoing interface.
40 */
41
ip_options_build(struct sk_buff * skb,struct ip_options * opt,__be32 daddr,struct rtable * rt,int is_frag)42 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
43 __be32 daddr, struct rtable *rt, int is_frag)
44 {
45 unsigned char *iph = skb_network_header(skb);
46
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);
50
51 if (opt->srr)
52 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
53
54 if (!is_frag) {
55 if (opt->rr_needaddr)
56 ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, skb, rt);
57 if (opt->ts_needaddr)
58 ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, skb, rt);
59 if (opt->ts_needtime) {
60 struct timespec tv;
61 __be32 midtime;
62 getnstimeofday(&tv);
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);
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
80 /*
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
ip_options_echo(struct ip_options * dopt,struct sk_buff * skb)89 int ip_options_echo(struct ip_options *dopt, struct sk_buff *skb)
90 {
91 const struct ip_options *sopt;
92 unsigned char *sptr, *dptr;
93 int soffset, doffset;
94 int optlen;
95 __be32 daddr;
96
97 memset(dopt, 0, sizeof(struct ip_options));
98
99 sopt = &(IPCB(skb)->opt);
100
101 if (sopt->optlen == 0)
102 return 0;
103
104 sptr = skb_network_header(skb);
105 dptr = dopt->__data;
106
107 daddr = skb_rtable(skb)->rt_spec_dst;
108
109 if (sopt->rr) {
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)
116 return -EINVAL;
117 dptr[2] = soffset + 4;
118 dopt->rr_needaddr = 1;
119 }
120 dptr += optlen;
121 dopt->optlen += optlen;
122 }
123 if (sopt->ts) {
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)
131 return -EINVAL;
132 dopt->ts_needaddr = 1;
133 soffset += 4;
134 }
135 if (sopt->ts_needtime) {
136 if (soffset + 3 > optlen)
137 return -EINVAL;
138 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
139 dopt->ts_needtime = 1;
140 soffset += 4;
141 } else {
142 dopt->ts_needtime = 0;
143
144 if (soffset + 7 <= optlen) {
145 __be32 addr;
146
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;
150 soffset += 8;
151 }
152 }
153 }
154 }
155 dptr[2] = soffset;
156 }
157 dptr += optlen;
158 dopt->optlen += optlen;
159 }
160 if (sopt->srr) {
161 unsigned char *start = sptr+sopt->srr;
162 __be32 faddr;
163
164 optlen = start[1];
165 soffset = start[2];
166 doffset = 0;
167 if (soffset > optlen)
168 soffset = optlen + 1;
169 soffset -= 4;
170 if (soffset > 3) {
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);
174 /*
175 * RFC1812 requires to fix illegal source routes.
176 */
177 if (memcmp(&ip_hdr(skb)->saddr,
178 &start[soffset + 3], 4) == 0)
179 doffset -= 4;
180 }
181 if (doffset > 3) {
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 }
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 }
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
ip_options_fragment(struct sk_buff * skb)213 void ip_options_fragment(struct sk_buff * skb)
214 {
215 unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
216 struct ip_options * opt = &(IPCB(skb)->opt);
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;
242 }
243
244 /*
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.
248 */
249
ip_options_compile(struct net * net,struct ip_options * opt,struct sk_buff * skb)250 int ip_options_compile(struct net *net,
251 struct ip_options * opt, struct sk_buff * skb)
252 {
253 int l;
254 unsigned char * iph;
255 unsigned char * optptr;
256 int optlen;
257 unsigned char * pp_ptr = NULL;
258 struct rtable *rt = NULL;
259
260 if (skb != NULL) {
261 rt = skb_rtable(skb);
262 optptr = (unsigned char *)&(ip_hdr(skb)[1]);
263 } else
264 optptr = opt->__data;
265 iph = optptr - sizeof(struct iphdr);
266
267 for (l = opt->optlen; l > 0; ) {
268 switch (*optptr) {
269 case IPOPT_END:
270 for (optptr++, l--; l>0; optptr++, l--) {
271 if (*optptr != IPOPT_END) {
272 *optptr = IPOPT_END;
273 opt->is_changed = 1;
274 }
275 }
276 goto eol;
277 case IPOPT_NOOP:
278 l--;
279 optptr++;
280 continue;
281 }
282 optlen = optptr[1];
283 if (optlen<2 || optlen>l) {
284 pp_ptr = optptr;
285 goto error;
286 }
287 switch (*optptr) {
288 case IPOPT_SSRR:
289 case IPOPT_LSRR:
290 if (optlen < 3) {
291 pp_ptr = optptr + 1;
292 goto error;
293 }
294 if (optptr[2] < 4) {
295 pp_ptr = optptr + 2;
296 goto error;
297 }
298 /* NB: cf RFC-1812 5.2.4.1 */
299 if (opt->srr) {
300 pp_ptr = optptr;
301 goto error;
302 }
303 if (!skb) {
304 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
305 pp_ptr = optptr + 1;
306 goto error;
307 }
308 memcpy(&opt->faddr, &optptr[3], 4);
309 if (optlen > 7)
310 memmove(&optptr[3], &optptr[7], optlen-7);
311 }
312 opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
313 opt->srr = optptr - iph;
314 break;
315 case IPOPT_RR:
316 if (opt->rr) {
317 pp_ptr = optptr;
318 goto error;
319 }
320 if (optlen < 3) {
321 pp_ptr = optptr + 1;
322 goto error;
323 }
324 if (optptr[2] < 4) {
325 pp_ptr = optptr + 2;
326 goto error;
327 }
328 if (optptr[2] <= optlen) {
329 if (optptr[2]+3 > optlen) {
330 pp_ptr = optptr + 2;
331 goto error;
332 }
333 if (rt) {
334 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
335 opt->is_changed = 1;
336 }
337 optptr[2] += 4;
338 opt->rr_needaddr = 1;
339 }
340 opt->rr = optptr - iph;
341 break;
342 case IPOPT_TIMESTAMP:
343 if (opt->ts) {
344 pp_ptr = optptr;
345 goto error;
346 }
347 if (optlen < 4) {
348 pp_ptr = optptr + 1;
349 goto error;
350 }
351 if (optptr[2] < 5) {
352 pp_ptr = optptr + 2;
353 goto error;
354 }
355 if (optptr[2] <= optlen) {
356 unsigned char *timeptr = NULL;
357 if (optptr[2]+3 > optptr[1]) {
358 pp_ptr = optptr + 2;
359 goto error;
360 }
361 switch (optptr[3]&0xF) {
362 case IPOPT_TS_TSONLY:
363 if (skb)
364 timeptr = &optptr[optptr[2]-1];
365 opt->ts_needtime = 1;
366 optptr[2] += 4;
367 break;
368 case IPOPT_TS_TSANDADDR:
369 if (optptr[2]+7 > optptr[1]) {
370 pp_ptr = optptr + 2;
371 goto error;
372 }
373 if (rt) {
374 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
375 timeptr = &optptr[optptr[2]+3];
376 }
377 opt->ts_needaddr = 1;
378 opt->ts_needtime = 1;
379 optptr[2] += 8;
380 break;
381 case IPOPT_TS_PRESPEC:
382 if (optptr[2]+7 > optptr[1]) {
383 pp_ptr = optptr + 2;
384 goto error;
385 }
386 {
387 __be32 addr;
388 memcpy(&addr, &optptr[optptr[2]-1], 4);
389 if (inet_addr_type(net, addr) == RTN_UNICAST)
390 break;
391 if (skb)
392 timeptr = &optptr[optptr[2]+3];
393 }
394 opt->ts_needtime = 1;
395 optptr[2] += 8;
396 break;
397 default:
398 if (!skb && !capable(CAP_NET_RAW)) {
399 pp_ptr = optptr + 3;
400 goto error;
401 }
402 break;
403 }
404 if (timeptr) {
405 struct timespec tv;
406 u32 midtime;
407 getnstimeofday(&tv);
408 midtime = (tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC;
409 put_unaligned_be32(midtime, timeptr);
410 opt->is_changed = 1;
411 }
412 } else {
413 unsigned overflow = optptr[3]>>4;
414 if (overflow == 15) {
415 pp_ptr = optptr + 3;
416 goto error;
417 }
418 if (skb) {
419 optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
420 opt->is_changed = 1;
421 }
422 }
423 opt->ts = optptr - iph;
424 break;
425 case IPOPT_RA:
426 if (optlen < 4) {
427 pp_ptr = optptr + 1;
428 goto error;
429 }
430 if (optptr[2] == 0 && optptr[3] == 0)
431 opt->router_alert = optptr - iph;
432 break;
433 case IPOPT_CIPSO:
434 if ((!skb && !capable(CAP_NET_RAW)) || opt->cipso) {
435 pp_ptr = optptr;
436 goto error;
437 }
438 opt->cipso = optptr - iph;
439 if (cipso_v4_validate(skb, &optptr)) {
440 pp_ptr = optptr;
441 goto error;
442 }
443 break;
444 case IPOPT_SEC:
445 case IPOPT_SID:
446 default:
447 if (!skb && !capable(CAP_NET_RAW)) {
448 pp_ptr = optptr;
449 goto error;
450 }
451 break;
452 }
453 l -= optlen;
454 optptr += optlen;
455 }
456
457 eol:
458 if (!pp_ptr)
459 return 0;
460
461 error:
462 if (skb) {
463 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
464 }
465 return -EINVAL;
466 }
467 EXPORT_SYMBOL(ip_options_compile);
468
469 /*
470 * Undo all the changes done by ip_options_compile().
471 */
472
ip_options_undo(struct ip_options * opt)473 void ip_options_undo(struct ip_options * opt)
474 {
475 if (opt->srr) {
476 unsigned char * optptr = opt->__data+opt->srr-sizeof(struct iphdr);
477 memmove(optptr+7, optptr+3, optptr[1]-7);
478 memcpy(optptr+3, &opt->faddr, 4);
479 }
480 if (opt->rr_needaddr) {
481 unsigned char * optptr = opt->__data+opt->rr-sizeof(struct iphdr);
482 optptr[2] -= 4;
483 memset(&optptr[optptr[2]-1], 0, 4);
484 }
485 if (opt->ts) {
486 unsigned char * optptr = opt->__data+opt->ts-sizeof(struct iphdr);
487 if (opt->ts_needtime) {
488 optptr[2] -= 4;
489 memset(&optptr[optptr[2]-1], 0, 4);
490 if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
491 optptr[2] -= 4;
492 }
493 if (opt->ts_needaddr) {
494 optptr[2] -= 4;
495 memset(&optptr[optptr[2]-1], 0, 4);
496 }
497 }
498 }
499
ip_options_get_alloc(const int optlen)500 static struct ip_options_rcu *ip_options_get_alloc(const int optlen)
501 {
502 return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
503 GFP_KERNEL);
504 }
505
ip_options_get_finish(struct net * net,struct ip_options_rcu ** optp,struct ip_options_rcu * opt,int optlen)506 static int ip_options_get_finish(struct net *net, struct ip_options_rcu **optp,
507 struct ip_options_rcu *opt, int optlen)
508 {
509 while (optlen & 3)
510 opt->opt.__data[optlen++] = IPOPT_END;
511 opt->opt.optlen = optlen;
512 if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
513 kfree(opt);
514 return -EINVAL;
515 }
516 kfree(*optp);
517 *optp = opt;
518 return 0;
519 }
520
ip_options_get_from_user(struct net * net,struct ip_options_rcu ** optp,unsigned char __user * data,int optlen)521 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
522 unsigned char __user *data, int optlen)
523 {
524 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
525
526 if (!opt)
527 return -ENOMEM;
528 if (optlen && copy_from_user(opt->opt.__data, data, optlen)) {
529 kfree(opt);
530 return -EFAULT;
531 }
532 return ip_options_get_finish(net, optp, opt, optlen);
533 }
534
ip_options_get(struct net * net,struct ip_options_rcu ** optp,unsigned char * data,int optlen)535 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
536 unsigned char *data, int optlen)
537 {
538 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
539
540 if (!opt)
541 return -ENOMEM;
542 if (optlen)
543 memcpy(opt->opt.__data, data, optlen);
544 return ip_options_get_finish(net, optp, opt, optlen);
545 }
546
ip_forward_options(struct sk_buff * skb)547 void ip_forward_options(struct sk_buff *skb)
548 {
549 struct ip_options * opt = &(IPCB(skb)->opt);
550 unsigned char * optptr;
551 struct rtable *rt = skb_rtable(skb);
552 unsigned char *raw = skb_network_header(skb);
553
554 if (opt->rr_needaddr) {
555 optptr = (unsigned char *)raw + opt->rr;
556 ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
557 opt->is_changed = 1;
558 }
559 if (opt->srr_is_hit) {
560 int srrptr, srrspace;
561
562 optptr = raw + opt->srr;
563
564 for ( srrptr=optptr[2], srrspace = optptr[1];
565 srrptr <= srrspace;
566 srrptr += 4
567 ) {
568 if (srrptr + 3 > srrspace)
569 break;
570 if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
571 break;
572 }
573 if (srrptr + 3 <= srrspace) {
574 opt->is_changed = 1;
575 ip_hdr(skb)->daddr = opt->nexthop;
576 ip_rt_get_source(&optptr[srrptr-1], skb, rt);
577 optptr[2] = srrptr+4;
578 } else if (net_ratelimit())
579 pr_crit("%s(): Argh! Destination lost!\n", __func__);
580 if (opt->ts_needaddr) {
581 optptr = raw + opt->ts;
582 ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
583 opt->is_changed = 1;
584 }
585 }
586 if (opt->is_changed) {
587 opt->is_changed = 0;
588 ip_send_check(ip_hdr(skb));
589 }
590 }
591
ip_options_rcv_srr(struct sk_buff * skb)592 int ip_options_rcv_srr(struct sk_buff *skb)
593 {
594 struct ip_options *opt = &(IPCB(skb)->opt);
595 int srrspace, srrptr;
596 __be32 nexthop;
597 struct iphdr *iph = ip_hdr(skb);
598 unsigned char *optptr = skb_network_header(skb) + opt->srr;
599 struct rtable *rt = skb_rtable(skb);
600 struct rtable *rt2;
601 unsigned long orefdst;
602 int err;
603
604 if (!rt)
605 return 0;
606
607 if (skb->pkt_type != PACKET_HOST)
608 return -EINVAL;
609 if (rt->rt_type == RTN_UNICAST) {
610 if (!opt->is_strictroute)
611 return 0;
612 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
613 return -EINVAL;
614 }
615 if (rt->rt_type != RTN_LOCAL)
616 return -EINVAL;
617
618 for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
619 if (srrptr + 3 > srrspace) {
620 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
621 return -EINVAL;
622 }
623 memcpy(&nexthop, &optptr[srrptr-1], 4);
624
625 orefdst = skb->_skb_refdst;
626 skb_dst_set(skb, NULL);
627 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
628 rt2 = skb_rtable(skb);
629 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
630 skb_dst_drop(skb);
631 skb->_skb_refdst = orefdst;
632 return -EINVAL;
633 }
634 refdst_drop(orefdst);
635 if (rt2->rt_type != RTN_LOCAL)
636 break;
637 /* Superfast 8) loopback forward */
638 iph->daddr = nexthop;
639 opt->is_changed = 1;
640 }
641 if (srrptr <= srrspace) {
642 opt->srr_is_hit = 1;
643 opt->nexthop = nexthop;
644 opt->is_changed = 1;
645 }
646 return 0;
647 }
648 EXPORT_SYMBOL(ip_options_rcv_srr);
649