[IPV4]: Add 'rtable' field in struct sk_buff to alias 'dst' and avoid casts
[linux-2.6.git] / 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  * 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
14 #include <linux/capability.h>
15 #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>
26 #include <net/route.h>
27 #include <net/cipso_ipv4.h>
28
29 /*
30  * Write options to IP header, record destination address to
31  * source route option, address of outgoing interface
32  * (we should already know it, so that this  function is allowed be
33  * called only after routing decision) and timestamp,
34  * if we originate this datagram.
35  *
36  * daddr is real destination address, next hop is recorded in IP header.
37  * saddr is address of outgoing interface.
38  */
39
40 void ip_options_build(struct sk_buff * skb, struct ip_options * opt,
41                             __be32 daddr, struct rtable *rt, int is_frag)
42 {
43         unsigned char *iph = skb_network_header(skb);
44
45         memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
46         memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
47         opt = &(IPCB(skb)->opt);
48         opt->is_data = 0;
49
50         if (opt->srr)
51                 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
52
53         if (!is_frag) {
54                 if (opt->rr_needaddr)
55                         ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, rt);
56                 if (opt->ts_needaddr)
57                         ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, rt);
58                 if (opt->ts_needtime) {
59                         struct timeval tv;
60                         __be32 midtime;
61                         do_gettimeofday(&tv);
62                         midtime = htonl((tv.tv_sec % 86400) * 1000 + tv.tv_usec / 1000);
63                         memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
64                 }
65                 return;
66         }
67         if (opt->rr) {
68                 memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
69                 opt->rr = 0;
70                 opt->rr_needaddr = 0;
71         }
72         if (opt->ts) {
73                 memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
74                 opt->ts = 0;
75                 opt->ts_needaddr = opt->ts_needtime = 0;
76         }
77 }
78
79 /*
80  * Provided (sopt, skb) points to received options,
81  * build in dopt compiled option set appropriate for answering.
82  * i.e. invert SRR option, copy anothers,
83  * and grab room in RR/TS options.
84  *
85  * NOTE: dopt cannot point to skb.
86  */
87
88 int ip_options_echo(struct ip_options * dopt, struct sk_buff * skb)
89 {
90         struct ip_options *sopt;
91         unsigned char *sptr, *dptr;
92         int soffset, doffset;
93         int     optlen;
94         __be32  daddr;
95
96         memset(dopt, 0, sizeof(struct ip_options));
97
98         dopt->is_data = 1;
99
100         sopt = &(IPCB(skb)->opt);
101
102         if (sopt->optlen == 0) {
103                 dopt->optlen = 0;
104                 return 0;
105         }
106
107         sptr = skb_network_header(skb);
108         dptr = dopt->__data;
109
110         daddr = skb->rtable->rt_spec_dst;
111
112         if (sopt->rr) {
113                 optlen  = sptr[sopt->rr+1];
114                 soffset = sptr[sopt->rr+2];
115                 dopt->rr = dopt->optlen + sizeof(struct iphdr);
116                 memcpy(dptr, sptr+sopt->rr, optlen);
117                 if (sopt->rr_needaddr && soffset <= optlen) {
118                         if (soffset + 3 > optlen)
119                                 return -EINVAL;
120                         dptr[2] = soffset + 4;
121                         dopt->rr_needaddr = 1;
122                 }
123                 dptr += optlen;
124                 dopt->optlen += optlen;
125         }
126         if (sopt->ts) {
127                 optlen = sptr[sopt->ts+1];
128                 soffset = sptr[sopt->ts+2];
129                 dopt->ts = dopt->optlen + sizeof(struct iphdr);
130                 memcpy(dptr, sptr+sopt->ts, optlen);
131                 if (soffset <= optlen) {
132                         if (sopt->ts_needaddr) {
133                                 if (soffset + 3 > optlen)
134                                         return -EINVAL;
135                                 dopt->ts_needaddr = 1;
136                                 soffset += 4;
137                         }
138                         if (sopt->ts_needtime) {
139                                 if (soffset + 3 > optlen)
140                                         return -EINVAL;
141                                 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
142                                         dopt->ts_needtime = 1;
143                                         soffset += 4;
144                                 } else {
145                                         dopt->ts_needtime = 0;
146
147                                         if (soffset + 8 <= optlen) {
148                                                 __be32 addr;
149
150                                                 memcpy(&addr, sptr+soffset-1, 4);
151                                                 if (inet_addr_type(&init_net, addr) != RTN_LOCAL) {
152                                                         dopt->ts_needtime = 1;
153                                                         soffset += 8;
154                                                 }
155                                         }
156                                 }
157                         }
158                         dptr[2] = soffset;
159                 }
160                 dptr += optlen;
161                 dopt->optlen += optlen;
162         }
163         if (sopt->srr) {
164                 unsigned char * start = sptr+sopt->srr;
165                 __be32 faddr;
166
167                 optlen  = start[1];
168                 soffset = start[2];
169                 doffset = 0;
170                 if (soffset > optlen)
171                         soffset = optlen + 1;
172                 soffset -= 4;
173                 if (soffset > 3) {
174                         memcpy(&faddr, &start[soffset-1], 4);
175                         for (soffset-=4, doffset=4; soffset > 3; soffset-=4, doffset+=4)
176                                 memcpy(&dptr[doffset-1], &start[soffset-1], 4);
177                         /*
178                          * RFC1812 requires to fix illegal source routes.
179                          */
180                         if (memcmp(&ip_hdr(skb)->saddr,
181                                    &start[soffset + 3], 4) == 0)
182                                 doffset -= 4;
183                 }
184                 if (doffset > 3) {
185                         memcpy(&start[doffset-1], &daddr, 4);
186                         dopt->faddr = faddr;
187                         dptr[0] = start[0];
188                         dptr[1] = doffset+3;
189                         dptr[2] = 4;
190                         dptr += doffset+3;
191                         dopt->srr = dopt->optlen + sizeof(struct iphdr);
192                         dopt->optlen += doffset+3;
193                         dopt->is_strictroute = sopt->is_strictroute;
194                 }
195         }
196         if (sopt->cipso) {
197                 optlen  = sptr[sopt->cipso+1];
198                 dopt->cipso = dopt->optlen+sizeof(struct iphdr);
199                 memcpy(dptr, sptr+sopt->cipso, optlen);
200                 dptr += optlen;
201                 dopt->optlen += optlen;
202         }
203         while (dopt->optlen & 3) {
204                 *dptr++ = IPOPT_END;
205                 dopt->optlen++;
206         }
207         return 0;
208 }
209
210 /*
211  *      Options "fragmenting", just fill options not
212  *      allowed in fragments with NOOPs.
213  *      Simple and stupid 8), but the most efficient way.
214  */
215
216 void ip_options_fragment(struct sk_buff * skb)
217 {
218         unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
219         struct ip_options * opt = &(IPCB(skb)->opt);
220         int  l = opt->optlen;
221         int  optlen;
222
223         while (l > 0) {
224                 switch (*optptr) {
225                 case IPOPT_END:
226                         return;
227                 case IPOPT_NOOP:
228                         l--;
229                         optptr++;
230                         continue;
231                 }
232                 optlen = optptr[1];
233                 if (optlen<2 || optlen>l)
234                   return;
235                 if (!IPOPT_COPIED(*optptr))
236                         memset(optptr, IPOPT_NOOP, optlen);
237                 l -= optlen;
238                 optptr += optlen;
239         }
240         opt->ts = 0;
241         opt->rr = 0;
242         opt->rr_needaddr = 0;
243         opt->ts_needaddr = 0;
244         opt->ts_needtime = 0;
245         return;
246 }
247
248 /*
249  * Verify options and fill pointers in struct options.
250  * Caller should clear *opt, and set opt->data.
251  * If opt == NULL, then skb->data should point to IP header.
252  */
253
254 int ip_options_compile(struct ip_options * opt, struct sk_buff * skb)
255 {
256         int l;
257         unsigned char * iph;
258         unsigned char * optptr;
259         int optlen;
260         unsigned char * pp_ptr = NULL;
261         struct rtable *rt = skb ? skb->rtable : NULL;
262
263         if (!opt) {
264                 opt = &(IPCB(skb)->opt);
265                 iph = skb_network_header(skb);
266                 opt->optlen = ((struct iphdr *)iph)->ihl*4 - sizeof(struct iphdr);
267                 optptr = iph + sizeof(struct iphdr);
268                 opt->is_data = 0;
269         } else {
270                 optptr = opt->is_data ? opt->__data :
271                                         (unsigned char *)&(ip_hdr(skb)[1]);
272                 iph = optptr - sizeof(struct iphdr);
273         }
274
275         for (l = opt->optlen; l > 0; ) {
276                 switch (*optptr) {
277                       case IPOPT_END:
278                         for (optptr++, l--; l>0; optptr++, l--) {
279                                 if (*optptr != IPOPT_END) {
280                                         *optptr = IPOPT_END;
281                                         opt->is_changed = 1;
282                                 }
283                         }
284                         goto eol;
285                       case IPOPT_NOOP:
286                         l--;
287                         optptr++;
288                         continue;
289                 }
290                 optlen = optptr[1];
291                 if (optlen<2 || optlen>l) {
292                         pp_ptr = optptr;
293                         goto error;
294                 }
295                 switch (*optptr) {
296                       case IPOPT_SSRR:
297                       case IPOPT_LSRR:
298                         if (optlen < 3) {
299                                 pp_ptr = optptr + 1;
300                                 goto error;
301                         }
302                         if (optptr[2] < 4) {
303                                 pp_ptr = optptr + 2;
304                                 goto error;
305                         }
306                         /* NB: cf RFC-1812 5.2.4.1 */
307                         if (opt->srr) {
308                                 pp_ptr = optptr;
309                                 goto error;
310                         }
311                         if (!skb) {
312                                 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
313                                         pp_ptr = optptr + 1;
314                                         goto error;
315                                 }
316                                 memcpy(&opt->faddr, &optptr[3], 4);
317                                 if (optlen > 7)
318                                         memmove(&optptr[3], &optptr[7], optlen-7);
319                         }
320                         opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
321                         opt->srr = optptr - iph;
322                         break;
323                       case IPOPT_RR:
324                         if (opt->rr) {
325                                 pp_ptr = optptr;
326                                 goto error;
327                         }
328                         if (optlen < 3) {
329                                 pp_ptr = optptr + 1;
330                                 goto error;
331                         }
332                         if (optptr[2] < 4) {
333                                 pp_ptr = optptr + 2;
334                                 goto error;
335                         }
336                         if (optptr[2] <= optlen) {
337                                 if (optptr[2]+3 > optlen) {
338                                         pp_ptr = optptr + 2;
339                                         goto error;
340                                 }
341                                 if (skb) {
342                                         memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
343                                         opt->is_changed = 1;
344                                 }
345                                 optptr[2] += 4;
346                                 opt->rr_needaddr = 1;
347                         }
348                         opt->rr = optptr - iph;
349                         break;
350                       case IPOPT_TIMESTAMP:
351                         if (opt->ts) {
352                                 pp_ptr = optptr;
353                                 goto error;
354                         }
355                         if (optlen < 4) {
356                                 pp_ptr = optptr + 1;
357                                 goto error;
358                         }
359                         if (optptr[2] < 5) {
360                                 pp_ptr = optptr + 2;
361                                 goto error;
362                         }
363                         if (optptr[2] <= optlen) {
364                                 __be32 *timeptr = NULL;
365                                 if (optptr[2]+3 > optptr[1]) {
366                                         pp_ptr = optptr + 2;
367                                         goto error;
368                                 }
369                                 switch (optptr[3]&0xF) {
370                                       case IPOPT_TS_TSONLY:
371                                         opt->ts = optptr - iph;
372                                         if (skb)
373                                                 timeptr = (__be32*)&optptr[optptr[2]-1];
374                                         opt->ts_needtime = 1;
375                                         optptr[2] += 4;
376                                         break;
377                                       case IPOPT_TS_TSANDADDR:
378                                         if (optptr[2]+7 > optptr[1]) {
379                                                 pp_ptr = optptr + 2;
380                                                 goto error;
381                                         }
382                                         opt->ts = optptr - iph;
383                                         if (skb) {
384                                                 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
385                                                 timeptr = (__be32*)&optptr[optptr[2]+3];
386                                         }
387                                         opt->ts_needaddr = 1;
388                                         opt->ts_needtime = 1;
389                                         optptr[2] += 8;
390                                         break;
391                                       case IPOPT_TS_PRESPEC:
392                                         if (optptr[2]+7 > optptr[1]) {
393                                                 pp_ptr = optptr + 2;
394                                                 goto error;
395                                         }
396                                         opt->ts = optptr - iph;
397                                         {
398                                                 __be32 addr;
399                                                 memcpy(&addr, &optptr[optptr[2]-1], 4);
400                                                 if (inet_addr_type(&init_net, addr) == RTN_UNICAST)
401                                                         break;
402                                                 if (skb)
403                                                         timeptr = (__be32*)&optptr[optptr[2]+3];
404                                         }
405                                         opt->ts_needtime = 1;
406                                         optptr[2] += 8;
407                                         break;
408                                       default:
409                                         if (!skb && !capable(CAP_NET_RAW)) {
410                                                 pp_ptr = optptr + 3;
411                                                 goto error;
412                                         }
413                                         break;
414                                 }
415                                 if (timeptr) {
416                                         struct timeval tv;
417                                         __be32  midtime;
418                                         do_gettimeofday(&tv);
419                                         midtime = htonl((tv.tv_sec % 86400) * 1000 + tv.tv_usec / 1000);
420                                         memcpy(timeptr, &midtime, sizeof(__be32));
421                                         opt->is_changed = 1;
422                                 }
423                         } else {
424                                 unsigned overflow = optptr[3]>>4;
425                                 if (overflow == 15) {
426                                         pp_ptr = optptr + 3;
427                                         goto error;
428                                 }
429                                 opt->ts = optptr - iph;
430                                 if (skb) {
431                                         optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
432                                         opt->is_changed = 1;
433                                 }
434                         }
435                         break;
436                       case IPOPT_RA:
437                         if (optlen < 4) {
438                                 pp_ptr = optptr + 1;
439                                 goto error;
440                         }
441                         if (optptr[2] == 0 && optptr[3] == 0)
442                                 opt->router_alert = optptr - iph;
443                         break;
444                       case IPOPT_CIPSO:
445                         if ((!skb && !capable(CAP_NET_RAW)) || opt->cipso) {
446                                 pp_ptr = optptr;
447                                 goto error;
448                         }
449                         opt->cipso = optptr - iph;
450                         if (cipso_v4_validate(&optptr)) {
451                                 pp_ptr = optptr;
452                                 goto error;
453                         }
454                         break;
455                       case IPOPT_SEC:
456                       case IPOPT_SID:
457                       default:
458                         if (!skb && !capable(CAP_NET_RAW)) {
459                                 pp_ptr = optptr;
460                                 goto error;
461                         }
462                         break;
463                 }
464                 l -= optlen;
465                 optptr += optlen;
466         }
467
468 eol:
469         if (!pp_ptr)
470                 return 0;
471
472 error:
473         if (skb) {
474                 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
475         }
476         return -EINVAL;
477 }
478
479
480 /*
481  *      Undo all the changes done by ip_options_compile().
482  */
483
484 void ip_options_undo(struct ip_options * opt)
485 {
486         if (opt->srr) {
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);
490         }
491         if (opt->rr_needaddr) {
492                 unsigned  char * optptr = opt->__data+opt->rr-sizeof(struct  iphdr);
493                 optptr[2] -= 4;
494                 memset(&optptr[optptr[2]-1], 0, 4);
495         }
496         if (opt->ts) {
497                 unsigned  char * optptr = opt->__data+opt->ts-sizeof(struct  iphdr);
498                 if (opt->ts_needtime) {
499                         optptr[2] -= 4;
500                         memset(&optptr[optptr[2]-1], 0, 4);
501                         if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
502                                 optptr[2] -= 4;
503                 }
504                 if (opt->ts_needaddr) {
505                         optptr[2] -= 4;
506                         memset(&optptr[optptr[2]-1], 0, 4);
507                 }
508         }
509 }
510
511 static struct ip_options *ip_options_get_alloc(const int optlen)
512 {
513         return kzalloc(sizeof(struct ip_options) + ((optlen + 3) & ~3),
514                        GFP_KERNEL);
515 }
516
517 static int ip_options_get_finish(struct ip_options **optp,
518                                  struct ip_options *opt, int optlen)
519 {
520         while (optlen & 3)
521                 opt->__data[optlen++] = IPOPT_END;
522         opt->optlen = optlen;
523         opt->is_data = 1;
524         if (optlen && ip_options_compile(opt, NULL)) {
525                 kfree(opt);
526                 return -EINVAL;
527         }
528         kfree(*optp);
529         *optp = opt;
530         return 0;
531 }
532
533 int ip_options_get_from_user(struct ip_options **optp, unsigned char __user *data, int optlen)
534 {
535         struct ip_options *opt = ip_options_get_alloc(optlen);
536
537         if (!opt)
538                 return -ENOMEM;
539         if (optlen && copy_from_user(opt->__data, data, optlen)) {
540                 kfree(opt);
541                 return -EFAULT;
542         }
543         return ip_options_get_finish(optp, opt, optlen);
544 }
545
546 int ip_options_get(struct ip_options **optp, unsigned char *data, int optlen)
547 {
548         struct ip_options *opt = ip_options_get_alloc(optlen);
549
550         if (!opt)
551                 return -ENOMEM;
552         if (optlen)
553                 memcpy(opt->__data, data, optlen);
554         return ip_options_get_finish(optp, opt, optlen);
555 }
556
557 void ip_forward_options(struct sk_buff *skb)
558 {
559         struct   ip_options * opt       = &(IPCB(skb)->opt);
560         unsigned char * optptr;
561         struct rtable *rt = skb->rtable;
562         unsigned char *raw = skb_network_header(skb);
563
564         if (opt->rr_needaddr) {
565                 optptr = (unsigned char *)raw + opt->rr;
566                 ip_rt_get_source(&optptr[optptr[2]-5], rt);
567                 opt->is_changed = 1;
568         }
569         if (opt->srr_is_hit) {
570                 int srrptr, srrspace;
571
572                 optptr = raw + opt->srr;
573
574                 for ( srrptr=optptr[2], srrspace = optptr[1];
575                      srrptr <= srrspace;
576                      srrptr += 4
577                      ) {
578                         if (srrptr + 3 > srrspace)
579                                 break;
580                         if (memcmp(&rt->rt_dst, &optptr[srrptr-1], 4) == 0)
581                                 break;
582                 }
583                 if (srrptr + 3 <= srrspace) {
584                         opt->is_changed = 1;
585                         ip_rt_get_source(&optptr[srrptr-1], rt);
586                         ip_hdr(skb)->daddr = rt->rt_dst;
587                         optptr[2] = srrptr+4;
588                 } else if (net_ratelimit())
589                         printk(KERN_CRIT "ip_forward(): Argh! Destination lost!\n");
590                 if (opt->ts_needaddr) {
591                         optptr = raw + opt->ts;
592                         ip_rt_get_source(&optptr[optptr[2]-9], rt);
593                         opt->is_changed = 1;
594                 }
595         }
596         if (opt->is_changed) {
597                 opt->is_changed = 0;
598                 ip_send_check(ip_hdr(skb));
599         }
600 }
601
602 int ip_options_rcv_srr(struct sk_buff *skb)
603 {
604         struct ip_options *opt = &(IPCB(skb)->opt);
605         int srrspace, srrptr;
606         __be32 nexthop;
607         struct iphdr *iph = ip_hdr(skb);
608         unsigned char *optptr = skb_network_header(skb) + opt->srr;
609         struct rtable *rt = skb->rtable;
610         struct rtable *rt2;
611         int err;
612
613         if (!opt->srr)
614                 return 0;
615
616         if (skb->pkt_type != PACKET_HOST)
617                 return -EINVAL;
618         if (rt->rt_type == RTN_UNICAST) {
619                 if (!opt->is_strictroute)
620                         return 0;
621                 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
622                 return -EINVAL;
623         }
624         if (rt->rt_type != RTN_LOCAL)
625                 return -EINVAL;
626
627         for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
628                 if (srrptr + 3 > srrspace) {
629                         icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
630                         return -EINVAL;
631                 }
632                 memcpy(&nexthop, &optptr[srrptr-1], 4);
633
634                 rt = skb->rtable;
635                 skb->rtable = NULL;
636                 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
637                 rt2 = skb->rtable;
638                 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
639                         ip_rt_put(rt2);
640                         skb->rtable = rt;
641                         return -EINVAL;
642                 }
643                 ip_rt_put(rt);
644                 if (rt2->rt_type != RTN_LOCAL)
645                         break;
646                 /* Superfast 8) loopback forward */
647                 memcpy(&iph->daddr, &optptr[srrptr-1], 4);
648                 opt->is_changed = 1;
649         }
650         if (srrptr <= srrspace) {
651                 opt->srr_is_hit = 1;
652                 opt->is_changed = 1;
653         }
654         return 0;
655 }