Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
[linux-2.6] / net / ipv4 / ip_options.c
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 #include <linux/capability.h>
13 #include <linux/module.h>
14 #include <linux/types.h>
15 #include <asm/uaccess.h>
16 #include <linux/skbuff.h>
17 #include <linux/ip.h>
18 #include <linux/icmp.h>
19 #include <linux/netdevice.h>
20 #include <linux/rtnetlink.h>
21 #include <net/sock.h>
22 #include <net/ip.h>
23 #include <net/icmp.h>
24 #include <net/route.h>
25 #include <net/cipso_ipv4.h>
26
27 /*
28  * Write options to IP header, record destination address to
29  * source route option, address of outgoing interface
30  * (we should already know it, so that this  function is allowed be
31  * called only after routing decision) and timestamp,
32  * if we originate this datagram.
33  *
34  * daddr is real destination address, next hop is recorded in IP header.
35  * saddr is address of outgoing interface.
36  */
37
38 void ip_options_build(struct sk_buff * skb, struct ip_options * opt,
39                             __be32 daddr, struct rtable *rt, int is_frag)
40 {
41         unsigned char *iph = skb_network_header(skb);
42
43         memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
44         memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
45         opt = &(IPCB(skb)->opt);
46
47         if (opt->srr)
48                 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
49
50         if (!is_frag) {
51                 if (opt->rr_needaddr)
52                         ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, rt);
53                 if (opt->ts_needaddr)
54                         ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, rt);
55                 if (opt->ts_needtime) {
56                         struct timespec tv;
57                         __be32 midtime;
58                         getnstimeofday(&tv);
59                         midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
60                         memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
61                 }
62                 return;
63         }
64         if (opt->rr) {
65                 memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
66                 opt->rr = 0;
67                 opt->rr_needaddr = 0;
68         }
69         if (opt->ts) {
70                 memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
71                 opt->ts = 0;
72                 opt->ts_needaddr = opt->ts_needtime = 0;
73         }
74 }
75
76 /*
77  * Provided (sopt, skb) points to received options,
78  * build in dopt compiled option set appropriate for answering.
79  * i.e. invert SRR option, copy anothers,
80  * and grab room in RR/TS options.
81  *
82  * NOTE: dopt cannot point to skb.
83  */
84
85 int ip_options_echo(struct ip_options * dopt, struct sk_buff * skb)
86 {
87         struct ip_options *sopt;
88         unsigned char *sptr, *dptr;
89         int soffset, doffset;
90         int     optlen;
91         __be32  daddr;
92
93         memset(dopt, 0, sizeof(struct ip_options));
94
95         sopt = &(IPCB(skb)->opt);
96
97         if (sopt->optlen == 0) {
98                 dopt->optlen = 0;
99                 return 0;
100         }
101
102         sptr = skb_network_header(skb);
103         dptr = dopt->__data;
104
105         daddr = skb->rtable->rt_spec_dst;
106
107         if (sopt->rr) {
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)
114                                 return -EINVAL;
115                         dptr[2] = soffset + 4;
116                         dopt->rr_needaddr = 1;
117                 }
118                 dptr += optlen;
119                 dopt->optlen += optlen;
120         }
121         if (sopt->ts) {
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)
129                                         return -EINVAL;
130                                 dopt->ts_needaddr = 1;
131                                 soffset += 4;
132                         }
133                         if (sopt->ts_needtime) {
134                                 if (soffset + 3 > optlen)
135                                         return -EINVAL;
136                                 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
137                                         dopt->ts_needtime = 1;
138                                         soffset += 4;
139                                 } else {
140                                         dopt->ts_needtime = 0;
141
142                                         if (soffset + 8 <= optlen) {
143                                                 __be32 addr;
144
145                                                 memcpy(&addr, sptr+soffset-1, 4);
146                                                 if (inet_addr_type(dev_net(skb->dst->dev), addr) != RTN_LOCAL) {
147                                                         dopt->ts_needtime = 1;
148                                                         soffset += 8;
149                                                 }
150                                         }
151                                 }
152                         }
153                         dptr[2] = soffset;
154                 }
155                 dptr += optlen;
156                 dopt->optlen += optlen;
157         }
158         if (sopt->srr) {
159                 unsigned char * start = sptr+sopt->srr;
160                 __be32 faddr;
161
162                 optlen  = start[1];
163                 soffset = start[2];
164                 doffset = 0;
165                 if (soffset > optlen)
166                         soffset = optlen + 1;
167                 soffset -= 4;
168                 if (soffset > 3) {
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);
172                         /*
173                          * RFC1812 requires to fix illegal source routes.
174                          */
175                         if (memcmp(&ip_hdr(skb)->saddr,
176                                    &start[soffset + 3], 4) == 0)
177                                 doffset -= 4;
178                 }
179                 if (doffset > 3) {
180                         memcpy(&start[doffset-1], &daddr, 4);
181                         dopt->faddr = faddr;
182                         dptr[0] = start[0];
183                         dptr[1] = doffset+3;
184                         dptr[2] = 4;
185                         dptr += doffset+3;
186                         dopt->srr = dopt->optlen + sizeof(struct iphdr);
187                         dopt->optlen += doffset+3;
188                         dopt->is_strictroute = sopt->is_strictroute;
189                 }
190         }
191         if (sopt->cipso) {
192                 optlen  = sptr[sopt->cipso+1];
193                 dopt->cipso = dopt->optlen+sizeof(struct iphdr);
194                 memcpy(dptr, sptr+sopt->cipso, optlen);
195                 dptr += optlen;
196                 dopt->optlen += optlen;
197         }
198         while (dopt->optlen & 3) {
199                 *dptr++ = IPOPT_END;
200                 dopt->optlen++;
201         }
202         return 0;
203 }
204
205 /*
206  *      Options "fragmenting", just fill options not
207  *      allowed in fragments with NOOPs.
208  *      Simple and stupid 8), but the most efficient way.
209  */
210
211 void ip_options_fragment(struct sk_buff * skb)
212 {
213         unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
214         struct ip_options * opt = &(IPCB(skb)->opt);
215         int  l = opt->optlen;
216         int  optlen;
217
218         while (l > 0) {
219                 switch (*optptr) {
220                 case IPOPT_END:
221                         return;
222                 case IPOPT_NOOP:
223                         l--;
224                         optptr++;
225                         continue;
226                 }
227                 optlen = optptr[1];
228                 if (optlen<2 || optlen>l)
229                   return;
230                 if (!IPOPT_COPIED(*optptr))
231                         memset(optptr, IPOPT_NOOP, optlen);
232                 l -= optlen;
233                 optptr += optlen;
234         }
235         opt->ts = 0;
236         opt->rr = 0;
237         opt->rr_needaddr = 0;
238         opt->ts_needaddr = 0;
239         opt->ts_needtime = 0;
240         return;
241 }
242
243 /*
244  * Verify options and fill pointers in struct options.
245  * Caller should clear *opt, and set opt->data.
246  * If opt == NULL, then skb->data should point to IP header.
247  */
248
249 int ip_options_compile(struct net *net,
250                        struct ip_options * opt, struct sk_buff * skb)
251 {
252         int l;
253         unsigned char * iph;
254         unsigned char * optptr;
255         int optlen;
256         unsigned char * pp_ptr = NULL;
257         struct rtable *rt = NULL;
258
259         if (skb != NULL) {
260                 rt = skb->rtable;
261                 optptr = (unsigned char *)&(ip_hdr(skb)[1]);
262         } else
263                 optptr = opt->__data;
264         iph = optptr - sizeof(struct iphdr);
265
266         for (l = opt->optlen; l > 0; ) {
267                 switch (*optptr) {
268                       case IPOPT_END:
269                         for (optptr++, l--; l>0; optptr++, l--) {
270                                 if (*optptr != IPOPT_END) {
271                                         *optptr = IPOPT_END;
272                                         opt->is_changed = 1;
273                                 }
274                         }
275                         goto eol;
276                       case IPOPT_NOOP:
277                         l--;
278                         optptr++;
279                         continue;
280                 }
281                 optlen = optptr[1];
282                 if (optlen<2 || optlen>l) {
283                         pp_ptr = optptr;
284                         goto error;
285                 }
286                 switch (*optptr) {
287                       case IPOPT_SSRR:
288                       case IPOPT_LSRR:
289                         if (optlen < 3) {
290                                 pp_ptr = optptr + 1;
291                                 goto error;
292                         }
293                         if (optptr[2] < 4) {
294                                 pp_ptr = optptr + 2;
295                                 goto error;
296                         }
297                         /* NB: cf RFC-1812 5.2.4.1 */
298                         if (opt->srr) {
299                                 pp_ptr = optptr;
300                                 goto error;
301                         }
302                         if (!skb) {
303                                 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
304                                         pp_ptr = optptr + 1;
305                                         goto error;
306                                 }
307                                 memcpy(&opt->faddr, &optptr[3], 4);
308                                 if (optlen > 7)
309                                         memmove(&optptr[3], &optptr[7], optlen-7);
310                         }
311                         opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
312                         opt->srr = optptr - iph;
313                         break;
314                       case IPOPT_RR:
315                         if (opt->rr) {
316                                 pp_ptr = optptr;
317                                 goto error;
318                         }
319                         if (optlen < 3) {
320                                 pp_ptr = optptr + 1;
321                                 goto error;
322                         }
323                         if (optptr[2] < 4) {
324                                 pp_ptr = optptr + 2;
325                                 goto error;
326                         }
327                         if (optptr[2] <= optlen) {
328                                 if (optptr[2]+3 > optlen) {
329                                         pp_ptr = optptr + 2;
330                                         goto error;
331                                 }
332                                 if (skb) {
333                                         memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
334                                         opt->is_changed = 1;
335                                 }
336                                 optptr[2] += 4;
337                                 opt->rr_needaddr = 1;
338                         }
339                         opt->rr = optptr - iph;
340                         break;
341                       case IPOPT_TIMESTAMP:
342                         if (opt->ts) {
343                                 pp_ptr = optptr;
344                                 goto error;
345                         }
346                         if (optlen < 4) {
347                                 pp_ptr = optptr + 1;
348                                 goto error;
349                         }
350                         if (optptr[2] < 5) {
351                                 pp_ptr = optptr + 2;
352                                 goto error;
353                         }
354                         if (optptr[2] <= optlen) {
355                                 __be32 *timeptr = NULL;
356                                 if (optptr[2]+3 > optptr[1]) {
357                                         pp_ptr = optptr + 2;
358                                         goto error;
359                                 }
360                                 switch (optptr[3]&0xF) {
361                                       case IPOPT_TS_TSONLY:
362                                         opt->ts = optptr - iph;
363                                         if (skb)
364                                                 timeptr = (__be32*)&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                                         opt->ts = optptr - iph;
374                                         if (skb) {
375                                                 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
376                                                 timeptr = (__be32*)&optptr[optptr[2]+3];
377                                         }
378                                         opt->ts_needaddr = 1;
379                                         opt->ts_needtime = 1;
380                                         optptr[2] += 8;
381                                         break;
382                                       case IPOPT_TS_PRESPEC:
383                                         if (optptr[2]+7 > optptr[1]) {
384                                                 pp_ptr = optptr + 2;
385                                                 goto error;
386                                         }
387                                         opt->ts = optptr - iph;
388                                         {
389                                                 __be32 addr;
390                                                 memcpy(&addr, &optptr[optptr[2]-1], 4);
391                                                 if (inet_addr_type(net, addr) == RTN_UNICAST)
392                                                         break;
393                                                 if (skb)
394                                                         timeptr = (__be32*)&optptr[optptr[2]+3];
395                                         }
396                                         opt->ts_needtime = 1;
397                                         optptr[2] += 8;
398                                         break;
399                                       default:
400                                         if (!skb && !capable(CAP_NET_RAW)) {
401                                                 pp_ptr = optptr + 3;
402                                                 goto error;
403                                         }
404                                         break;
405                                 }
406                                 if (timeptr) {
407                                         struct timespec tv;
408                                         __be32  midtime;
409                                         getnstimeofday(&tv);
410                                         midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
411                                         memcpy(timeptr, &midtime, sizeof(__be32));
412                                         opt->is_changed = 1;
413                                 }
414                         } else {
415                                 unsigned overflow = optptr[3]>>4;
416                                 if (overflow == 15) {
417                                         pp_ptr = optptr + 3;
418                                         goto error;
419                                 }
420                                 opt->ts = optptr - iph;
421                                 if (skb) {
422                                         optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
423                                         opt->is_changed = 1;
424                                 }
425                         }
426                         break;
427                       case IPOPT_RA:
428                         if (optlen < 4) {
429                                 pp_ptr = optptr + 1;
430                                 goto error;
431                         }
432                         if (optptr[2] == 0 && optptr[3] == 0)
433                                 opt->router_alert = optptr - iph;
434                         break;
435                       case IPOPT_CIPSO:
436                         if ((!skb && !capable(CAP_NET_RAW)) || opt->cipso) {
437                                 pp_ptr = optptr;
438                                 goto error;
439                         }
440                         opt->cipso = optptr - iph;
441                         if (cipso_v4_validate(skb, &optptr)) {
442                                 pp_ptr = optptr;
443                                 goto error;
444                         }
445                         break;
446                       case IPOPT_SEC:
447                       case IPOPT_SID:
448                       default:
449                         if (!skb && !capable(CAP_NET_RAW)) {
450                                 pp_ptr = optptr;
451                                 goto error;
452                         }
453                         break;
454                 }
455                 l -= optlen;
456                 optptr += optlen;
457         }
458
459 eol:
460         if (!pp_ptr)
461                 return 0;
462
463 error:
464         if (skb) {
465                 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
466         }
467         return -EINVAL;
468 }
469
470
471 /*
472  *      Undo all the changes done by ip_options_compile().
473  */
474
475 void ip_options_undo(struct ip_options * opt)
476 {
477         if (opt->srr) {
478                 unsigned  char * optptr = opt->__data+opt->srr-sizeof(struct  iphdr);
479                 memmove(optptr+7, optptr+3, optptr[1]-7);
480                 memcpy(optptr+3, &opt->faddr, 4);
481         }
482         if (opt->rr_needaddr) {
483                 unsigned  char * optptr = opt->__data+opt->rr-sizeof(struct  iphdr);
484                 optptr[2] -= 4;
485                 memset(&optptr[optptr[2]-1], 0, 4);
486         }
487         if (opt->ts) {
488                 unsigned  char * optptr = opt->__data+opt->ts-sizeof(struct  iphdr);
489                 if (opt->ts_needtime) {
490                         optptr[2] -= 4;
491                         memset(&optptr[optptr[2]-1], 0, 4);
492                         if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
493                                 optptr[2] -= 4;
494                 }
495                 if (opt->ts_needaddr) {
496                         optptr[2] -= 4;
497                         memset(&optptr[optptr[2]-1], 0, 4);
498                 }
499         }
500 }
501
502 static struct ip_options *ip_options_get_alloc(const int optlen)
503 {
504         return kzalloc(sizeof(struct ip_options) + ((optlen + 3) & ~3),
505                        GFP_KERNEL);
506 }
507
508 static int ip_options_get_finish(struct net *net, struct ip_options **optp,
509                                  struct ip_options *opt, int optlen)
510 {
511         while (optlen & 3)
512                 opt->__data[optlen++] = IPOPT_END;
513         opt->optlen = optlen;
514         if (optlen && ip_options_compile(net, opt, NULL)) {
515                 kfree(opt);
516                 return -EINVAL;
517         }
518         kfree(*optp);
519         *optp = opt;
520         return 0;
521 }
522
523 int ip_options_get_from_user(struct net *net, struct ip_options **optp,
524                              unsigned char __user *data, int optlen)
525 {
526         struct ip_options *opt = ip_options_get_alloc(optlen);
527
528         if (!opt)
529                 return -ENOMEM;
530         if (optlen && copy_from_user(opt->__data, data, optlen)) {
531                 kfree(opt);
532                 return -EFAULT;
533         }
534         return ip_options_get_finish(net, optp, opt, optlen);
535 }
536
537 int ip_options_get(struct net *net, struct ip_options **optp,
538                    unsigned char *data, int optlen)
539 {
540         struct ip_options *opt = ip_options_get_alloc(optlen);
541
542         if (!opt)
543                 return -ENOMEM;
544         if (optlen)
545                 memcpy(opt->__data, data, optlen);
546         return ip_options_get_finish(net, optp, opt, optlen);
547 }
548
549 void ip_forward_options(struct sk_buff *skb)
550 {
551         struct   ip_options * opt       = &(IPCB(skb)->opt);
552         unsigned char * optptr;
553         struct rtable *rt = skb->rtable;
554         unsigned char *raw = skb_network_header(skb);
555
556         if (opt->rr_needaddr) {
557                 optptr = (unsigned char *)raw + opt->rr;
558                 ip_rt_get_source(&optptr[optptr[2]-5], rt);
559                 opt->is_changed = 1;
560         }
561         if (opt->srr_is_hit) {
562                 int srrptr, srrspace;
563
564                 optptr = raw + opt->srr;
565
566                 for ( srrptr=optptr[2], srrspace = optptr[1];
567                      srrptr <= srrspace;
568                      srrptr += 4
569                      ) {
570                         if (srrptr + 3 > srrspace)
571                                 break;
572                         if (memcmp(&rt->rt_dst, &optptr[srrptr-1], 4) == 0)
573                                 break;
574                 }
575                 if (srrptr + 3 <= srrspace) {
576                         opt->is_changed = 1;
577                         ip_rt_get_source(&optptr[srrptr-1], rt);
578                         ip_hdr(skb)->daddr = rt->rt_dst;
579                         optptr[2] = srrptr+4;
580                 } else if (net_ratelimit())
581                         printk(KERN_CRIT "ip_forward(): Argh! Destination lost!\n");
582                 if (opt->ts_needaddr) {
583                         optptr = raw + opt->ts;
584                         ip_rt_get_source(&optptr[optptr[2]-9], rt);
585                         opt->is_changed = 1;
586                 }
587         }
588         if (opt->is_changed) {
589                 opt->is_changed = 0;
590                 ip_send_check(ip_hdr(skb));
591         }
592 }
593
594 int ip_options_rcv_srr(struct sk_buff *skb)
595 {
596         struct ip_options *opt = &(IPCB(skb)->opt);
597         int srrspace, srrptr;
598         __be32 nexthop;
599         struct iphdr *iph = ip_hdr(skb);
600         unsigned char *optptr = skb_network_header(skb) + opt->srr;
601         struct rtable *rt = skb->rtable;
602         struct rtable *rt2;
603         int err;
604
605         if (!opt->srr)
606                 return 0;
607
608         if (skb->pkt_type != PACKET_HOST)
609                 return -EINVAL;
610         if (rt->rt_type == RTN_UNICAST) {
611                 if (!opt->is_strictroute)
612                         return 0;
613                 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
614                 return -EINVAL;
615         }
616         if (rt->rt_type != RTN_LOCAL)
617                 return -EINVAL;
618
619         for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
620                 if (srrptr + 3 > srrspace) {
621                         icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
622                         return -EINVAL;
623                 }
624                 memcpy(&nexthop, &optptr[srrptr-1], 4);
625
626                 rt = skb->rtable;
627                 skb->rtable = NULL;
628                 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
629                 rt2 = skb->rtable;
630                 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
631                         ip_rt_put(rt2);
632                         skb->rtable = rt;
633                         return -EINVAL;
634                 }
635                 ip_rt_put(rt);
636                 if (rt2->rt_type != RTN_LOCAL)
637                         break;
638                 /* Superfast 8) loopback forward */
639                 memcpy(&iph->daddr, &optptr[srrptr-1], 4);
640                 opt->is_changed = 1;
641         }
642         if (srrptr <= srrspace) {
643                 opt->srr_is_hit = 1;
644                 opt->is_changed = 1;
645         }
646         return 0;
647 }