blob: cbcae6544622932cbb7f0a57d648338bd699ceb9 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 * Version: $Id: ip_options.c,v 1.21 2001/09/01 00:31:50 davem Exp $
9 *
10 * Authors: A.N.Kuznetsov
11 *
12 */
13
Randy Dunlap4fc268d2006-01-11 12:17:47 -080014#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/module.h>
16#include <linux/types.h>
17#include <asm/uaccess.h>
18#include <linux/skbuff.h>
19#include <linux/ip.h>
20#include <linux/icmp.h>
21#include <linux/netdevice.h>
22#include <linux/rtnetlink.h>
23#include <net/sock.h>
24#include <net/ip.h>
25#include <net/icmp.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020026#include <net/route.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027
28/*
29 * Write options to IP header, record destination address to
30 * source route option, address of outgoing interface
31 * (we should already know it, so that this function is allowed be
32 * called only after routing decision) and timestamp,
33 * if we originate this datagram.
34 *
35 * daddr is real destination address, next hop is recorded in IP header.
36 * saddr is address of outgoing interface.
37 */
38
39void ip_options_build(struct sk_buff * skb, struct ip_options * opt,
40 u32 daddr, struct rtable *rt, int is_frag)
41{
42 unsigned char * iph = skb->nh.raw;
43
44 memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
45 memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
46 opt = &(IPCB(skb)->opt);
47 opt->is_data = 0;
48
49 if (opt->srr)
50 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
51
52 if (!is_frag) {
53 if (opt->rr_needaddr)
54 ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, rt);
55 if (opt->ts_needaddr)
56 ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, rt);
57 if (opt->ts_needtime) {
58 struct timeval tv;
59 __u32 midtime;
60 do_gettimeofday(&tv);
61 midtime = htonl((tv.tv_sec % 86400) * 1000 + tv.tv_usec / 1000);
62 memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
63 }
64 return;
65 }
66 if (opt->rr) {
67 memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
68 opt->rr = 0;
69 opt->rr_needaddr = 0;
70 }
71 if (opt->ts) {
72 memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
73 opt->ts = 0;
74 opt->ts_needaddr = opt->ts_needtime = 0;
75 }
76}
77
78/*
79 * Provided (sopt, skb) points to received options,
80 * build in dopt compiled option set appropriate for answering.
81 * i.e. invert SRR option, copy anothers,
82 * and grab room in RR/TS options.
83 *
84 * NOTE: dopt cannot point to skb.
85 */
86
87int ip_options_echo(struct ip_options * dopt, struct sk_buff * skb)
88{
89 struct ip_options *sopt;
90 unsigned char *sptr, *dptr;
91 int soffset, doffset;
92 int optlen;
93 u32 daddr;
94
95 memset(dopt, 0, sizeof(struct ip_options));
96
97 dopt->is_data = 1;
98
99 sopt = &(IPCB(skb)->opt);
100
101 if (sopt->optlen == 0) {
102 dopt->optlen = 0;
103 return 0;
104 }
105
106 sptr = skb->nh.raw;
107 dptr = dopt->__data;
108
109 if (skb->dst)
110 daddr = ((struct rtable*)skb->dst)->rt_spec_dst;
111 else
112 daddr = skb->nh.iph->daddr;
113
114 if (sopt->rr) {
115 optlen = sptr[sopt->rr+1];
116 soffset = sptr[sopt->rr+2];
117 dopt->rr = dopt->optlen + sizeof(struct iphdr);
118 memcpy(dptr, sptr+sopt->rr, optlen);
119 if (sopt->rr_needaddr && soffset <= optlen) {
120 if (soffset + 3 > optlen)
121 return -EINVAL;
122 dptr[2] = soffset + 4;
123 dopt->rr_needaddr = 1;
124 }
125 dptr += optlen;
126 dopt->optlen += optlen;
127 }
128 if (sopt->ts) {
129 optlen = sptr[sopt->ts+1];
130 soffset = sptr[sopt->ts+2];
131 dopt->ts = dopt->optlen + sizeof(struct iphdr);
132 memcpy(dptr, sptr+sopt->ts, optlen);
133 if (soffset <= optlen) {
134 if (sopt->ts_needaddr) {
135 if (soffset + 3 > optlen)
136 return -EINVAL;
137 dopt->ts_needaddr = 1;
138 soffset += 4;
139 }
140 if (sopt->ts_needtime) {
141 if (soffset + 3 > optlen)
142 return -EINVAL;
143 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
144 dopt->ts_needtime = 1;
145 soffset += 4;
146 } else {
147 dopt->ts_needtime = 0;
148
149 if (soffset + 8 <= optlen) {
150 __u32 addr;
151
152 memcpy(&addr, sptr+soffset-1, 4);
153 if (inet_addr_type(addr) != RTN_LOCAL) {
154 dopt->ts_needtime = 1;
155 soffset += 8;
156 }
157 }
158 }
159 }
160 dptr[2] = soffset;
161 }
162 dptr += optlen;
163 dopt->optlen += optlen;
164 }
165 if (sopt->srr) {
166 unsigned char * start = sptr+sopt->srr;
167 u32 faddr;
168
169 optlen = start[1];
170 soffset = start[2];
171 doffset = 0;
172 if (soffset > optlen)
173 soffset = optlen + 1;
174 soffset -= 4;
175 if (soffset > 3) {
176 memcpy(&faddr, &start[soffset-1], 4);
177 for (soffset-=4, doffset=4; soffset > 3; soffset-=4, doffset+=4)
178 memcpy(&dptr[doffset-1], &start[soffset-1], 4);
179 /*
180 * RFC1812 requires to fix illegal source routes.
181 */
182 if (memcmp(&skb->nh.iph->saddr, &start[soffset+3], 4) == 0)
183 doffset -= 4;
184 }
185 if (doffset > 3) {
186 memcpy(&start[doffset-1], &daddr, 4);
187 dopt->faddr = faddr;
188 dptr[0] = start[0];
189 dptr[1] = doffset+3;
190 dptr[2] = 4;
191 dptr += doffset+3;
192 dopt->srr = dopt->optlen + sizeof(struct iphdr);
193 dopt->optlen += doffset+3;
194 dopt->is_strictroute = sopt->is_strictroute;
195 }
196 }
197 while (dopt->optlen & 3) {
198 *dptr++ = IPOPT_END;
199 dopt->optlen++;
200 }
201 return 0;
202}
203
204/*
205 * Options "fragmenting", just fill options not
206 * allowed in fragments with NOOPs.
207 * Simple and stupid 8), but the most efficient way.
208 */
209
210void ip_options_fragment(struct sk_buff * skb)
211{
Wei Yongjun63cbd2f2006-05-09 15:18:50 -0700212 unsigned char * optptr = skb->nh.raw + sizeof(struct iphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 struct ip_options * opt = &(IPCB(skb)->opt);
214 int l = opt->optlen;
215 int optlen;
216
217 while (l > 0) {
218 switch (*optptr) {
219 case IPOPT_END:
220 return;
221 case IPOPT_NOOP:
222 l--;
223 optptr++;
224 continue;
225 }
226 optlen = optptr[1];
227 if (optlen<2 || optlen>l)
228 return;
229 if (!IPOPT_COPIED(*optptr))
230 memset(optptr, IPOPT_NOOP, optlen);
231 l -= optlen;
232 optptr += optlen;
233 }
234 opt->ts = 0;
235 opt->rr = 0;
236 opt->rr_needaddr = 0;
237 opt->ts_needaddr = 0;
238 opt->ts_needtime = 0;
239 return;
240}
241
242/*
243 * Verify options and fill pointers in struct options.
244 * Caller should clear *opt, and set opt->data.
245 * If opt == NULL, then skb->data should point to IP header.
246 */
247
248int ip_options_compile(struct ip_options * opt, struct sk_buff * skb)
249{
250 int l;
251 unsigned char * iph;
252 unsigned char * optptr;
253 int optlen;
254 unsigned char * pp_ptr = NULL;
255 struct rtable *rt = skb ? (struct rtable*)skb->dst : NULL;
256
257 if (!opt) {
258 opt = &(IPCB(skb)->opt);
259 memset(opt, 0, sizeof(struct ip_options));
260 iph = skb->nh.raw;
261 opt->optlen = ((struct iphdr *)iph)->ihl*4 - sizeof(struct iphdr);
262 optptr = iph + sizeof(struct iphdr);
263 opt->is_data = 0;
264 } else {
265 optptr = opt->is_data ? opt->__data : (unsigned char*)&(skb->nh.iph[1]);
266 iph = optptr - sizeof(struct iphdr);
267 }
268
269 for (l = opt->optlen; l > 0; ) {
270 switch (*optptr) {
271 case IPOPT_END:
272 for (optptr++, l--; l>0; optptr++, l--) {
273 if (*optptr != IPOPT_END) {
274 *optptr = IPOPT_END;
275 opt->is_changed = 1;
276 }
277 }
278 goto eol;
279 case IPOPT_NOOP:
280 l--;
281 optptr++;
282 continue;
283 }
284 optlen = optptr[1];
285 if (optlen<2 || optlen>l) {
286 pp_ptr = optptr;
287 goto error;
288 }
289 switch (*optptr) {
290 case IPOPT_SSRR:
291 case IPOPT_LSRR:
292 if (optlen < 3) {
293 pp_ptr = optptr + 1;
294 goto error;
295 }
296 if (optptr[2] < 4) {
297 pp_ptr = optptr + 2;
298 goto error;
299 }
300 /* NB: cf RFC-1812 5.2.4.1 */
301 if (opt->srr) {
302 pp_ptr = optptr;
303 goto error;
304 }
305 if (!skb) {
306 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
307 pp_ptr = optptr + 1;
308 goto error;
309 }
310 memcpy(&opt->faddr, &optptr[3], 4);
311 if (optlen > 7)
312 memmove(&optptr[3], &optptr[7], optlen-7);
313 }
314 opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
315 opt->srr = optptr - iph;
316 break;
317 case IPOPT_RR:
318 if (opt->rr) {
319 pp_ptr = optptr;
320 goto error;
321 }
322 if (optlen < 3) {
323 pp_ptr = optptr + 1;
324 goto error;
325 }
326 if (optptr[2] < 4) {
327 pp_ptr = optptr + 2;
328 goto error;
329 }
330 if (optptr[2] <= optlen) {
331 if (optptr[2]+3 > optlen) {
332 pp_ptr = optptr + 2;
333 goto error;
334 }
335 if (skb) {
336 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
337 opt->is_changed = 1;
338 }
339 optptr[2] += 4;
340 opt->rr_needaddr = 1;
341 }
342 opt->rr = optptr - iph;
343 break;
344 case IPOPT_TIMESTAMP:
345 if (opt->ts) {
346 pp_ptr = optptr;
347 goto error;
348 }
349 if (optlen < 4) {
350 pp_ptr = optptr + 1;
351 goto error;
352 }
353 if (optptr[2] < 5) {
354 pp_ptr = optptr + 2;
355 goto error;
356 }
357 if (optptr[2] <= optlen) {
358 __u32 * timeptr = NULL;
359 if (optptr[2]+3 > optptr[1]) {
360 pp_ptr = optptr + 2;
361 goto error;
362 }
363 switch (optptr[3]&0xF) {
364 case IPOPT_TS_TSONLY:
365 opt->ts = optptr - iph;
366 if (skb)
367 timeptr = (__u32*)&optptr[optptr[2]-1];
368 opt->ts_needtime = 1;
369 optptr[2] += 4;
370 break;
371 case IPOPT_TS_TSANDADDR:
372 if (optptr[2]+7 > optptr[1]) {
373 pp_ptr = optptr + 2;
374 goto error;
375 }
376 opt->ts = optptr - iph;
377 if (skb) {
378 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
379 timeptr = (__u32*)&optptr[optptr[2]+3];
380 }
381 opt->ts_needaddr = 1;
382 opt->ts_needtime = 1;
383 optptr[2] += 8;
384 break;
385 case IPOPT_TS_PRESPEC:
386 if (optptr[2]+7 > optptr[1]) {
387 pp_ptr = optptr + 2;
388 goto error;
389 }
390 opt->ts = optptr - iph;
391 {
392 u32 addr;
393 memcpy(&addr, &optptr[optptr[2]-1], 4);
394 if (inet_addr_type(addr) == RTN_UNICAST)
395 break;
396 if (skb)
397 timeptr = (__u32*)&optptr[optptr[2]+3];
398 }
399 opt->ts_needtime = 1;
400 optptr[2] += 8;
401 break;
402 default:
403 if (!skb && !capable(CAP_NET_RAW)) {
404 pp_ptr = optptr + 3;
405 goto error;
406 }
407 break;
408 }
409 if (timeptr) {
410 struct timeval tv;
411 __u32 midtime;
412 do_gettimeofday(&tv);
413 midtime = htonl((tv.tv_sec % 86400) * 1000 + tv.tv_usec / 1000);
414 memcpy(timeptr, &midtime, sizeof(__u32));
415 opt->is_changed = 1;
416 }
417 } else {
418 unsigned overflow = optptr[3]>>4;
419 if (overflow == 15) {
420 pp_ptr = optptr + 3;
421 goto error;
422 }
423 opt->ts = optptr - iph;
424 if (skb) {
425 optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
426 opt->is_changed = 1;
427 }
428 }
429 break;
430 case IPOPT_RA:
431 if (optlen < 4) {
432 pp_ptr = optptr + 1;
433 goto error;
434 }
435 if (optptr[2] == 0 && optptr[3] == 0)
436 opt->router_alert = optptr - iph;
437 break;
438 case IPOPT_SEC:
439 case IPOPT_SID:
440 default:
441 if (!skb && !capable(CAP_NET_RAW)) {
442 pp_ptr = optptr;
443 goto error;
444 }
445 break;
446 }
447 l -= optlen;
448 optptr += optlen;
449 }
450
451eol:
452 if (!pp_ptr)
453 return 0;
454
455error:
456 if (skb) {
457 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
458 }
459 return -EINVAL;
460}
461
462
463/*
464 * Undo all the changes done by ip_options_compile().
465 */
466
467void ip_options_undo(struct ip_options * opt)
468{
469 if (opt->srr) {
470 unsigned char * optptr = opt->__data+opt->srr-sizeof(struct iphdr);
471 memmove(optptr+7, optptr+3, optptr[1]-7);
472 memcpy(optptr+3, &opt->faddr, 4);
473 }
474 if (opt->rr_needaddr) {
475 unsigned char * optptr = opt->__data+opt->rr-sizeof(struct iphdr);
476 optptr[2] -= 4;
477 memset(&optptr[optptr[2]-1], 0, 4);
478 }
479 if (opt->ts) {
480 unsigned char * optptr = opt->__data+opt->ts-sizeof(struct iphdr);
481 if (opt->ts_needtime) {
482 optptr[2] -= 4;
483 memset(&optptr[optptr[2]-1], 0, 4);
484 if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
485 optptr[2] -= 4;
486 }
487 if (opt->ts_needaddr) {
488 optptr[2] -= 4;
489 memset(&optptr[optptr[2]-1], 0, 4);
490 }
491 }
492}
493
Arnaldo Carvalho de Melo4c6ea292005-08-16 19:46:48 -0300494static struct ip_options *ip_options_get_alloc(const int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495{
Arnaldo Carvalho de Melo4c6ea292005-08-16 19:46:48 -0300496 struct ip_options *opt = kmalloc(sizeof(*opt) + ((optlen + 3) & ~3),
497 GFP_KERNEL);
498 if (opt)
499 memset(opt, 0, sizeof(*opt));
500 return opt;
501}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502
Arnaldo Carvalho de Melo4c6ea292005-08-16 19:46:48 -0300503static int ip_options_get_finish(struct ip_options **optp,
504 struct ip_options *opt, int optlen)
505{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 while (optlen & 3)
507 opt->__data[optlen++] = IPOPT_END;
508 opt->optlen = optlen;
509 opt->is_data = 1;
510 opt->is_setbyuser = 1;
511 if (optlen && ip_options_compile(opt, NULL)) {
512 kfree(opt);
513 return -EINVAL;
514 }
Jesper Juhla51482b2005-11-08 09:41:34 -0800515 kfree(*optp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 *optp = opt;
517 return 0;
518}
519
Arnaldo Carvalho de Melo4c6ea292005-08-16 19:46:48 -0300520int ip_options_get_from_user(struct ip_options **optp, unsigned char __user *data, int optlen)
521{
522 struct ip_options *opt = ip_options_get_alloc(optlen);
523
524 if (!opt)
525 return -ENOMEM;
526 if (optlen && copy_from_user(opt->__data, data, optlen)) {
527 kfree(opt);
528 return -EFAULT;
529 }
530 return ip_options_get_finish(optp, opt, optlen);
531}
532
533int ip_options_get(struct ip_options **optp, unsigned char *data, int optlen)
534{
535 struct ip_options *opt = ip_options_get_alloc(optlen);
536
537 if (!opt)
538 return -ENOMEM;
539 if (optlen)
540 memcpy(opt->__data, data, optlen);
541 return ip_options_get_finish(optp, opt, optlen);
542}
543
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544void ip_forward_options(struct sk_buff *skb)
545{
546 struct ip_options * opt = &(IPCB(skb)->opt);
547 unsigned char * optptr;
548 struct rtable *rt = (struct rtable*)skb->dst;
549 unsigned char *raw = skb->nh.raw;
550
551 if (opt->rr_needaddr) {
552 optptr = (unsigned char *)raw + opt->rr;
553 ip_rt_get_source(&optptr[optptr[2]-5], rt);
554 opt->is_changed = 1;
555 }
556 if (opt->srr_is_hit) {
557 int srrptr, srrspace;
558
559 optptr = raw + opt->srr;
560
561 for ( srrptr=optptr[2], srrspace = optptr[1];
562 srrptr <= srrspace;
563 srrptr += 4
564 ) {
565 if (srrptr + 3 > srrspace)
566 break;
567 if (memcmp(&rt->rt_dst, &optptr[srrptr-1], 4) == 0)
568 break;
569 }
570 if (srrptr + 3 <= srrspace) {
571 opt->is_changed = 1;
572 ip_rt_get_source(&optptr[srrptr-1], rt);
573 skb->nh.iph->daddr = rt->rt_dst;
574 optptr[2] = srrptr+4;
575 } else if (net_ratelimit())
576 printk(KERN_CRIT "ip_forward(): Argh! Destination lost!\n");
577 if (opt->ts_needaddr) {
578 optptr = raw + opt->ts;
579 ip_rt_get_source(&optptr[optptr[2]-9], rt);
580 opt->is_changed = 1;
581 }
582 }
583 if (opt->is_changed) {
584 opt->is_changed = 0;
585 ip_send_check(skb->nh.iph);
586 }
587}
588
589int ip_options_rcv_srr(struct sk_buff *skb)
590{
591 struct ip_options *opt = &(IPCB(skb)->opt);
592 int srrspace, srrptr;
593 u32 nexthop;
594 struct iphdr *iph = skb->nh.iph;
595 unsigned char * optptr = skb->nh.raw + opt->srr;
596 struct rtable *rt = (struct rtable*)skb->dst;
597 struct rtable *rt2;
598 int err;
599
600 if (!opt->srr)
601 return 0;
602
603 if (skb->pkt_type != PACKET_HOST)
604 return -EINVAL;
605 if (rt->rt_type == RTN_UNICAST) {
606 if (!opt->is_strictroute)
607 return 0;
608 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
609 return -EINVAL;
610 }
611 if (rt->rt_type != RTN_LOCAL)
612 return -EINVAL;
613
614 for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
615 if (srrptr + 3 > srrspace) {
616 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
617 return -EINVAL;
618 }
619 memcpy(&nexthop, &optptr[srrptr-1], 4);
620
621 rt = (struct rtable*)skb->dst;
622 skb->dst = NULL;
623 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
624 rt2 = (struct rtable*)skb->dst;
625 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
626 ip_rt_put(rt2);
627 skb->dst = &rt->u.dst;
628 return -EINVAL;
629 }
630 ip_rt_put(rt);
631 if (rt2->rt_type != RTN_LOCAL)
632 break;
633 /* Superfast 8) loopback forward */
634 memcpy(&iph->daddr, &optptr[srrptr-1], 4);
635 opt->is_changed = 1;
636 }
637 if (srrptr <= srrspace) {
638 opt->srr_is_hit = 1;
639 opt->is_changed = 1;
640 }
641 return 0;
642}